In order to detect space small targets from sequence image

a improved dynamic programming algorithm is presented. First

the development and current situation of dynamic programming algorithm used in small targets detection was introduced. Second

Focus on the case of defocus which is common in project for obtaining the sub-pixel center of target

the recursive equation of score function was improved and multi-points accumulation with direction weight algorithm was presented

and the original multispeed plane calculation was also simplified. Finally

the key processes which were encountered in algorithm realization

but skipped over by previous literatures were discussed

include states initialization

constant false alarm rate threshold and track data structure. Analysis indicates that the computational complexity reduced about 50% under the condition of the certain parameters. Experimental results indicate that targets were detected in 5 frames with the multi-points algorithm

in contrast

a number of fake tracks were found until to 10th frame when use the original algorithm. The score SNR of 20 tracks with maximum score in 5th frame of multi-points algorithm is 0.87 higher than that of origin. The multi-points accumulation with direction weight algorithm is able to reduce the false alarm caused by speckle noise efficiently and raise the detection ability of dynamic programming.