Moving Target Capture And Tracking Algotithm Based On OpenCV

With the rapid advancement of machine vision technology,the research on the capture and tracking of moving targets is gradually becoming an important research area in this field.At present,many universities,research institutes and even large-scale science and technology companies have devoted a lot of energy to research and exploration.The research value has been self-evident.The choice of the subject direction of this paper is precisely under such a background to start moving target capture and tracking algorithm.Capture depends on the detection effect,and the traditional moving target detection algorithm can not adapt to many complex scenes,especially in the face of more noise,light conversion and other conditions difficult to meet the requirements of the test results.The traditional target tracking algorithm also applies only to certain situations,speed,background lighting,occlusion and other factors will affect the target tracking effect.To solve these problems,based on the study and study of the traditional algorithms,some algorithms are optimized according to the characteristics of the algorithm and the corresponding problems to make it more adaptive.The work done is as follows:(1)Moving target capture section.The advantages and disadvantages of traditional methods are compared,and a new method based on edge is proposed Moving Target Detection Algorithm Based on Information and Gaussian Mixture.The algorithm makes use of the two adjacent three frames of images in the video for making differential operations or binary operations,morphological processing,canny edge detection on the intermediate frames,and then performing the or operation on the two results.After morphological processing get a more complete silhouette.Mixed Gaussian mixture was used to extract the foreground from the middle frame of the three frames.After binarization and canny detection were performed,the moving targets were obtained after morphological processing and hole filling.Through the comparison of the experimental results,the optimized method can detect a clearer target contour with higher stability and adaptability.(2)moving target tracking part.The traditional CamShift algorithm is analyzed and two optimized algorithms are implemented.One is based on the foreground and the two-dimensional histogram optimization algorithm,using H-S two-dimensionalhistogram to obtain the reverse projection and the detection part of the obtained foreground and get a new projection,the new projection of the CamShift center;The other is the integration of Kalman to correct the CamShift center.Both methods have a better tracking effect than the original algorithm.However,the CamShift-based method has higher target and background color features.In this paper,CamShift is incorporated into the spatio-temporal context algorithm.When the occlusion occurs,the center is modified using the CamShift algorithm to track the center to update the spatio-temporal context to update the local context.It can continue tracking.Finally,through experimental comparison and algorithm analysis,this method has the least fluctuation and the best effect.