Target Tracking Based On Infrared And Visible Image Fusion

Visible image cannot work at night and cannot “see”the targets obscured by smoke, while infraredimage can see objects in that environment but with blur edge information, So the trackingperformance can be improved by fusing infrared and visible image.In this paper, correlation tracking based on Kalman predictor and adaptive template update isresearched. The block time is determined by using the correlation coefficient, the possible search areais predicted by using Kalman predictor and the update strategy of template buffer is used to update thetarget templateCamshift has received wide-spread attention and a wide range of applications because it has theadvantages of less computation, anti-target deformation and rotation, anti-block of small size and soon. But it is difficult to rule out interference of similar colors goal and track fast-moving goal only bythis algorithm. This paper presents an improved camshift tracking based on Kalman predictoralgorithm. For the interference of similar color, color target model based on space gradientinformation is employed; for the time of the target block, coefficient of Bhattacharyya and changes oftarget area are used. Kalman predictor is used to predict the search of quick movement target and thetarget location during the process of the block.After studying the algorithm of using Kalman predictor to improve the camshift,target trackingbased on the fusion of infrared and visible image is researched.Firstly, tracking on the basis of fusionof the weighted average of pixel level is studied. Secondly, because of different characteristics ofinfrared and visible light target, tracking algorithm based on feature fusion is used. Gray value andgradient are extracted in infrared image, and information of color and gradient are extracted in visiblelight image. Their respective weights are calculated by the coefficient of each Bhattacharyyacoefficient. Then the final target location is computed according to each weight.The theories, methods and applications of single-mode or dual-mode target tracking based onmulti-sensor are researched comprehensively and deeply. And the algorithm of target tracking basedon infrared and visible image fusion is achieved on this basis.