Research On Weak Target Detection And Tracking Algorithm In Complex Background

As a public consciousness of humans in the 21st century,environmental protection has also greatly improved about protection of birds,but at the same time,due to the reproduction of birds,the bird's descent flashover,short-circuit of bird's nest material and short-circuit of birds are caused to the transmission line.Traditional bird-repelling methods such as bird-proof plates,bird-repelling covers,sound and light,and ultrasonic bird repellers have played a certain role in preventing birds,but they have problems such as high maintenance costs and equipment loss.The rapid development of computer vision makes intelligent video analysis technology widely penetrate into practical applications.Therefore,based on the computer vision correlation algorithm,this paper acquires and analyzes the bird behavior information near the transmission line,and provides the algorithmic theoretical basis for the research and design of the subsequent intelligent bird repelling device.The main research contents ot this paper are as follows:Weak bird target detection and extraction in complex scenes of transmission lines.The mixed Gaussian background modeling,CodeBook,ViBe(visual background extraction)and PBAS(The Pixel-Based Adaptive Segmenter)algorithms are studied.PBAS is selected as the basic detection algorithm and several optimizations are proposed.Combined preprocessing reduces illumination interferepce,improved foreground/background decision threshold is used to expand the adaptability range of the background complexity,the detected foreground pixels are subjected to a second judgment to ghost,and a counter is introduced to detect the bird target resting on the wire.Compared with other algorithms,the proposed algorithm has higher accuracy in detecting weak birds in complex backgrounds,and can reach more than 96%.Real-time tracking of bird targets near the transmission line can facilitate analysis of the behavior information of the birds,thereby roughly determining the point of landing of the bird-repelling device.In this paper,an algorithm combining Camshift and Kalman filtering is proposed,which can solve the problems of occlusion,fast flight speed and similar color interference in the tracking process.Finally,based on the tracking results,the number of birds in a period of time is used to determine whether a certain area is suitable for the bird-removal device,and a method of combining the virtual detection line and the regional marker is proposed to perform the weak bird target statistics,and the accuracy can be achieved to more than 85%.