Research On Video Tracking Algorithm Based On Fish-eye Lens

In recent years,video monitoring system has been widely used.Monitoring systems usually use multiple camera lenses or rotating camera lenses to increase the monitoring angle.However,these methods were difficult to realize simultaneous and synchronous monitoring.Aiming at this problem,the monitoring system based on fish-eye lens has gained popularity.The angle of fish-eye lens usually reaches or exceeds 180°,and the angle of view was wider than that of ordinary lens,but its imaging process was seriously distorted.The target of fish-eye video not only has the problem of size,posture and occlusion,but also influenced by shape distortion,so it will be difficult to achieve fast and accurate tracking of targets in fish-eye video.The main contents of this paper are as follows:(1)Aiming at the problem that the fish-eye lens distortion is not uniform and the fixed prediction area is easy to cause tracking error,an area prediction method based on fish-eye lens imaging model is proposed.On the basis of the fish-eye imaging model,the motion characteristics of the projection point are derived,and the maximum motion area of the projection point is determined when the space point is arbitrary.Finally,the prediction area of the fish-eye video target is obtained according to the characteristics of the target orientation and the projection point of the previous frame.(2)Aiming at the problem of limited target tracking speed in a simple background,this paper presents a fast and accurate fish-eye video target tracking method.The algorithm first uses OTSU algorithm to segment the prediction area,and uses morphological processing to get some connected domains.Then the size of all connected domains is calculated and compared with the target size of the previous frame,and the target candidate regions are obtained.Finally,the color histogram of the target in the last frame is compared with the color histogram of the current frame candidate region,and the highest position of the similarity is obtained,that is the target position of the current frame.Experimental results show that fast and accurate tracking of fish-eye video in simple background can be achieved.(3)In view of the low contrast between the target and the background,the tracking process is easily disturbed by the background object.In this paper,a video tracking algorithm for fish-eye in complex background is proposed.The algorithm first combines the grayscale features and the relative gradient features to obtain the high-dimensional features of the target,and then compresses the features of the target with the average dimension reduction.Then the prediction area of the target to the vertex is determined according to the motion characteristics of the projection points.In order to adapt to the scale change,on the basis of block matching motion estimation,the vertices of the target tracking frame are located from coarse to fine,and the exact location of the target is calculated by the compression feature and the Bias classifier.Experimental results show that the tracking accuracy of the algorithm is much better than that of other comparison algorithms.