Chinese Version Light学术出版中心
Latest Issue
Issue 2 ， 2021
-
Automatic offset-frequency locking of external cavity diode laser in wide wavelength range Cover Article
Abstract：To construct different synthetic wavelengths for multi-wavelength interferometric absolute distance measurements， an automatic offset-frequency locking method was proposed to lock the frequency of a tunable external cavity diode laser （ECDL） to a femtosecond optical frequency comb （OFC） in a wide wavelength range. First， based on the grating and double convex lens comb filtering， a beat signal detection unit was designed. In addition， the beat signal detection between the ECDL and target comb modes in a wide wavelength range was realized. A phase-frequency detector was then designed based on the principle of a lock-in amplifier； phase-locking with a large capture range， wide monotonic phase range， and high accuracy was achieved. Finally， multiple closed-loop controls were adopted to realize automatic adjustment and offset-frequency locking of the ECDL over a wide wavelength range. Experimental results show that the proposed method can automatically lock the ECDL to OFC in the range of 10 nm， and the average signal-to-noise ratio of the beat signal is approximately 35.9 dB. Over 4 h， the standard deviation of the ECDL laser frequency is approximately 1.49 kHz， and the relative Allan deviation is approximately 4.76 × 10－12 at 1 s. The proposed automatic offset-frequency locking method can satisfy the requirements of wide-range wavelength adjustment and high-precision frequency stabilization in precision interferometry.Keywords：external cavity diode laser;laser frequency stabilization;offset-frequency locking;beat signal detection;femtosecond optical frequency comb;phase-frequency detector- 580
- |
- 203
- |
- 3
Published：2021-03-05 -
Snapshot spectral light-field imaging technology
Abstract：Realizing high-dimensional imaging for a snapshot multi-dimensional imaging system is difficult. Therefore， this study proposes a snapshot spectral light-field imaging method that can simultaneously record the light field and spectral information of a target scene using a single detector. This method introduces the paradigm of light-field imaging to Fourier transform imaging spectroscopy. A raw image composed of the light-field image and interference is then obtained. A convolutional neural network is developed to decouple the light-field image and interference. The depth and spectral information of the scene can then be reconstructed. The study also constructs a snapshot spectral-volumetric imaging system， and its performance in spatial， spectral， and temporal dimensions is verified through experiments. Experimental results showed that the root mean square error （RMSE） of depth reconstruction was 7.7 mm， and the normalized RMSE of spectral reconstruction was 6.87% when the acquired information was extended to seven dimensions.Keywords：instrumentation;snapshot spectral imaging;light-field imaging;multi-dimensional imaging- 652
- |
- 208
- |
- 0
Published：2021-03-05 -
Numerical calculation of transmittance of fogs based on dynamic particle parameter selection
Abstract：To investigate the influence of multiple scattering effects on the transmission characteristics of radiation through fog， a polydisperse fog layer consisting of spherical water particles of different radii was proposed in this study. The radii of fog particles were first divided into several radius regions， and the transport approximation optical parameters of fog particles in each radius region were calculated. A Monte Carlo method based on dynamic particle parameter selection was next proposed， and the propagation of 0.85- and 4-µm radiations through heavy advection fog， moderate advection fog， heavy radiation fog， and moderate radiation fog was simulated. The influence of the propagation distance on the transmittance and computational time was analyzed， and the results were compared with those of the ordinary Monte Carlo method. The results show that when compared with 4-µm radiation， 0.85-µm radiation has better transmission properties in fog. The trends of transmittance versus propagation distance obtained using the Monte Carlo method based on dynamic particle parameter selection and the ordinary Monte Carlo method are similar. When the propagation distance is 150 m， the absolute errors of the two methods for transmittance of 0.85- and 4-µm radiations through heavy advection fog are 0.048 3 and 0.001 5， respectively. Compared with that of the ordinary Monte Carlo method， the computational efficiency of the Monte Carlo method based on dynamic particle parameter selection for 4-µm radiation through heavy advection fog is as high as 44%. The Monte Carlo method based on dynamic particle parameter selection should be used for radiation transfer calculation in fog for long propagation distance cases， which can effectively shorten the computational time. The results of this study are significant for target recognition and detection in foggy weather.Keywords：atmospheric optics;multiple scattering;fog;transport approximation;Monte Carlo method;transmittance- 188
- |
- 37
- |
- 0
Published：2021-03-05 -
Factors influencing stone displacement in holmium laser lithotripsy
Abstract：To reduce the displacement of stones caused by holmium laser lithotripsy， the effects of peak pulse power， laser cauterization size， and working distance on stone displacement were explored in this study. First， domestic and foreign studies were reviewed to determine the factors that may affect stone displacement. Second， an in vitro model device was built to simulate the process of holmium laser lithotripsy in the ureter. The experiment used a 0.25-g cubic gypsum model to replace human stones. The parameters required for the experiments were then determined. The holmium laser power was selected to be 12 W （0.6 J × 20 Hz， 0.8 J × 15 Hz， 1.2 J × 10 Hz） and 20 W （0.8 J × 25 Hz， 1.0 J × 20 Hz， 2.0 J × 10 Hz）， The pulse duration was 200 and 800 μs， the fiber sizes were 272 and 550 μm， and the working distance was in the range of 0 to 5 mm. Finally， the in vitro model was placed in water， with the room temperature kept constant during the experiment. Results showed that stone displacement increased with an increase in pulse peak power. A longer pulse duration could reduce the displacement of stones by 16.95%–27.27%. The 272-μm fiber could reduce the displacement by 35.59%–54.17% as compared with the 550-μm fiber. When the working distance increased， both the stone displacement and ablation efficiency decreased. However， when the stone model contacted the fiber tip， the phenomenon of adsorption struggle occurred. This study can be a reference for improving the design of domestic holmium lasers and the fine settings of holmium laser parameters. It also has significance for clinical surgery.Keywords：medical optics;domestic holmium laser;laser ablation;stone displacement;vaporization bubble- 174
- |
- 28
- |
- 0
Published：2021-03-05 -
Multi-bottle beam optical trap and its light field regulation
Abstract：In this study， the relationship between the position of an optical trap and the deflection angle of a reflector was studied using an optical system that can produce a multi-bottle optical trap. The position of the optical trap could be arbitrarily changed by adjusting the deflection angles of both the mirror and reflector for accurate capture and trapping of particles. Based on the theory of diffraction integral and matrix optics， the light field distribution of the incident light source passing through an optical element was analyzed and calculated. When the relationship of angles is θ1 – θ2 = 90°， the positions of the transformed optical traps are all on inclined straight lines. When the angles of deflection of the two mirrors are θ1 < 112.5° and θ2 < 22.5 °， respectively， the four-bottle trap can be transformed to one of 12 bottles. The gradient and scattering forces of the formed optical traps were calculated， and the Monte Carlo method was used to verify that the Rayleigh particles had stable captivity in a very small area in the center of the optical trap. This work shows that studying the precise trapping of multiple particles in multi-bottle optical traps is of great significance.Keywords：optical tweezer;laser technique;multi-bottle beam;reflector;position transformation;particle trapping- 147
- |
- 22
- |
- 1
Published：2021-03-05 -
Design of optical system for laser fusion shock velocity measurement
Abstract：To meet the demands of laser fusion shock velocity measurements， a composite system that integrates a passive scanning optical pyrometer with an active shock wave velocity interferometer was designed in this study. The common-path lens was achromatic and designed in the range of 400–700 nm by using several types of radiation resistance optical materials. The mirrors of a Mach⁃Zehnder interferometer were motorized and could be controlled remotely， thereby simplifying the operation of the system. In addition， the system magnification could be altered by using different focal lengths with a streak camera coupling lens. The field of view of the combined system is 2 mm， and the magnification settings for the active velocity measuring system are 10×， 20×， and 30×. The static interference fringes are straight， and their contrast is greater than 0.69. The system has a 4.72 μm spatial resolution. The system can be applied to dynamic measurement of shock velocity in a laser fusion device.Keywords：optical system design;velocity interferometer;Doppler shift;scanning optical pyrometer- 118
- |
- 24
- |
- 1
Published：2021-03-05 -
Development of photoelectric integrated single photon avalanche photodiode and front-end circuit based on BCD technology
Abstract：A type of single-photon avalanche photodiode （SPAD） and front-end quench reset circuit （QRC）， based on bipolar-CMOS-DMOS （BCD） technology， for detecting weak optical signals was proposed in this study. To reduce the risk of edge breakdown and to improve the responsivity， a circular SPAD with a P+/Nwell/deep-Nwell （DNwell） structure was designed. The pn junction formed between the DNwell and substrate could effectively reduce the dark current flowing from the P substrate to the avalanche region. It could also reduce the dark count rate， ensure a small longitudinal transit time， and improve the response speed. Simultaneously， a Pwell protection ring was designed to increase the breakdown voltage of the device. A two-dimensional simulation of the device was conducted using SILVACO， and the proposed structure was then compared with the traditional P+/Nwell and P+/Nwell/DNwell structures. The advantages of the design structure in terms of breakdown voltage and responsivity were verified. To realize a cooperative design of photodetectors and integrated circuits， the equivalent circuit model of avalanche photodiode optoelectronic devices was improved. Accordingly， an active quenching reset circuit was designed. The dead time is determined to be approximately 2.6 ns， which can achieve the purpose of rapid detection. Results show that the avalanche breakdown voltage of the P+/Nwell/DNwell structure is 15.8 V. When the excess bias voltage is 0.2 V， the responsivity is approximately 0.80 A/W， and the dark count rate is 20 kHz.Keywords：single photon avalanche photodiode;optoelectronic integration;BCD technology;responsivity;time-of-flight sensor- 134
- |
- 24
- |
- 0
Published：2021-03-05 -
High-power spectral instrument for chronic soft tissue injury
Abstract：In the course of phototherapy， the spectral band and therapeutic light intensity are the key factors that determine the indications， therapeutic depth， and curative effect. Based on the principle of water-filtered infrared A， a high-power halogen lamp was used as the light sourcein this study. A high-power spectral therapeutic apparatus， comprising a beam shaper， a closed liquid filter， and an electronic control system， was designed. Its basic performance was equivalent to those of advanced foreign products. The spectrum， optical power density， electrical safety， and electromagnetic compatibility passed the medical device registration test. The device was used clinically for 71 patients to treat chronic soft tissue injury， inflammation， and pain. After （6±1） days of treatment， the total effective rate is 100%， and the significant efficiency is 80.28%， without adverse reactions being recorded. The therapeutic apparatus improves the effective spectral width and the therapeutic light intensity, reduces the thermal effect, increases the penetration depth and curative effect, and can be widely used in soft tissue injuries, chronic diseases, pain and wound healing.Keywords：medical optics;spectrum therapy;waveband;light intensity;treatment depth- 132
- |
- 27
- |
- 0
Published：2021-03-05
Modern Applied Optics
-
Magnetorheological finishing method that combines mechanical and virtual axes
Abstract：The conventional magnetorheological finishing employs a fixed position of the polishing ribbon to perform normal processing on the workpiece. However， the processing area of the workpiece is greatly restricted by the limited stroke of the machine tool rotary axis. Recently， shortcomings of the current polishing method have been identified based on the principle of equivalent magnetic fields， which implies a high cost for realizing an equivalent magnetic field and insufficient polishing ability due to the fact that the mechanical and virtual axes are not combined. A method for processing a high-gradient curved surface was thus proposed in this study. The characteristics of the magnetic field that ensure the stability of the removal function were then analyzed， and the magnetic field stability range was verified through a magnetic field measurement experiment. The study also conducted a spot-taking experiment， which determined that the feasible range of the virtual axis for the removal function stability was ±12°. A machining method that combines the virtual and mechanical axes was then proposed， and a coordinate calculation for the machining method was realized based on rigid body transformation. Finally， the study conducted a spherical polishing experiment with an greater inclination angle. Results shows that the peak-valley（PV） value of the spherical workpiece with a 95% aperture size converges to 0.096λ， and the root mean square（RMS） value converges to 0.012λ， indicating that the proposed composite polishing method of virtual and mechanical axes can obtain a high-gradient surface.Keywords：magnetorheological finishing;variable normal-contact;virtual axis;mechanical axis;high-gradient surface- 211
- |
- 39
- |
- 1
Published：2021-03-05 -
column:Three-DoF weakly coupled resonator micro-balance for mass detection
Abstract：To measure micro-masses effectively， a three degrees-of-freedom （3-DoF） mode localized weakly coupled resonator micro-mass sensor was proposed in this study. The equations of motion of three resonators were established， the system equations of mass disturbance were obtained， and a MATLAB/SIMULINK model of a 3-DoF coupled resonance system was established. The study determined that the variation in amplitude ratio has the highest normalized sensitivity compared with the variation in amplitude and frequency shift. The sensor was designed and fabricated using micro-electromechanical system （MEMS） technologies， and a mass measurement system was also developed. Experimental results reveal that the normalized sensitivity of the change in voltage amplitude ratio is approximately four times the change in amplitude. In addition， approximately two orders of magnitude of resonant frequency shift occurred， and the minimum detectable mass is 2 ng. Simulation and experimental results prove that the 3-DoF mode localized weakly coupled resonator can effectively measure micro-masses and thus has good prospects for bio-sensing applications.Keywords：micro-balance;3-DoF;mode localization;weakly coupled;resonator- 132
- |
- 16
- |
- 0
Published：2021-03-05 -
Kinematic decoupling and velocity adaptive planning of parallel robot in large overall translation
Abstract：Parallel robot is a kind of nonlinear strong coupling system with many branches and joints. It has obvious advantages of high speed， high stiffness and large load. So， it is widely used to industrial field. However， with the number of joints increasing， it becomes more difficult to finish kinematic decoupling and precision control. In order to realize automatic trajectory tracking and stable control of 3-RRRU parallel robot， kinematic decoupling and velocity adaptive planning method are systematically and deeply researched. The kinematic equations are derived by DH theory and space vector method. Inverse kinematics calculation is completed based on structural constraints. The velocity correction mechanism is added to the S-type control strategy， and the maximum speed parameters can be calculated and revised automatically according to different trajectories. It realizes the adaptive optimization ofthe speed control strategy. The laser tracker is used to track the robot's path， and the accuracy is compared and analyzed under control of standard S-type and ladder type. The maximum error reaches to 4.513 mm， which is 3 times of the S-type control strategy. There are several error peaks on the position curve which shows that the motion stability is poorer due to the sudden change of speed. On this basis， the second experiment is carried out. We test the stability and tracking error of the robot before and after adopting correction mechanism for the S-type speed control under the circular trajectory. The experimental results have shown that when robot can’t achieve the maximum speed on the preset path， the original max speed is still control parameter which cause the asymmetric distribution of the speed control curve. Therefore， when the robot reaches the end of the trajectory， it also owns a large inertia speed and results in a peak error of 2.676 mm which is 2.4 times of the after correction. Robot stops accompanied by obvious shaking. After adaptive correction， the position errors of the starting point and the ending point are 0.722 mm and 0.382 mm respectively， and the relative position errors of the two position are only 0.34 mm. It effectively solves the inertia speed problem when the robot reaches the end of the trajectory， and greatly improves the overall tracking accuracy and motion stability of the robot.Keywords：large overall translation;parallel robot;kinematic decoupling;Adaptive control;Accuracy- 148
- |
- 45
- |
- 3
Published：2021-03-05 -
Parameter calibration of 6-dof parallel mechanism using orthogonal displacement measurement system
Abstract：In order to simplify the parameter calibration process of 6-DOF parallel mechanism， improve the calibration efficiency and reduce the calibration， a novel device and method of pose measurement based on orthogonal displacement measurement system is described.First， the pose calculation method of the device is studied， and the forward and reverse kinematic solution is solved by using the method of spatial analytic geometry. Secondary， the error model of the combination of parallel mechanism and orthogonal displacement measurement system is established by using the method of micro displacement synthesis. Then， based on the error model， the model of optimal problem of parameter identification is constructed. The minimum square sum of the sensor indication error is the objective function， and the parameter error of the combination is the variables. Finally， the pose of the parallel mechanism are measured by orthogonal displacement measurement system ，and the optimal solution of the parameter error is found by OASIS， and model parameter in the control system is compensated .the calibration of parallel mechanism is completed.The comparison of the pose error before and after calibration show that： the maximum position error is reduced by 58%-96%， the maximum attitude error is reduced by 92%-97%.Using the orthogonal displacement measurement system to calibration the parameters of the parallel mechanism can not only improve the position accuracy of the parallel mechanism ，but also simplify the calibration work， improve the calibration efficiency and reduce the cost.Keywords：6-DOF parallel mechanism;orthogonal displacement measurement system;parameter identification;optimization algorithm;Optimization Assisted System Integration Software（OASIS）- 144
- |
- 45
- |
- 1
Published：2021-03-05 -
Two-dimensional abbe error analysis and modeling of CNC machine tool XY table unidirectional movement
Abstract：In this study， an error compensation model is established for improving the machining accuracy of computer numerical control （CNC） machine tools. The error correlations between the various error components of the XY table rail system of the CNC machine tools are studied under the influence of temperature. First， the correlation between the translation error and angle error and the variation law of the two-dimensional Abbe error of the x-direction guide rail system are analyzed. Then， an Abbe error calculation model is established for the y-axis and x-axis guide system error correlation. Finally， the correlation between the x-axis and y-axis guideway angle errors is verified， and the positioning error compensation effect of the guide rail system is investigated. The experimental results show that the overall error of "positioning error 2" after compensation by using the two-dimensional Abbe error of the workbench shows a decreasing trend； the maximum difference is approximately 3 μm from that of "positioning error 1" after compensation by the traditional method， which accounts for one-third of the total error in the compensated workbench. The established two-dimensional Abbe error model of the guideway system is more suitable for identifying the actual variation principle of the XY worktable error for CNC machine tools. The compensation effect for the worktable linkage positioning error based on the model guideway is better than that based on the traditional compensation method. Thus， this study lays a theoretical foundation for the real-time compensation of CNC machine tool errors and can be used to improve the machining precision and measurement accuracy of on-machine measurement systems.Keywords：Computerized Numerecal Controlled（CNC） machine tool;stacked worktable;two-dimensional abbe error;correlation;Error modeling- 164
- |
- 39
- |
- 2
Published：2021-03-05
Micro/Nano Technology and Fine Mechanics
-
Blind image deconvolution via cross-scale dictionary learning
Abstract：Blind image deconvolution recovers a sharp image from a blurred image when the blur kernel is unknown. To solve this underdetermined inverse problem， most existing methods exploit various image priors to constrain the solution. In this study， we propose a blind deconvolution method based on cross-scale dictionary learning， in which the down-sampled blurry image is used to learn a dictionary as training samples and the texture region is represented sparsely over the dictionary as the regularization term. Because the down-sampling process weakens the blur of the image， it will result in the formation of redundant cross-scale similar patches. To ensure that a sharp image is represented sparsely， sharper image patches from the down-sampled image in this study were used to learn the dictionary as training samples. The results showed that the sparse representation error of the texture patch from the sharp image was less than that from the blurred image， further diminishing the sparse representation error over the dictionary， and the intermediate latent image approached the sharp image. The mean peak signal-to-noise ratio of the results by our method on the dataset of Kohler et al. is 29.54 dB. Experimental results on blurry images demonstrated that our method can estimate large blur kernels accurately and that it has good robustness.Keywords：blind deconvolution;sparse representation;dictionary learning;cross-scale;self-similarity- 172
- |
- 33
- |
- 1
Published：2021-03-05 -
Low illumination panoramic image enhancement algorithm based on simulated multi-exposure fusion
Abstract：Panoramic images captured under low-illumination conditions suffer from low contrast and poor visual effects. To address these problems， we propose a low-illumination panoramic image enhancement algorithm based on simulated multi-exposure fusion. First， the original image is converted to HSV color space； then， the optimal exposure rate is estimated by using a metric of image information entropy， and the V component is enhanced by using an intensity transform function to obtain an overexposed image. Second， a medium-exposure image is generated by using an exposure interpolation method， which utilizes the low-light image and overexposed image as input. Third， the fused image is obtained by employing a multi-fusion strategy in which the original low-illumination image， medium-exposure image， and overexposed image are fused. Finally， the detailed information is enhanced by using a multi-scale detail boosting method. The proposed method exhibits better performance compared with NPE， LIME， SRIE， Li， Ying， and RtinexNet algorithms. In case of panoramic images of different scenes， the lightness order error is 322， natural image quality evaluator is 2.32， blind/referenceless image spatial quality evaluator is 5.71， and structure similarity index is 0.82. The comprehensive performance of the proposed method is found to be better than that of other comparison algorithms. Experimental results show that the quality of the low-illumination panoramic image can be improved effectively by using the proposed algorithm.Keywords：image enhancement;low-illumination panoramic image;multi-exposure fusion;exposure interpolation;image entropy- 230
- |
- 49
- |
- 2
Published：2021-03-05 -
Salient region weighted correlation filter for object tracking
Abstract：To improve the positioning accuracy of target positions in object tracking， an object tracking algorithm based on a salient region weighted correlation filter is proposed in this study. Using the tracking framework of efficient convolution operators （ECO） for tracking， we first apply SE-ResNet， which is a pre-trained improved residual network， to extract the multi-resolution features of different layers and fully utilize the different characteristics of the shallow and deep features to enhance feature expression. Next， a background object model is used to obtain a saliency map of the target. The saliency map is then applied to weight the response map of the correlation filter to improve positioning accuracy. Finally， compared with eight popular tracking algorithms employed at the Visual Object Tracking （VOT） challenge， the expected average overlap scores of VOT2016 and VOT2017 are determined to be 0.415 7 and 0.341 2， respectively， which are better than those of the other algorithms. Experimental results show that the proposed algorithm can effectively improve positioning accuracy and tracking performance.Keywords：convolutional neural network;correlation filter;salient region;object tracking- 189
- |
- 45
- |
- 0
Published：2021-03-05 -
Building outer boundary extraction from ALS point clouds using neighbor point direction distribution
Abstract：Factors such as uneven point density and various irregular and complex shapes of buildings are common in airborne laser scanning point clouds and cause difficulty in setting parameters and low adaptability of existing building-outline-extraction methods. To address these problems， a novel building-outer-boundary-extraction method using a neighbor point direction distribution is proposed in this study. First， a specific number of neighbor points are used to analyze the neighborhood point direction distribution to obtain the angles between different directions. The concept of a potential boundary point is defined， all potential boundary points are regarded as boundary points based on the characteristics of the building outlines， and the initial building boundary points are retrieved. Because disorder boundary points are difficult to use in actual tasks， a triangulated irregular network is constructed to obtain the edges between points， and processes such as deletion and addition are performed on the edges to acquire edges that link only boundary points. After the deletion of points that cause obvious zigzag， ordered and smooth building boundaries are finally retrieved by point tracing based on unfixed-length edge scanning. Multi-group of simulated and real-scene building point clouds with different point density distributions and shapes are employed in the experiment. The results show that the proposed method can achieve good boundary extraction when the same parameter is set in different scenes. In addition， the F1 scores are found to be higher than 90.88%. The proposed method can ensure high F1 scores of extracted boundaries while effectively overcoming the difficulties in setting parameters and low adaptability. The method can provide stable and reliable building outline information for applications such as three-dimensional building model reconstruction.Keywords：building outer boundary;neighborhood direction distribution;extraction and tracing;edge scanning;airborne LiDAR- 137
- |
- 28
- |
- 1
Published：2021-03-05 -
Research on fundus image classification based on transfer learning and feature fusion
Abstract：Given the many fundus images that must be collected and the uneven distribution of experienced ophthalmologists， which lead to low accuracy and lengthy examinations of patients with fundus diseases， this study proposes an image classification method based on transfer learning. We first modified and fine-tuned the EfficientNet-B0 and EfficientNet-B7 models for later use as feature extractors on fundus images. Feature fusion was then performed and a deep neural network classifier was finally implemented to detect abnormal fundus. In addition， a visual representation that used gradient-weighted class activation mapping was produced to explain why the model predicted that the fundus images would be abnormal. The proposed method obtained an average accuracy， sensitivity， and area under curve of 95.74%， 96.46%， and 0.987， respectively， for internal data using a 10-fold cross-validation. It also achieved an accuracy of 97.04% and sensitivity of 97.14% on the public JSIEC dataset. The results demonstrated that this method can be used for large-scale screening of abnormal fundus and can assist doctors in performing efficient diagnoses.Keywords：fundus images;deep neural network classifier;EfficientNet fine-tune;feature extract;feature fusion;visual explanation- 209
- |
- 52
- |
- 5
Published：2021-03-05 -
Single image dehazing combining sky segmentation and transmission mapping
Abstract：To solve the contour effect and color distortion problems in the sky area of the dark channel prior algorithm， a dehazing algorithm for sky area segmentation and transmission mapping of different areas is proposed. First， the sky area of an image is roughly segmented using the adaptive threshold method， and the atmospheric light value is estimated in the sky area. Second， the dark channel is improved by applying the super-pixel segmentation method to obtain the initial transmission， and refined transmission is obtained using the guided filtering method. Adaptive threshold segmentation is performed on the refined transmission， and the largest connected domain is retained to achieve fine segmentation of the sky area. Finally， different transmission mapping methods are proposed for the sky and non-sky areas to obtain the final transmission， and the atmospheric scattering model is used to restore the image. Experimental results showed that the restored image performed well in terms of both subjective vision and objective indicators. This effectively solves the defect whereby the dark channel prior algorithm easily fails in the sky area. The proposed dehazing algorithm can restore a more natural sky and weaken the halo effect in the edge area.Keywords：image dehazing;superpixels segmentation;sky segmentation;transmission mapping;improve dark channel- 163
- |
- 51
- |
- 1
Published：2021-03-05 -
Improved Zigzag traversal and Lorenz chaotic construction of hash
Abstract：To construct a measurement matrix for encryption algorithm based on chaotic system， it only USES simple chaotic initial value to iterate， in order to make the correlation degree of the information of plain ciphertext more closely， it has the ability to resist the attack of selective plain ciphertext.This paper proposes a method to determine zigzag initial position by using plaintext information， which greatly improves the sensitivity of plaintext by forming an association with the original image pixel. In addition， SHA-512， a secure hash algorithm， is adopted to generate the hash value associated with the clear text， which is used to construct the initial value of the system. The measurement matrix in the compressed perception is constructed through the chaotic system， which greatly reduces the space resources and transmission costs caused by the need to store the measurement matrix. In the information entropy， ciphertext sensitivity and correlation coefficient and other characteristics of visual analysis.The realization results show that the information entropy of ciphertext image is 7.998 6， which is closer to 8， and the correlation coefficient is close to 0.It can realize the security encryption of gray image and resist plain ciphertext attack. It has certain robustness， and has strong practicality for gray image encryption.Keywords：Zigzag traversal;Hash function;chaotic mapping;Compressed sensing;image encryption- 122
- |
- 25
- |
- 2
Published：2021-03-05 -
Human action recognition based on improved two-stream spatiotemporal network
Abstract：Traditional two-stream networks cannot capture the temporal relationship in video sequences， which leads to poor performance of temporal-dependent action recognition. To solve this problem， a human action recognition method based on an improved two-stream spatiotemporal network is proposed. First， a convolutional neural network is used to model the temporal relationship by using the temporal shift idea to efficiently capture the spatiotemporal information in the video. Further， an attention mechanism is used to improve the downward learning ability of spatial features caused by the movement of channel information in the temporal dimension. On this basis， a two-stream network is designed that includes spatiotemporal apparent information flow and spatiotemporal motion information flow. Finally， the weighted average method is used to fuse the two streams to obtain the final result. Experiments on UCF101 and HMDB51 datasets exhibited accuracies of 96.3% and 77.7%， respectively. The results demonstrate improved accuracy compared with that of the traditional two-stream network method， which verifies that the proposed algorithm can effectively capture the temporal relationship， enhance the ability of network feature expression， and improve the accuracy of temporal-dependent action and similar actions.Keywords：computer vision;action recognition;two-stream convolution networks;temporal shift;attention mechanism- 199
- |
- 23
- |
- 4
Published：2021-03-05 -
Road extraction technology based on multi-source remote sensing data： review and prospects
Abstract：Extracting road information from remote sensing （RS） images is a necessary step in traditional RS applications， such as detecting land usage and updating geographic information systems. It is also critical for new infrastructures， such as digital cities and intelligent transportation. Based on their development processes and the different data sources， existing road extraction methods for RS images can mainly be divided into high-resolution imaging， multispectral/hyperspectral imaging， laser/point cloud imaging， and synthetic aperture radar （SAR） imaging. In this review， we introduce the application status， applicable scope， and method characteristics of four RS technologies. Additionally， we analyze the development and current application effects of road-based methods on hyperspectral data from different platforms. Finally， we summarize the content of this paper and discuss future development trends.Keywords：hyperspectral;multi-source remote sensing data;road extraction;new infrastructure- 218
- |
- 45
- |
- 1
Published：2021-03-05 -
Image edge detection based on guided filter Retinex and adaptive Canny
Abstract：When machine vision recognizes the edges of metal products， uneven surface brightness can easily cause edges to be misidentified. Traditional edge detection algorithms denoise while also suppressing a considerable amount of edge information， which reduces the quality of edge detection. This study proposes an edge detection method that combines guided filtering-based Retinex and adaptive Canny for metal images. The guided filtering-based Retinex method is first used to obtain the reflection component of a metal image. Next， the image contrast of the reflection component is improved using adaptive gamma correction with weighted distribution， and an adaptive anisotropic diffusion filter is employed to denoise the enhanced image and suppress the noise and low-contrast texture. The improved four-direction Sobel gradient template is then adopted to extract the edges of the image. Finally， the non-maximum suppression and dual-threshold segmentation methods of the traditional Canny algorithm are applied to further refine the edges. The test results showed that when the proposed algorithm was used to detect typical metal parts， the image sharpness index increased from 47.11 in the original image to 68.39， and the brightness standard deviation of the metal surface decreased from 44.76 to 20.16. In addition， the noise assessment index dropped from 1.1 in the original images to approximately 0.15， and the sharpness of the image edges was well preserved. The new method effectively solves the edge misrecognition problem caused by uneven brightness in metal surface images， with the extracted edges exhibiting better connectivity.Keywords：image processing;uneven brightness;metal products;image enhancement;edge detection;improved Canny- 230
- |
- 63
- |
- 6
Published：2021-03-05 -
Combined exposure computational imaging system and image restoration method
Abstract：In traditional optical imaging， when high-speed moving objects are photographed under low-light conditions， balancing the contrasting differences between energy acquisition and high-speed motion blur by configuring the integration time is difficult. To solve this problem， a computational imaging method based on combined exposure was proposed. A high-frame-rate camera was used to acquire two adjacent frames to form a combined exposure image pair. Based on the complementary information between the two frames， the motion blur point spread function was estimated， and the restored image with a high signal-to-noise ratio （SNR） was then recovered. Experimental results show that the proposed computational imaging method can solve the blur problem of high-speed object imaging under low-light conditions. Compared with the original degraded image， the restored image obtained by the restoration algorithm is significantly improved in terms of detailed texture. The objective evaluation indices of peak SNR and structural similarity index measure are also improved by greater than 10% over those of the degraded image. Overall， the performance of the proposed computational imaging method is better than that of the existing non-depth learning restoration method， and the image proved to be clear and reliable， exhibiting a good subjective visual effect.Keywords：computational imaging;image restoration;combined exposure;non-uniform point spread function estimation- 167
- |
- 51
- |
- 0
Published：2021-03-05
Information Sciences
Loading...
0