Research On Moving Object Detection Algorithm For Countermeasure System

The military developments in the world make it necessary for our army to improve themilitary training. The countermeasure training system referenced here offers the soldiers almostreal combat scenes, allowing them to shoot with real bullet and being able to avoid casualties. It’sa big innovation in military training. Motion detection is an important module of thecountermeasure training. The complex background and soldiers’ camouflage make it difficult todetect targets accurately.Firstly, popular motion detection algorithms are analyzed here, including fame diff, Gaussianmodel, optical flow and ViBe. Compare them in accuracy, time complexity, target integrity andapplicable scene. ViBe is chose as the moving objects detection algorithm for the shooting trainingsystem since ViBe is easy to implement and it has low memory cost, low time cost and gooddetection result. It’s the best method in terms of both computation speed and detection rate.ViBe algorithm is described in details, including the background model’s principle, the pixelclassification process, the background model initialization process and its updating processing.But ViBe isn’t perfect when applied to the shooting training system. Ghost, shadow, little objectand caves in the image will affect the accuracy. An improved method based on optical flow isproposed in this paper to remove ghost. Optical flow computation is high time consuming, but thenew method just needs compute a few pixels’ optical flow. It will not affect the system’s efficiency.In the training scene, there always exist glass, plash which will cause shadow. To suppress shadow,an algorithm based on the pixels’ gradient information in the HSV color model is proposed in the paper. Little objects are removed based on their connected region’s area, utilizing filling algorithmto fill the caves in the image.Finally, a flow chart is displayed to show how the moving objects detection system works.The experimental data shows that the system’s detection result is real-time, robust and accurate.The work in this paper lays a good foundation for target tracking.