Small Moving Target Detection And Tracking Under Complex Background

Moving target detection and tracking has been a hot and key directions in the field ofimage processing and artificial intelligence research. With the target detection scenariosand situations changing more complex, the detection demands becoming increasingly highand the technology deepening, it has become a hotspot in the field of detecting targetsunder certain special conditions. small moving targets detection and tracking undercomplex background is one of them. In recent years, it has been researched and applied inmilitary interception, security monitoring, biological detection and many other area.In this paper, single frame image detection technology and multi-frame imagesequences detection technology for small targets detection have been researched. On thebasis of current detection methods, the detection effects and problems are analysed in thecompare of high-pass filtering, morphological filtering, layerd threshold filtering, wavelettransform, neural network filtering and other classic algorithms. An improved single framedetection algorithm is proposed, weightedly combining the two spatial filtering and layerdthreshold filtering. This method ensures not only the background filtering effect, but alsothat the small target being incorrectly filtered in the case of low SNR.In the multi-frame tracking detection step, by adding the Kalman prediction ofmotion information, the lost problem when small target motion changes in the pipelinefiltering detection has been improved. In the multi-frame tracking detection of smalltargets, pipeline method fully use the correlation of small targets between frames. It hashigh tracking efficiency, but it easily lost the target when the motion changes suddenly. Inthe new method, by separating the motion into X direction and Y direction and predictingthrough those two direction, it makes up the disadvantage of pipeline filter. The result forthe experiment show that the improved method effectively improve the success rate of small target detection and tracking.