Research On Detection And Tracking Of Weak Small Target In Visible Light Image

Birds activity in transmission lines induces bad influence to safety of power supply.According to statistics data in recent years,the total quantity of power line failures caused by birds activities is just less than lightning and external forces,that is to say,it takes third place and goes up gradually.The current common bird repellent schemes are passive ways of repelling birds.For example,Acousto-optic repellent,ultrasonic repellent,pulse repellent.But after a long period of adjustment,birds will adapt to this situation,so that the bird repelling effect is reduced.What’s more,ultrasonic repellent have passive effect on citizen daily life,pulse repellent is harmful to staff those who in charge of maintenance.In view of the repellent is narrowly applicable,and it needs to work continuously and get much expense of mainrenance,we realize a new thought of repellent,and get it a pair of smart eyes,detecting and tracking on flying birds,by video monitoringIn this paper,it give a detailed introduction of the characteristics and difficulties when detection and tracking weak small targets in the visible light image,Summarized and analyzed detection effect of moving target detection algorithm on flying targets.Through precision analysis and actual environmental testing,it proposed the algorithm of adaptive moving average combine with motion history image for weak small targets detection.It used adaptive moving average method to extract the foreground target for the first step,and then use accurate detection of the target which based on the Motion history image of the target in second step.The results indicate that this algorithm is better than the single target detection algorithm in the test environment.The accuracy of single bird target detection is over 90%.The tracking study of bird targets can further confirm the stay of birds in the vicinity of transmission lines.In matters of tracking weak small targets by visible images,we can get the corner feature and color feature as matching features,according to their characteristics of high frequency in the neighborhood,The particle filter combined with Mean-shift tracking algorithm is used as the tracking algorithm in this paper,and the test result indicates that itkeeps excellent effect.In terms of the tracking window of Mean-shift keeps same size,we can judge whether the target position changes or not,by calculating the offset of two adjacent frames in the tracking window,and realize the stay detection of bird target.