Research On Underwater Moving Target Detection Method Based On Polarization Image

In the Marine environment,it is difficult to accurately judge the position of moving target by using light intensity imaging in the fields of Marine military,Marine rescue and Marine resource survey,etc.,and the unclear and indistinct target will bring great uncertainty in the Marine work.Therefore,the correct detection of Marine target is an urgent problem to be solved in the Marine field.As a special target imaging method different from light intensity imaging,polarization imaging can obtain more target image information under dark background,effectively solve the problems of detector pixel saturation caused by strong background reflective interference and bright light incident,and effectively improve the ability of underwater target detection.Based on the principle of active polarization imaging detection,an underwater active polarization imaging device is designed in this paper.The device is mainly composed of four parts: underwater environment simulation device,polarization imaging device,polarization active lighting device and image processing system.Due to the short transmission distance of underwater light,irregular attenuation of energy and less polarization component reflected by the target in underwater passive imaging,the quality of passive imaging is poor.Compared with the land environment,it is more difficult to control the direction and position of the light source and to acquire images under water.Therefore,this paper adopts the threeposition observation method of light source and camera,which is side to side,side to side and back to side.The device can collect polarization images of the target under different light source conditions at different positions,and collect polarization images of blue(B),green(G)and red(R)in visible light bands through tunable filter.In addition,the split-focal plane camera in the polarization imaging device can collect real-time video frame images of moving targets,detect the moving targets by using the target detection method,and then detect the target motion direction by combining the output image of the target detection algorithm with dense optical flow method.The experimental results show that the polarization image acquired by active polarization light source has more polarization information than that acquired by non-polarization light source,and the image mean value and information entropy have been improved significantly.In the underwater environment with high visibility,the imaging effect of blue(B)and green(G)band underwater targets is better.In terms of moving target detection,the target with the strongest noise suppression ability is relatively complete in the underwater environment where the light and shade change greatly,while the target with the mixed Gaussian background model is more complete in the underwater environment where the light and shade change little.In the detection of target motion direction,the three-frame difference method has fewer errors and is not sensitive to light and shade changes,so the combination of the three-frame difference method and dense optical flow algorithm has higher robustness.