Chinese Version Light学术出版中心
Latest Issue
Issue 10s ， 2017
-
Design and analysis of concave grating concentric spectrometer with shifting object and image plane
Abstract：The object plane and image plane of concave grating concentric spectrometers are in the same plane, and the vertical axis distance between object and image is only a few millimeters, which limits the arrangement of focal plane detectors and front lens. In order to solve this problem, a plane mirror was introduced into the optical system to deflect the light, and the separation of the object surface and the image surface was realized. According to the deflection light, two improved concave grating concentric spectrometers with shift object plane or shift image plane were designed and then simulated by simulation software. The simulation results show that the two spectrometers can achieve the separation of object surface and image plane, the separation distance is greater than 16 mm, and the aberrations are also well controlled, within the 400-800 nm the aberration curve is almost zero. Those results meet the application requirements of low aberration, high image quality, image plane separation, which can effectively solve the problem of device arrangement.Keywords：optical design;optical grating spectrometer;Concave grating;concentric;shift object plane;shift image plane- 31
- |
- 7
- |
- 0
Published：2020-08-13 -
Analyzing microfiltration process of material under laser loading by ultrahigh-speed framing photography
Abstract：As a research focus, the micro-capsule fracture of the material under strong shock wave loading plays an important role in the field of defense. Herein, selecting high purity aluminum, high purity tin and helium bubble aluminum as the research object, an experimental platform was built by using self developed ultra high-speed photoelectric framing camera and a sophisticated diagnostic test technique of high speed photograph was established. The micro-filtration process of the material under high energy laser loading was studied, and a clear image of the broken process in 107~108/s was captured for analyzing the experimental phenomena. The results show that the measurement technique established in this paper can be employed to study crushing characteristics of the material, which provides effective experimental data for the further study of micro fracture of the material.Keywords：photoelectronic imaging;high speed photography;microfiltration;dynamic fragmentation;microfiltration process- 71
- |
- 10
- |
- 0
Published：2020-08-13 -
Autofocus system for invisible pulsed laser
Abstract：It is more difficult for invisible laser to position locating and local autofocus in the invisible pulse laser processing system. In this paper, a useful laser processing system was introduced, and the methods of the position locating and autofocus were proposed based on the system. First, the image of the interaction between one pulse laser and the sample surface was captured by CCD. Then, the image was turned into binary image by a proper threshold value. Extracting the area of the light spot, the coordinate of the subpixel of the centroid of the focused laser spot was calculated by the gray centroid method. Meanwhile several operators were employed to evaluate the sharpness of the image in time domain, thus analyzing different components of the imagein the operators. Subsequently, the autofocus algorithm was presented based on the evaluation of green component of the sample surface with Sobel gradient operator.Furthermore, the local autofocus algorithm was proposed based on the local image centered with the focused laser spot rather than the whole image to reduce the error of the global autofocus algorithm. Experimental results show that the precision of the position locating algorithm is 1 pixeland the accuracy of the autofocus algorithm is 1 μm, which indicatethat the proposed autofocus methodcan improve the performance of laser processing systems to meet the increasing need of the laser processing.Keywords：laser processing;invisible pulsed laser;position locating;autofocus;local autofocus;computer vision- 77
- |
- 11
- |
- 0
Published：2020-08-13 -
Theoretical analysis and numerical simulation for wavelength switchable Tm:YAG laser modulated by Tm:YAG crystal length
Abstract：Based on the analysis of quasi-three-level side pumped Tm:YAG laser system, the dependence of the laser wavelength was studied theoretically and experimentally, which shows central wavelength is switchable between 2.07 and 2.02 μm with different Tm:YAG crystal lengths or rod temperatures. The length of the laser medium corresponding to the wavelength red shift was calculated quantitatively. The laser oscillation at 2.02 μm with larger stimulated emission sections is suppressed when the crystal length is larger than~85 mm with a 5% output coupling, then an efficient side-diode-pumped rod Tm:YAG laser operating at 2.07 μm is realized. The experimental results show that the wavelength of the single-module Tm:YAG laser with Tm:YAG crystal length of 69 mm is located at 2.02 μm, while the wavelength of two-modules Tm:YAG laser with the total Tm:YAG crystal length of 138 mm is located at 2.07 μm. The experiments confirm the accuracy of the model.Keywords：quasi-three-level;pump threshold;wavelength switchable;length of laser medium- 56
- |
- 10
- |
- 0
Published：2020-08-13 -
Field inspection of rhodamine B in prickly ash by laser Raman
Abstract：Prickly ash, dry pepper, chili sauce and other condiment are widely used in food industry and daily life. However these condiments are illegally added with rhodamine B, alkaline orange Ⅱ and other harzardous additives, which seriously affect human health. In order to inspect these illegal additives with low-cost and high-efficiency, a field detection system for rhodamine B in prickly ash was designed based on the principle of laser Raman. It mainly consisted of laser, spectrometer, laser probe, vacuum pump, analysis software and portable PC. A portable spectrometer based on Raman principle was achieved. Moreover, two baseline removal algorithms including piecewise linear fitting method as well as fault polynomial fitting method and a result matching algorithm were proposed. The samples were pretreated with silver ion Raman enhancer to obtain the corresponding Raman signals, which were used to match the results. The experiments prove that this method can accurately detect 1×10-6 rhodamine B in prickly ash sample with whole inspection time less than 10 min, which can basically meet the requirements of field rapid inspection of illegal additives in condiments.Keywords：laser Raman;Rhodamine B;field inspection;baseline removal algorithm;sample pretreatment- 60
- |
- 12
- |
- 0
Published：2020-08-13 -
Optimization of correlation algorithms for small array pixels tracking sensors
Abstract：In order to improve the tracking accuracy of irregular targets in compound axis tracking system,the correlation algorithm was introduced into array photon counters to process the slope error. Similar to the correlation algorithms used in conventional CCD, part of a frame of a array photon counter was used as a template. After the template was symmetrically optimized, the template was matched on the frame.With the SAD algorithm, the position of the template on the measured frame was obtained by fitting, and the offset of light spot was calculated. Furthermore, the offset was corrected by the centroid location of light spots on the template to achieve stable tracking. The experimental results show that the modified correlation algorithm can achieve stable tracking, and the symmetrical optimization of the template can also improve the fitting accuracy.With optimized template, the measurement error of the SAD algorithm is 0.018 pixel, and the random error is 0.046 pixel. Employing the correlation algorithm to small array detectors, the stable tracking can be achieved by choosing appropriate size of the template and correcting the template.Keywords：adaptive optics;correlation algorithm;wavefront detection;parabola fitting;sub-pixel- 59
- |
- 15
- |
- 0
Published：2020-08-13 -
Study on growth characteristics of microalgae biofilm based on hyperspectral imaging
Abstract：To obtain the optimal environmental conditions for microalgae biofilm culture, the spectral information of the biofilms under different culture conditions including light intensity, pH value and temperature, were monitored online and in real time, using hyperspectral imaging technology. The characteristic variables of biofilm spectral characteristics were extracted, and the correlation coefficient between the characteristic variables and dry weight of the biofilms was calculated. Furthermore, the characteristic variables of the biofilms were normalized based on the quadratic function and were combined weight by using the exponential function, thus establishing the prediction model of biofilm growth process. Then the biofilm growth process was predicted by the established model. The results show that the optimal culture conditions of the biofilm were the temperature of 28℃, light intensity of 3 500 lx, and PH of 8; the mature biofilm, which will no longer grow, can be obtained under the optimized culture conditions after 7 days in culture. The growth characteristics of biofilm can be quickly and non-destructively monitored by the hyperspectral imaging method.Keywords：microalgae biofilm;growth characteristics;hyperspectral;characteristic spectra;prediction- 76
- |
- 13
- |
- 0
Published：2020-08-13 -
Center of gravity testing for giant steerable science mirror
Abstract：The tertiary mirror of Thirty Meter Telescope (TMT) is a Giant Steerable Science Mirror (GSSM), the center of gravity (CG) of which significantly influences the control accuracy of the telescope. However, the GSSM is so large that the traditional lying down method is not suitable for the CG measurement. To measure the CG of the GSSM at small tipping angles, a method for testing the CG with a laser tracker and load cells was proposed and verified theoretically and experimentally. In the experiment, a 1/4 scale prototype of the TMT was constructed and its CG was located. Furthermore, the error of the measuring result was analyzed considering the laser tracker error and equipment error, whereas the gravity sag of GSSMP was ignored. The mass of GSSMP is 805.800 kg, and the location of CG is (-7.169 mm,12.900 mm,526 mm) with an accuracy of 0.501 mm.Keywords：Thirty Meter Telescope(TMT);tertiary mirror;center of gravity testing;error analysis;laser tracker- 48
- |
- 8
- |
- 0
Published：2020-08-13 -
Lidar simultaneous localization and mapping with low-drift
Abstract：As conventional 2D lidars can't reconstruct three-dimensional map in real time without other sensors, this paper constructed a three-dimensional map from point clouds perceived by a 2D laser lidar scanning in six degrees of freedom. First, the location of radar was taken as the origin of world coordinate system at the begging of the program. The laser spot scanned by the radar was combined with the rotation angle of the motor so that the laser point has three-dimensional coordinate information, which was registrated into the radar coordinate system. The bilateral filtering algorithm was adopted to filter the point cloud, which was then processed by lidar ranging algorithms to calculate the motion of the lidar between two successive scans. The estimated motion was used to correct the distortion of the point cloud. The drawing algorithm further processes the output of the ranging algorithm at a lower frequency, fine-graining the corrected cloud to the map. Finally, the two algorithms were integrated to map and post the radar posture transformation in real time, thus outputting the position of lidar in world coordinate system. Experimental results show that the method can accurately estimate the trajectory of the lidar with six degrees of freedom in real time based on the laser radar original data and the motor position information, and meanwhile construct high-quality 3D point cloud image. The proposed method exhibits good real-time performance, and the relative precision of the indoor test is approximately 2%.Keywords：lidar;three-dimensional mapping;Simultaneous Localization and Mapping(SLAM);posture transform;bilateral filtering- 61
- |
- 18
- |
- 0
Published：2020-08-13 -
Hartmann wavefront sensor for improving SNR of centroid measurement under non-uniform illumination using zone-sharing exposure
Abstract：In order to reduce the influence of non-uniform illumination on the wavefront measurement accuracy of Hartmann wavefront sensor, the absolute calibration principle of Hartmann wavefront sensor as well as the relationship between the centroid measurement error and the signal light energy were analyzed, then a method for zone-sharing exposure of Hartmann wavefront sensor was proposed. Under the condition of a static wavefront, the signal-to-noise ratio of centroid measurement in a single subaperture was improved by setting different exposure time for different subaperture regions of the Hartmann wavefront sensor. Finally, the 35×35 sub apertures used for absolute calibration of the Hartmann wavefront sensor was exposed as zone-sharing exposure. Simulation results show that the centroid measurement error within a single sub aperture is reduced to 0.1 pixel. The method of divisional exposure can effectively reduce the centroid measurement error of the spot in non-uniform illumination, thus improving the accuracy of the Hartmann wavefront sensor in non-uniform illumination.Keywords：adaptive optics;wavefront measurement;centroid detection;zone-sharing exposure;Hartmann wavefront sensors- 55
- |
- 16
- |
- 0
Published：2020-08-13 -
Design of optical system with wide spectrum and large field of view
Abstract：An optical System with wide spectrum and large field of view was designed to meet the needs of the large detection range of space cameras. For further study of the structure, performance index and imaging quality of the optical system, three parallel optical tubes were used for ideal image plane mosaic with one image, so as to realize simulation detection and image scanning of infinite targets of the camera. Then the scanning image was compared with the simulated image, thus achieving dynamic parameter detection of the camera in laboratories. The simulation and experiment results show that the focal length of the optical system is 100 mm, the F is 27 and the aperture diameter is 25 mm. The target image is clear with high quality, the optical transfer function reaches the diffraction limit and the distortion is better than 1%. The system has important significance for dynamic detection of space cameras and provides technical guarantee for product test and finalization.Keywords：optical system design;deep space exploration;wide spectrum;Image stitching- 88
- |
- 24
- |
- 0
Published：2020-08-13 -
Multi-aperture and large field imaging experimental system based on artificial compound eyes
Abstract：In order to realize artificial compound eyes with small volume, large field of view and sensitive to moving objects, a multi-aperture and large angle imaging experiment system was established, which used a fiber optic panel as deflector optical elements. The imaging unit of this system, the distribution angle of the ommatidia and the performance of the deflector element were analyzed and studied, effectively improving the utilization rate of off-axis light and image quality of the compound eye imaging system. First, according to the whole field angle requirement of the system, nine ommatidia channels were used to calculate the distribution and angle of each imaging channel. Then, the design of single ommatidia channel was optimized to ensure the system field angle and imaging quality. Finally, matching problems in imaging lens, deflector elements and light detection array were analyzed, thus selecting the best deflector program. Experiments show that the multi-channel imaging experiment system with fiber optic panel as the folding element can realize the large-angle imaging function, and the viewing angle can reach 120°. Compared with the traditional single-aperture imaging system, the system can realize the large angle imaging function under a small volume.Keywords：compound eye imaging;optical design;fibre-optic faceplate;large field of view- 123
- |
- 18
- |
- 0
Published：2020-08-13 -
Design of fuzzy clustering system for FTIR spectroscopy of tea
Abstract：In order to classify tea varieties, a fuzzy clustering system was established using Fourier transform infrared spectroscopy (FTIR) technology and its fuzzy clustering algorithms such as fuzzy C-means clustering algorithm (FCM), possibilistic C-means clustering (PCM) and possibilistic fuzzy C-means clustering(PFCM) were investigated in clustering FTIR spectra of tea. Firstly, FTIR spectra of different varieties of tea samples were collected. Secondly, multiplicative scatter correction (MSC) was applied to preprocess the spectra. Thirdly, dimensionality of FTIR spectra was reduced by Principal Component Analysis (PCA) and linear discriminant analysis (LDA). Finally, the spectral data of tea samples were analyzed by fuzzy clustering algorithms. Experimental results indicate that as the weight value m increases, the clustering accuracies of FCM and PFCM clustering algorithms increases remarkably while that of PCM has no obvious change with its accuracy remaining at 75.76% in the whole process. A fuzzy clustering system can be designed to classify tea varieties by FTIR technology coupled with MSC, PCA, LDA, FCM or PFCM with high detection speed and accuracy.Keywords：Fourier transform infrared spectroscopy;fuzzy clustering;Principal Component Analysis(PCA);linear discriminant analysis;multiplicative scatter correction- 48
- |
- 9
- |
- 0
Published：2020-08-13 -
Non-contact high-precision defect detection and 3D reconstruction of object surface
Abstract：Aimed at the requirement of non-contact and high-precision detection on the surface of optical devices, a set of geometric precision measurement system was constructed by using the principle of spectral confocal to realize flaw detection and 3D reconstruction on the surface of optical devices such as microscope lens and so on.A set of 3D scanning system of object surface was constructed combined with high-precision displacement platform to analyze structure and principle of system after introducing the measurement principle of spectral confocal displacement sensor. A self-adaptive method based on dichotomy was adopted to control sampling rate of system for the problem of inaccurate reading of spectral confocal displacement sensor caused by sampling rate,and overall measurement error of system was analyzed by repeated measurement of the same standard measured mass. Finally, high-precision measurement on the surface of microscope lens was realized. Experimental result shows that the method realizes micro-grade measurement of transparent surfaces, and still has better measurement result in the edges with larger surface gradient change. Maximum error of measurement is 0.624 μm, average error is 0.167 μm and uncertainty of measurement is 0.633 μm.Analysis of morphology and size can be realized for micro defects existing on the surface of lens and accurate 3D model on the surface of lens to be detected can be obtained.Keywords：spectral confocal;3D surface;micro defect;precision measurement- 124
- |
- 16
- |
- 0
Published：2020-08-13 -
Testing and analysis of residual stresses in quenched 7075 aluminum alloy thick plate
Abstract：After solution treatment of 7075 aluminum alloy plates, the residual stresses were respectively investigated in the processes of spraying quenching and immersion quenching with the same medium. Firstly, the finite element model was established for the spray-and immersion-quenching. The blind-hole method and crack compliance method were respectively adopted to measure the residual stresses, thus validating the FEM of quenching processes. The comparison analysis show that the simulated values of the residual stress distribution are in good agreement with the experimental measurement data. Secondly, the influence of plate thickness on the residual stress distribution was revealed for the spray quenching and immersion quenching. Obviously, the quenching cooling rate and plate thickness are critical to residual stresses. The residual stress remains unchanged when the plate thickness is more than 70 mm in the immersion quenching. However, the spray quenching can make the residual stress increase uniformly. In addition, the residual stress can increase with the increment of cooling rate. The residual stress in plates with thickness less than 30 mm is small. The surface compressive stresses, induced by the spray quenching, decrease faster than the interior tensile stresses when the plate thickness varies from 30 mm to 100 mm. Nevertheless the interior tensile stress decreased rarely if the plate thickness is from 30 mm to 60 mm. The difference between spraying quenching and immersion quenching gets smaller and smaller. Moreover, the advantage of spray quenching on immersion quenching is gradually decreased with the increase of plate thickness.Keywords：7075 aluminum alloy;quenching;residual stress;cooling temperature;Finite Element Model(FEM)- 72
- |
- 18
- |
- 0
Published：2020-08-13 -
Design of inductance-capacitance microfluidic oil detection chip
Abstract：An inductance-capacitance microfluidic chip was designed and fabricated to detect various pollutants in hydraulic oil. Herein, the position of microchannel, distance of coils and coil turns were simulated. Experiments for studying the influence of coil turns on pollutants detection signal were performed on chip. The results indicate, the inductance and capacitance amplitude are largest when the microchannel is located at the edge of the coil inner hole. The smaller distance between the two coils, the larger inductance and capacitance amplitude are achieved. When coil turns increase, the inductance amplitude increase and the capacitive amplitude decrease gradually. With the increase of coil turns, the inductance amplitude of iron particle and copper particle increase, while the capacitance amplitude of water droplet and air bubble decrease, and inductance and capacitance signal to noise ratio are significantly reduced. When the number of turns is 20, the signal to noise ratio of iron particle, water droplet and air bubble is approximately 3.23, 8.41 and 7.34 times of that of 60 turns respectively. The signal to noise ratio of copper particle is 0 at 60 turns. When the coil turns of the detection chip is 20, the detection signal to noise ratio of the four pollutants reaches the maximum. The microfluidic chip designed in this paper can detect ferromagnetic, non-ferromagnetic particles, water droplets and air bubbles in oil by using one sensor made up of two planar coils.Keywords：inductance-capacitance microfluidic chip;contamination detection;detection signal amplitude;detection signal to noise ratio- 51
- |
- 8
- |
- 0
Published：2020-08-13 -
Design and tests of shafting structure in giant steerable science mirror prototype system
Abstract：To develop a smooth tracking mechanism for Thirty Meter Telescope (TMT) in the gravity-invariant condition, a 1/4 scale, functionally accurate version of the Giant Steerable Science Mirror (GSSM) prototype was developed for pre-construction. The error sources influencing the rotation and pointing accuracies of the system were investigated. Due to complicated working conditions, a shafting system for the GSSM was designed with angular contact ball bearings as support of the azimuth-altitude system. Then, a detailed finite element model was constructed to verify the performance of this system. Wobble tests were performed. In addition, a repeatability test was designed and carried out to verify the pointing accuracy of the system. The first modal frequency is approximately 35.9 Hz, which meets the design requirement well. Moreover, the wobble error of the altitude axis is approximately 1.91"(PV), while the one for azimuth axis is approximately 2.5"(PV). The repeatability of the GSSM is up to 2.77"(RMS) by the test data. The results indicate that the proposed shafting system is feasible and can achieve the requirements of the TMT.Keywords：Thirty Meter Telescope(TMT);alt-azimuth shafting;Wobble error;harmonic theory;pointing repeatability- 67
- |
- 11
- |
- 0
Published：2020-08-13 -
Simulation tool for complex-engineered micro-optical electromechanical system
Abstract：In order to improve the design and analysis ability of complex microsystems with multi-physical domain coupling, an efficient design and modeling method based on system-level structure was presented. In comparison, it was difficult to simulate such kind of microsystem using conventional modeling tools. Such tools were either based on distributed-element analysis, which was time-consuming and too computationally intensive for complex systems, or based on traditional lumped-element analysis, which was limited by an inability to handle algebraic constraints in general. For design, a novel graphical user interface (GUI) was presented. It allowed users to quickly configure complex systems in 3D using a computer mouse or pen at a faster rate than it might be drawn with pencil and paper. The GUI was coupled to a novel and powerful netlist language for design flexibility. For modeling, recent advances in analytical system dynamics and differential-algebraic equations (DAEs) were applied into a framework that facilitates the systematic modeling of multidisciplinary systems that may comprise static or dynamic constraints. For a test case, a complex microsystem fabricated by Sandia National Laboratory (SNL) was efficiently modeled and simulated. This micro optics electro mechanical system comprises mirrors, gears, hinges, a slider, electronic components, comb drives, and electromechanical flexures. The results prove the effectiveness of the method, which presents the netlist language and modeling methodology are beneficial to configure and simulate one of the most complex-engineered micro systems to date.Keywords：modeling;simulation;DAE;system level;lumped analysis;micro system- 80
- |
- 13
- |
- 0
Published：2020-08-13 -
Improvement of output characteristic of oil detection chip
Abstract：The wear condition of lubrication system can be obtained through the lubrication oil, therefore metal particle is a very important parameter. Accurate detection of the metallic wear particle is vital to avoid catastrophic failure of rotating or reciprocating machine. To improve the sensitivity of inductive microfluidic detection chip, an oil detection microfluidic chip was improved. The structure of the chip was designed, and the electromagnetic model of one tilted planar coil was established by the Maxwell function. Then the experimental verification was performed. The experiment results show that the basic inductance is 7.894 15×10-6 H and the average inductance variation is 7.895 08×10-6 H. The inductance variation of the copper particle is up to 9.3×10-10 H, in comparing with 7.008 33×10-10 H in planar coil, with a 32.67% improvement in sensitivity.Keywords：microfluidic chip;inductive sensor;oil detection;metallic wear debris- 83
- |
- 10
- |
- 0
Published：2020-08-13 -
Effect of laser welding parameters on microstucture and mechanical properties of C-276/304 overlapping weld
Abstract：In order to realizereliable seal welding at the end of CAP1400 nuclear reactor coolant pimp can, Nd:YAG pulsed laser was adopted to implement overlapping weld of dissimilar materials of Hastelloy C-276 and austenite stainless steel 304. Optical microscope and scanning electron microscope were adopted to observe morphology and microstructure of the weld metal. EDS was used to detect elementary composition in weld, electron probe was used to measure element distribution in weld and mechanical property of welded joint was evaluated by mechanical tearing experiment. Result of mechanical tearing experiment indicates that fracture occurs in weld under different plus energy, and it is tension fracture. When plus energy is 3-4 J, mechanical tearing resistance strength of weld reaches the maximum and is about 78 N/mm. Optimized weld parameter scopes are:pulse energy 3-4 J, welding speed 125-200 mm/min, pulse frequency 30 Hz, pulse width 6-8 ms, and defocusing amount -1-0 mm. The result shows that the effect of pulse energy and pulse frequency on joint width of overlapping weld is relatively large, but excessively large pulse energy and pulse frequency and excessively small pulse width will cause collapse and serious spatter of weld surface. Along with increase of pulse energy, width of incomplete fusion zone in the side of 304 parent material decreases, content of precipitated phase in weld is reduced. Under relatively high pulse energy, columnar crystal and equiaxed structure in weld coarsen, quantity and size of micropore caused by element ablation is relatively large. Along with improvement of dilution rate of 304 parent material in weld, difference between element content and 304 parent material in weld is reduced, and uniformity of element distribution in weld decreases.Keywords：laser welding;Hastelloy C-276;dissimilar welding;microstructure;mechanical property- 58
- |
- 8
- |
- 0
Published：2020-08-13 -
Characteristics of double-coil resonant microfluidic chip for oil detection
Abstract：In order to realize the high precision detection of fine metal abrasive grains in hydraulic oil, a double-coil resonant microfluidic detection chip was designed and fabricated. The relationship between the excitation frequency and inductance change of the detection chip and the detection effect of metal abrasive grains were studied. First, according to the principle of LC resonance, a single-channel microfluidic detection chip with double solenoid inductor and capacitor in parallel was designed, and its detection principle was analyzed theoretically. Then based on the detection chip, the detection system formetal abrasive grain in oil wasset up. Finally, the experimental study on the inductance change of iron and copper particles in hydraulic oil at different excitation frequencies was carried out, and the detection experiment was carried out under the optimum excitation frequency. The experimental results show that when the frequency is lower than the resonant one, the iron particles produce positive inductance change pulse and the copper particles produce negative inductance pulse. When the frequency is higher than the resonant one, the iron particles produce negative inductance change pulse and the copper particles produce positive inductance pulse. When the frequency is close to the resonantone, the detection effect is the best. At this moment, the detection effect of double-coil resonant chip is better than that of the double coil chip, and the detection of 10 μm iron particles and 50 μm copper particles is realized in the experiment. The detection chip realizes the high precision detection of the metal abrasive grains in the hydraulic oil, and provides technical support for prevention and diagnosis of mechanical failure of the hydraulic system.Keywords：oil debris detection;microfluidic chip;double coil;resonance circuit;fault diagnosis- 58
- |
- 16
- |
- 0
Published：2020-08-13 -
Multi-sensor fusion UAV dynamic landing system
Abstract：In order to achieve the UAV without GPS guidance and high-precision dynamic landing, a dynamic and autonomous landing system for UAVs with multi-sensor fusion was established.UWB (Ultra Wideband) equipment was used to achieve three-dimensional positioning of UAVs,the landing near the labelwas guided by using position information. Then the visual processing algorithm was designed to quickly and accurately calculate the relative three-dimensional information of the UAV and landing label. Then, after analyzing the dynamic uniform model of the landing tag, Kalman filter was designed to estimate and correct the visual information in the horizontal direction and improve the landing accuracy. Finally, the filtered three-dimensional informations were utilized by the PID controller based on position control to realize the high-precision dynamic autonomous landing control of UAV.The experimental results show that the unmanned aerial machine can be guided to near the landing label and the landing accuracy within 5 cm.UAV in the absence of GPS signal guidance problem is resolved basically, the requirements of high precision dynamic and automatic landing is met.Keywords：UAV;dynamic landing;computer vision;UWB guided;multi-sensor fusion- 65
- |
- 13
- |
- 0
Published：2020-08-13 -
Improvement and application of RGB-D vision SLAM algorithm
Abstract：An improved visual SLAM algorithm of RGB-D was proposed aimed at prominent problem of accumulative position error and large calculated amount of traditional visual SLAM algorithm under the application background of wheeled mobile robot in the thesis. SURF algorithm was used to extract features, and random sample consensus (Ransac) and iterative closest point (ICP) algorithm were used to estimate and optimize movement process according to matching result in proposed algorithm; semi-random loop detection was introduced in the optimization process of back-end graph. Defects of recent detection of current frame and loop detection of historical frames were overcome by changing extraction method of key frames, which improved accuracy of matching and reduced calculated amount at the same time. Experimental result of open-source robot operating system (ROS) shows that speed of processing one frame data by proposed improved visual SLAM algorithm of RGB-D is 0.030 s approximately, which accelerates processing speed, saves calculated amount and strengthens real-time performance of the algorithm on the basis of satisfying accuracy and completeness of obtained 3D point cloud map.Keywords：RGB-D;Simultaneous Localization and Mapping(SLAM);feature point matching;loop closures detection- 79
- |
- 12
- |
- 0
Published：2020-08-13 -
DOA estimation of weighted smoothed l0 norm under multiple snapshots
Abstract：Aimed at the problem of low estimated accuracy of existing DOA estimation algorithms based on compressed sensing, a DOA estimation method of weighted smoothed L0 norm under multiple snapshots was proposed in the thesis. A new weighting method was adopted in the proposed method. After a proper smooth continuous function was constructed, a proper decreasing sequence of set was determined according to initial solution of receiving data, and the minimum value of approximation function of L0 norm was solved by the steepest descent method for every σ value; then the σ value was taken to be initial value of the next iteration, weight was updated at the beginning of each iteration, and minimum solution of approximation function namely minimum L0 norm of approximation was obtained by multiple iterations. Proposed method could implement effective estimation for DOA. It was easy to be realized with higher accuracy and had better estimation performance compared with unweighted DOA estimation method of smoothed L0 norm under multiple snapshots. Finally, the proposed method was verified by simulation experiment. The result shows that root mean square error of DOA estimation for two narrow-band target signal is 0.480 9° under the condition that snapshots with 32, signal noise ratio with -5 dB and array elements with 6 in the proposed method, which reaches design requirement of target estimation method in array signal processing basically.Keywords：array signal processing;Direction of Arrival (DOA) estimation;Compressed sensing;weighted smoothed l0 norm- 96
- |
- 17
- |
- 0
Published：2020-08-13 -
Geometric error measurement and integrated modeling for precision CNC machine tools of optical free-form surface
Abstract：In order to improve processing precision of independently developed precision CNC machine tools of optical free-form surface, geometrical error of the machine was measured and modeled. Geometric error of motion axis of precision CNC machine tools was measured by laser interferometer and basic geometric error term model was established according to Chebyshev polynomia. It was found that approximation level of basic geometric error model was high by comparing measured curves and matched curve. It is obtained by calculation that residual deviation bandwidth range of moving error of x, y and z axis is (0.63,2.84) μm, and residual deviation bandwidth range of angle error is (0.39,1.19) arc second, which shows that prediction precision of established model is high. The geometric error model of basic error was substituted into the model based on multi-body system theory to obtain integrated geometric error model under the operating condition of two-axis linkage. The modeling process is simple with easy program design and improves processing precision and efficiency of precision CNC of optical free-form surface.Keywords：CNC machine tools;geometric error;Chebyshev polynomial;error model- 87
- |
- 15
- |
- 0
Published：2020-08-13 -
Efficient matching technique for laser detection features of cable cutting traces
Abstract：A lot of line traces on the bearing surface of broken ends which usually present nonlinear morphological features and have strong randomness were left in the crime scene of cable cutting case. In order to implement trace feature matching and affiliated tool inference more rapidly, an efficient matching technique for laser detection features of cable cutting traces was designed:K-Means clustering was used to implement abnormal data correction for 1-D signals picked up on the surface of broken ends detected by single-point laser displacement sensor firstly, and then self-adaptation correction of rotation angle was implemented to unify matching datum. Finally, matching strategy based on threshold sequences was used to realize overlap ratio matching of trace feature similarity, thus realizing quick inference of corresponding tools, and cutting tool interference experiment by actual traces verifies practicability and effectiveness of the technique.Keywords：clamp tool;laser detection;abnormal correction;rotation correction;threshold sequence- 67
- |
- 10
- |
- 0
Published：2020-08-13 -
Quick defogging and cloud system realization for aerial image of guided filter
Abstract：A defogging method of aerial image based on guided filter was proposed, and application of defogging system based on cloud computing in visual navigation and reconnaissance was realized. Firstly, atmospheric curtain function of aerial image was calculated by shadow channel, and edge preservation estimation of atmospheric curtain function was calculated by guided filter; then, residual image of original image and atmospheric curtain function were calculated and bright channel was extracted; finally reflection coefficient of scene was obtained according to light reflection model as recovered image. It wasn't needed to obtain global atmospheric light and better image details and overall visibility by guided filtering algorithm. The defogging method was implemented by powerful computing resources of cloud computing. Compared with traditional airborne processing system by unmanned aerial vehicle (UAV), using powerful distributed computing power and massive storage capacity of cloud computing makes defogging system more efficient and quicker. It improves duration of flight of unmanned aerial vehicle, makes all-whether reconnaissance of real-time aerial photography become possible, and has extensive application prospect.- 63
- |
- 15
- |
- 0
Published：2020-08-13 -
Personalized image searching based on clustering analysis and user interest model
Abstract：Aimed at the problem that existing image searching algorithms were difficult to evaluate query purpose of users completely and the quality of image retrieval was low, a personalized image searching method based on clustering analysis and user interest model was proposed in the thesis. Input retrieval image was divided into 9 sub-blocks and transformed into HSV color space, and color distribution histogram of image was used to extract color feature information. And then, Gabor wavelet was used to extract texture features of the image, and obtained color features and texture features were fused to form multi-feature fusion similarity matrix of the image to calculate similarity among images. Then multi-feature fusion similarity matrix was used as input of multi-core dynamic clustering to cluster the images in the database. The clustering image was sent to LSSVM network to determine the classification surface and construct personalized user interest model. Finally, retrieved results were provided to users for independent choice according to comparison with similarity degree of user interest model. Experimental result shows that:average recall ratio and precision ratio of method in the thesis are improved by 8.2%, 11.42% and 19.7%, 26.08% compared with search algorithm of single color and texture features. It can promote quality of image searching effectively and has obvious application value.Keywords：image retrieval;relevance feedback;Computer simulation;LSSVM- 42
- |
- 13
- |
- 0
Published：2020-08-13 -
High-precision segmentation extraction of harbor in remote sensing images based on F-B template
Abstract：Detection precision of ship in harbor was extremely dependent on the effect of sea-land segmentation in the harbor. Docking of ships in the harbor was usually connected with embankment, which increased the difficulty of harbor segmentation and detection. A high-precision segmentation extraction algorithm of harbor was proposed according to remote sensing image features of the harbor. Interested harbor areas were roughly located by latitude and longitude coordinates contained in remote sensing image data, and then accurate registration from detection image to B (Binary) template was finished according to corresponding F (Feature) of harbor; finally, high-precision segmentation and extraction of harbor was finished according to B template. Resulting images in the thesis were consistent with size and direction of template images, had accurate sea-land segmentation and were more beneficial to high-precision detection and recognition of ships in subsequent harbors. Through experiment by multiple groups of remote sensing images of the harbor, the result shows that algorithm in the thesis has the highest segmentation detection precision of harbor compared with segmentation extraction algorithm of mainstream harbor such as MLE, OTSU, Mean-shift, LBE and LBP and it is superior to other similar methods in both objective evaluation and subjective evaluation. RUMA evaluation index is improved significantly compared with other similar methods, and the best method at present has increased by 2.5 times at average compared with multiple groups of experiments.Keywords：harbor segmentation extraction;harbor priori;F-B template;template registration- 52
- |
- 8
- |
- 0
Published：2020-08-13 -
Application of improved support vector machine in identification of pulmonary nodule
Abstract：In order to solve the problems of the unbalanced distribution of positive and negative samplesin data set in pulmonary nodule identification and overtime parameter optimization, a PSO-CSVM algorithm was proposed. ROI image of pulmonary nodule was extracted from the lung CT and then 13-dimensional characteristics was extracted from it. Finally, the proposed PSO-based cost-sensitive type SVM algorithm was used for identification. In the testing the accuracy rate of the identification reached 91.11% and the sensitivity reached 85.71%, specificity reached 93.55%, and the time of parameter optimization was 54.37 s. In order to further verify the effectiveness of the algorithm, the proposed algorithm was compared with the genetic optimizing algorithm and grid optimizing searching algorithm. The experimental result shows that the run-time of PSO-CSVM is shorter and the accuracy and sensitivity is optimal.It features short run-time, high accuracy rate of identification and detection rate and can meet the requirements of medical imaging for the identification of pulmonary nodule.Keywords：pulmonary nodule recognition;cost-sensitive support vector machine;particle swarm optimization;RBF kernel- 69
- |
- 12
- |
- 0
Published：2020-08-13 -
Deep learning target detection based on pre-segmentation and regression
Abstract：Aiming at the problem of difficult small target detection in high resolution images, combined with region-of-interest (ROI) extraction strategy in target detection method based on candidate region and regression strategy in target detection algorithm based on regression, deep learning target detection algorithm based on pre-segmentation and regression (Quad-ssd) was proposed. As fast-RCNN series implement image location and classification separately, small targets could be detected but detection time was too long. YOLO series method used regression method to implement classification and location for targets in images at the same time. As only high-level features were used, detection accuracy for small target was not enough. Therefore, quad tree was used to extract interest target of original images, and target detection method based on regression was used to implement detailed relocation and classification for targets in interested region. Compared with traditional Fast-RCNN method and deep learning method based on regression of YOLO series, target detection algorithm of deep learning based on quad tree has obvious advantages in accuracy and speed. The experimental results show that compared with Fast-RCNN, accuracy of Quad-ssd algorithm is improved by 6.5% and reaches 74.9% at the time of target detection. The detection speed is improved greatly; reaching 45 f/s, and can satisfyrequirements of timeliness.Keywords：quadtree;deep learning;small target detection;Quad-ssd- 124
- |
- 31
- |
- 0
Published：2020-08-13 -
Contention resolution and blocking fairness improvement based on limited delay preemption
Abstract：In order to improve blocking fairness of users, limited delay preemption scheme based on fairness improvement (LDP-FI) was proposed to satisfy the requirement of low discarding rate of high and low priority burst in optical burst switching network synchronously. In core nodes, priority level and real-time traffic states reached by LDP-FI according to groups decreased invalid discard of low priority users effectively under high traffic state by dynamic setting of traffic threshold and restriction of buffer and channel occupancy authority of high priority users under blocking condition. Numerical results show that:discarding rate is directly related to proportion of high and low priority users and set value of traffic threshold. When proportion of high priority users is less than 0.3, and dynamic traffic threshold value is 0.5-0.7, LDP-FI can guarantee that blocking rate under middle and high traffic state is lower than 10-2 and fairness index is kept to be 0.95. Compared with traditional preemption mechanism based on priority, LDP-FI gives consideration to blocking and fairness requirements of users at the time of supporting differentiated service, avoids invalid transmission and secondary blocking caused by discard of a large number of low priority burst and improves practicability and reliability of optical switching networks effectively.Keywords：optical burst switching;contention resolution;fairness;blocking rate;preemption;buffer- 43
- |
- 6
- |
- 0
Published：2020-08-13 -
Intensification of retinal image integrating dual-tree complex wavelet and improved morphology
Abstract：In order to solve the problem that the intensification of retinal image affected by chromatic aberration, contrast and dense blood vessels etc, was hard to achieve, an intensification method of combing dual free complex wavelet (DT-CWT) and advantages of morphology transform was proposed. The retinal image was decomposed by using the transformation of dual-tree complex wavelet (DT-CWT)at first, and then adaptive wavelet threshold de-noising was applied to the high-frequency part and for the low-frequency part, top-hat transformation of improved morphology was adopted for its processing, finally, the high-frequency and low-frequency part were reconstructed by adopting the inverse transformation of dual free complex wavelet and the intensified retinal image was obtained. In order to verify the algorithm, 40 retinal images in the DRIVE database were used as samples for simulation experiment. The data shows that standard deviation, mean value, information entropy and other evaluation indexes can be increased by more than 15% by using the algorithm of the thesis; for the image definition, diversity factor and degree of distortion are at least optimized by more than 5% compared with other algorithms. Arrivalling at a conclusion that the algorithm proposedperforms better in improving the overall visual effect of image.Keywords：Retinal image;dual tree complex wavelet transform;top-hat transform;image enhance- 44
- |
- 10
- |
- 0
Published：2020-08-13 -
Detection operator for traffic image in hazy weather
Abstract：A new edge detection method of traffic image in hazy weather was proposed pertinent to the problem that complete edge extraction cannot be achieved for traffic image in hazy weather in current edge detection method of image. Coarse edge of traffic image was extracted with newly proposed eight-direction edge detection operators pertinent to edge detection for traffic image in hazy weather, then weight of pixel points in partial directions were increased to overcome defocus blur due to refraction of haze particles in combination with features of hazy images; weight value was adjusted to design new refining operators of image edge so as to conduct refinement treatment on image edge and reduce quantity of fake edges; finally refined image edge was output for display. Experimental result indicates that most edge details of traffic image in hazy weather can be detected in the method proposed and more fine image edge can be achieved simultaneously compared with existing edge detection methods; it has better distinction degree and more accurate image edge positioning; image edge strength increases by over 50% compared with traditional method, thus effect of edge detection in hazy weather is effectively increased.Keywords：haze;traffic image;edge detection operator- 69
- |
- 15
- |
- 0
Published：2020-08-13 -
Tagging for improving feature points of human spine model by using Gaussian curvature and vector inner product
Abstract：A tagging method for improving feature points of human spine model by using Gauss curvature and the vector feature was proposed in the thesis to dynamically adjust physical coordinates of tagging points based on Gauss curvature values and vector inner products of various points located in local neighborhood in this model. Three-dimensional reconstruction and mesh dividing was used to generate 3D mesh model of thehuman spine; and then, a point in manual pick-up model was chosen as the initial pickup point. Spherical space with minimal radius R was regarded as the maximum approximate neighborhood of the pickup point, based on which absolute values of Gauss curvature for all vertexes in the neighborhood of the pickup point were calculated; n points with larger values from above Gauss curvature values were selected to calculate vector inner product between these n points and initial pickup point and further calculate the point from these points which had the minimum included angle with the initial pickup point, finally, coordinate of the initial pickup point would be corrected to the physical coordinate of the point which had the minimum included angle. The method makes sure that Gauss curvature value taken from the neighborhood is the largest possible value, it can make tagging results closer to the real feature pointat the same time. Experimental results show that the method improves the accuracy of positioning feature point of spinal 3D image by about 29%, besides, relatively good tagging results can be obtained by the method for smooth region of the spine model.Keywords：spine model;3D mesh model;feature point labeling;Gaussian curvature;vector inner product- 65
- |
- 9
- |
- 0
Published：2020-08-13 -
Research on improved fuzzy clustering segmentation in magnetic resonance image for brain
Abstract：The main defect for Fuzzy C mean clustering (FCM) was initialization sensitivity and trapping into the local optimum of clustering center, and nonlinear separable data could not be provided with clustering through this algorithm. Selection for initial clustering center was random selection, which causes relatively bad image segmentation results. In order to conquerthis defect, Gaussian kernel Fuzzy C mean clustering was proposed. Fuzzy C mean clustering algorithm for stimulated annealing genetic algorithm was proposed to guarantee traditional algorithm can rapidly converge to accurate and stable clustering center. Relatively strong local search ability for simulated annealing and relatively strong global search ability for genetic algorithm were utilized to effectively select initial clustering center and improved convergence speed. In order to strengthen and improve non-ideal problem of improved algorithm to clustering result, nonlinear transformation was conducted with the help of Gaussian kernel function to be mapped into higher dimensional space and be conversed into linear separable problem of higher dimensional space, which could improve brain MRI image segmentation effect. Under the effect of noises and biased field with different types, segmentation time for algorithm in the thesis saved 1-3 times than other algorithms, and it was at least 0.01 larger than Probabilistic Rand Index (PRI). It can be found that time, white matter and grey matter segmentation accuracies of algorithm in the thesis both have obvious advantages in clinical image segmentation results. For brain MRI image segmentation, the algorithm has better robustness and effectiveness in speed and accuracy compared with traditional FCM algorithm.Keywords：fuzzy C means clustering;brain MRI image;image segmentation;simulated annealing genetic algorithm;gauss kernel function- 59
- |
- 6
- |
- 0
Published：2020-08-13 -
Signal processing for double objectives to impact simultaneously on single linear CCD vertical target
Abstract：Currently several coordinate measurement methods for bullets to impact could not be used in the situation for two bullets to impact simultaneously, thus CCD vertical target measurement system of single linear array composed of components such as a CCD camera based on single black-and-white linear array, a double hyphen line laser and projection plank, etc was adopted to overcome the problem that measurement for double objectives to impact could not be achieved in traditional coordinate measurement method for bullets to impact. Theoretical analysis and actual handling were conducted on signal of double objectives captured by CCD vertical measurement system of single linear array and recognition was conducted on three different types of signals of double objectives according to two different signal feature including grey value of bullet signal and pixel width value of occupied CCD device. Feasibility and measurement precision of the system were verified through the method to simulate live ammunition verification and feasibility of signal handling method of double objectives was proved; standard deviations and of measurement error of coordinate σx and coordinate σy are respectively 2.0 mm and 6.8 mm when target surface is 0.5 m×1 m.Keywords：shooting range test;projectile;impacting coordinate;vertical target;dual-object;signal processing;measurement accuracy- 60
- |
- 9
- |
- 0
Published：2020-08-13 -
Application of BP neural network in obstacle avoidance of driverless car
Abstract：In order to achieve the reasonable matching of the environmental obstacle information of driverless car and the performed action command mode, the BP neural network applied to the obstacle avoidance of driverless car in the article and technology of obstacle avoidance of driverless car based on BP neural network was researched. The environmental information identification model of BP neural network was established. The 180° planar domain detected by single line laser radar in front of the driverless car was divided into 8 sub-domains and the scan range of each sub-domain was 22.5°. The information of environmental obstacles of the 8 sub-domains was taken as the input feature vector of BP neural network system and the action command of the driverless car controlled by identification was taken as the output vector of BP neural network system. The input environmental coding information matched with the performed action command by using BP neural network. The experimental result shows that the error between the obtained result of the obstacle avoidance model of driverless car based on BP neural network and expected target value is controlled within the range of 0.001 and the accurate and quick classification matching between environmental coding information and the performed action command is achieved, and the aim of driverless car reasonably and effectively avoiding obstacles is achieved.Keywords：environmental obstacle information;action mode instruction;BP neural network;single-line lidar- 64
- |
- 12
- |
- 0
Published：2020-08-13 -
Design of signal processing module for infrared focal plane detector
Abstract：Aimed at theneural network non-uniformity correction algorithm based on scenes requires high standards on hardware architecture data operation and transmission capacity, in order to achieve on-chip real-time operation, high-efficient data processing and other purposes, XC7K325T FPGA chip of Kintex 7 series and TMS320C6657 DSP chip of TI C66x series were adopted to construct signal processing module based on FPGA and dual-core DSP architecture. DDR3 with high speed was introduced in the design to improve overall system data throughput capacity, statistical interconnection between FPGA and DSP was designed and achieved by using SRIO interface technology, and theoretical transmission speed of 2.5 Gb/s was reached. At the same time, algorithm logic of neural network non-uniformity correction was realized in the interior of DSP processing chips, correction parameter matrix was calculated in real time on the chips, and stable operation was implemented under the mode of continuous work for 50 H of infrared detector with resolution ratio of 640×512. Finally, maximum single board area of integrated module is 90 mm×52 mm, integrated height of the whole module is less than 50 mm and stability of non-uniformity of infrared image output is less than 0.1%, which satisfies miniaturized and real-time design demands of signal processing module, and achieves engineering standard.Keywords：infrared focal plane detector;digital signal processor;non-uniformity correction(NUC);neural network- 77
- |
- 16
- |
- 0
Published：2020-08-13 -
Center positioning of target surface image for subgrade settlement monitoring system of high-speed rail
Abstract：Affects measurement accuracy of monitoring system was affected by the center positioning technology of target image. Aimed at the problem that laser image positioning of traditional laser strip center positioning technologycauses larger error when the distance between laser and target surface is relatively far, algorithm of positioning cross light spot image center with two steps by precise positioning using coarse positioning of cross light spot intersection and local gray centroid method was proposed:shape features of cross light spot image were extracted based on image preprocessing to obtain binary image firstly; secondly cross centre intersection was positioned coarsely according to shape features of cross light spot images, and image center was obtained by positioning with gray centroid method in local rectangular region. Experimental result shows that calculation time of the whole experimental process is 1 s and standard deviation of the experiment is 0.145. Compared with elliptic light spot positioning with gray centroid method, algorithm in the article has higher calculation accuracy with shorter time and is applicable to surface settlement monitoring system of ballastless track.Keywords：Ballastless track;settlement monitoring;cross sport;gravity center method;center positioning- 70
- |
- 18
- |
- 0
Published：2020-08-13 -
Denoising method for shape measurement based on three wavelength phase shift profilometry
Abstract：In order to reduce noise points in shape measurement of 3D objects using Phase Shift Profilometry (PSP) methods, a novel Three Wave Length PSP (TWPSP) method was investigated. Firstly, relevant problems of equivalent wavelength and unwrapped phase were analyzed. Then, the solution of unwrapping method for TWPSP was derived. Finally, global phase filtering based method was used to reduce the phase noises. The measurement system was designed to measure the calibration target as well as the complicated shape target. Experimental results show that the noisy points of three-dimensional graphics are reduced by 98.02%, and the speed of 3D reconstruction is raised by 12%. The experiment and analysis results show that this system has better robust and higher measurement accuracy, thus the noises in the shape measurement can be suppressed significantly.Keywords：three-dimensional tomography;phase shift profilometry;denoising;multi-wavelengths;phase unwrapping- 81
- |
- 14
- |
- 0
Published：2020-08-13 -
Blind restoration of atmospheric turbulence images based on P system
Abstract：The image process of ground-based telescopes is affected by complicated and variable factors, which seriously influences high-precision observation of space targets. Herein, a spatial object image restoration algorithm based on the P system was proposed based on the atmospheric turbulence transfer function model under long-term exposure. The non-reference single-frame image evaluation index was taken as the optimization objective function, and the P system optimization method was used to quickly obtain the atmospheric coherence length and spectral density ratio, thus reconstructing the image with the Wiener deconvolution algorithm. The algorithm was compared with five main blind restoration algorithms experimentally, which shows the proposed algorithm has the best turbulence deblurring image restoration effect. In terms of the simulated images, the average gradient and edge strength index of the proposed method are 3.74 and 39.92 respectively. In terms of real images, the entropy and edge strength of the method are 5.66 and 61.61 respectively. Generally, the average evaluation of the algorithm is more than 10% above the contrast method.Keywords：atmospheric turbulence;atmospheric coherence length;Wiener deconvolution;P system;image restoration- 62
- |
- 12
- |
- 0
Published：2020-08-13 -
Recognition of human tumbles based on improved normalized inertia moment
Abstract：In order to detect human tumble behavior in videos quickly and accurately, an improved Normalized Moment of Inertia (NMI) feature extraction algorithm was proposed, which used the idea of SURF to structure feature retrieve, and combine SURF with NMI to calculate the NMI values. Taking the feature points with the closet NMI values in the two images of video as the matching points, the total of matching points were calculated between the two images. Moreover, the number of the feature points was compared with the minimum of feature points in a tumble video, thus judging whether a tumble behavior has happened or not. The results show that accuracy rate of this improved NMI algorithm runs up to 96%, and the average recognition time is 0.138 s, which is faster than other algorithms. The algorithm occupies a little internal storage and detects quickly and accurately, which is a feasible and effective approach to detect human actions.Keywords：video detection;normalized moment of inertia;feature extraction;Feature points matching;fall behavior- 51
- |
- 10
- |
- 0
Published：2020-08-13 -
Obstacle region recognition using single image for planetary soft landing
Abstract：In order to overcome the problem of obstacle detection in complex terrain for planet soft landing, a passive visual obstacle detection method based on single image was proposed. First, a saliency map of the image was obtained using the saliency detection based on the phase spectrum method, and the variance map of the original image was calculated. Then, the two-dimensional histogram of terrain image was constructed by counting the gray level of two graphs, and the segmented Gamma correction was used to enhance the peak feature. Furthermore the two-dimensional histogram of terrain image was segmented by the two-dimensional Otsu method, which was used to obtain the binary image of the obstacle area. Herein, the various terrain grayscale images provided by HIRISE and the corresponding DEM data were employed to evaluate the accuracy of the obstacle detection algorithm. The average detection rate of TN + TP is over 80%. Experiment result indicates that the algorithm can effectively segment the obstacle area and the safety area in complex terrain environments, thus providing useful information for landing point selection.Keywords：obstacle detection;saliency;variance map;Gamma correction;two-dimensional Otsu method- 92
- |
- 162
- |
- 0
Published：2020-08-13 -
Filtering of UV images of UV-Vis dual-band imaging system
Abstract：In order to reduce image noise and improve registration accuracy in visible and UV image registration of UV-Vis dual-band imaging system, a UV image filtering algorithm based on correlation operation was designed. Firstly, the type of noise was confirmed by analyzing sources and characteristics of UV image noise. Then, correlation operation was completed between UV image and a continuous mask, and the noise was eliminated through a suitable threshold. Finally, the algorithm was verified by comparing the result of correlation filtering and other filtering methods, and the effect of the correlation filtering algorithm was further verified by detecting the corner feature of calibration board in UV image. The experiments prove that the peak signal to noise ratio(PSNR) of filtered UV image is 45.31 dB by correlation filter, and it's running time is 0.57 s measured by matlab. Furthermore, the corner feature of calibration board in UV image was detected throughly by Shi-Tomasi algorithm. Therefore, the correlation filter can eliminate the noise of UV image with high-speed and less pixel loss, which can satisfy the requirement of high accuracy image registration.Keywords：UV-Vis dual-band imaging system;correlation operation;image registration;image filtering;corner detection- 62
- |
- 13
- |
- 0
Published：2020-08-13
Loading...
0