Research Of Tracking Technology Based On Image Processing And Trajectory Prediction

The tracking module of optoelectronic tracking system is mainly composed of three modules: image processing,controller and predictive controller.The introduction of predictive controller is to reduce image processing lag and improve tracking stability.In the tracking process,when the target encounters occlusion problems such as clouds passing through poles,the image processing algorithm with better robustness can not track effectively when occlusion occurs,especially when severe occlusion occurs,or even extract occlusion objects or other false targets.Once the tracking system is converted to the tracking of false targets,the real target will miss the target.In order to improve the tracking performance of photoelectric tracking system in occlusion,this paper starts from the research method of combining image processing with predictive control.When the target is in occlusion,the predictive data provided by the predictive controller is used to track the target.In order to improve the tracking ability of image processing in occlusion,the predictive controller provides predictive data for image processing to ensure that the target is retreating.The occluded object does not need to retrieve the target,but directly restores the normal tracking.Because the key to solve the occlusion tracking problem is to determine the occlusion state by combining image processing and predictive control.Therefore,based on the principle of the algorithm and occlusion tracking results,the anti-occlusion performance of various popular tracking algorithms is analyzed.Because of the lack of occlusion detection mechanism and the lack of target feature information when occlusion occurs,occlusion or false targets are easily extracted when occlusion occurs.Because the template matching method based on Bhattacharyya coefficient has good real-time performance and is suitable for various tracking algorithms,this paper chooses the template matching method based on Bhattacharyya coefficient as the occlusion detection algorithm,but this detection algorithm is greatly affected by the background pixels and the target attitude transformation,so this paper makes two improvements to the original method.The first is to add background to the initial template of the target.Weighting factor is used to reduce the influence of background pixels.The second is to reduce the error detection caused by target attitude transformation by block matching.The simulation results show that the improved occlusion detection algorithm can effectively detect the occlusion state of the target.After deciding that the target is occluded,Kalman's predictive filtering method is used to predict and track the target.The approximate position of the target is synthesized by using the registered encoder information and miss distance data.Kalman predictive filter iteratively extrapolates the approximate position of the target to obtain the position information of the target in occlusion,so as to solve the occlusion problem of the target passing through clouds and rods.At the same time,the predictive controller sends the predictive data to the image processing to improve the occlusion tracking ability of the image processing.After the object exits the occlusion,it can directly recover the normal tracking without re-extracting the object.The experimental results show that the combination of image processing and Kalman prediction can improve the occlusion tracking ability of the system,and the prediction data provided by Kalman predictor can effectively improve the occlusion tracking ability of the image.