Bullet Spots Capturing And Tracking Based On Video Stream

With the improvement in information degree of our military, automatic target-scoring technology which based on image recognition method has been widely applied in the soldiers’ shooting training. A laser target shooting training system is proposed in this paper. Short focal and large angle of view are the characteristics of fisheye lens. By taking advantage of the characteristics, the volume of target box can be compressed, so it will be easy to carry by soldiers. The system is also equipped with functionalities of laser point real-time detection and tracking, it helps soldiers to improve shooting skills. The related theories and techniques of fisheye image correction and moving target detection and tracking are studied in this paper.Fisheye image correction mainly involves image preprocessing and correction algorithms. Image preprocessing includes image segmentation, edge detection, noise reduction, and mathematical morphology processing technology. The purpose is to calculate the center coordinate of the target. In image correction, after the center coordinate is obtained, the distance between target center and each ring can be calculated. The mapping relationship of the distorted and undistorted distance is approximated by least square method when center of blank is located at optical center. In case of none overlapping, center of blank and optical center are connected with a straight line. The distance between optical center and intersection points of the straight line and each ring are measured, and the mapping relationship is approximated. At last, the image is restored using pixel interpolation. Experimental result shows that it can accurately determine shooting scores with the precision of one tenth of ring width after correction.In moving object detection, because bullet spots are small and have higher gray scale than other parts, the detection and identification algorithm is implemented and then improved according to these characteristics. Experimental result shows that it can correctly detect laser points.Finally, moving target tracking algorithms have been studied, and their advantages and disadvantages are compared. On this basis, Kalman filter is chosen to track laser points. Experimental result shows that it can accurately track laser points. In addition, laser points are likely to overlap with each other and disappear in some scenarios, and the corresponded tracking algorithms are presented.