Research On Tracking Algorithm Based On Infrared Visible Light Binocular Environment

Object tracking has always been one of the hot spots in the field of computer vision research.In image sequences,a certain algorithm is used to locate and track the object and obtain the object trajectory.In certain situations,the fusion of visible light sensors and infrared sensors complements the advantages of the two and can overcome the poor tracking performance due to image intensity changes or insufficient texture in single sensor object tracking.This paper implements fusion tracking in the infrared visible binocular environment,and divides the process into three parts: preprocessing,image fusion and object tracking.In the image fusion part,the quality of the fused image is improved,and in the tracking part,the scale change and rotation problems are solved.Around the above,the main research work is as follows:(1)Pre-process the visible light image and infrared image to ensure the fusion of the two image sequences.In order to make the two images have the same color channel,the visible light image is grayed based on the HIS model.In order to make the image have a more uniform gray distribution,CLAHE is performed on the infrared image and the visible light image.In order to ensure that the resolutions of the two images are the same,bilinear interpolation is used to change the resolution of the infrared images.In order to deal with the geometric deviation between the two images,the infrared images were image registered.(2)An image fusion method based on improved gray wolf optimization is proposed,which optimizes the obtaining of image weights and improves the information content of the fused image.After the image is decomposed into a detail layer and a rough layer,the detail layer is fused using optimized weights.The range of the optimized weights is determined by complementary edges,which can effectively strengthen the target information and edge information.The optimization weights are obtained through the improved gray wolf optimization,and the cross genetic operation is introduced to the gray wolf optimization to improve the optimization effect.Finally,experiments have verified that the algorithm in this paper has good performance in entropy,standard deviation,spatial frequency and average gradient.And it has better visual effects and less noise than other optimization algorithms.(3)An improved tracking algorithm based on feature points is proposed to solve the problem of scale change or rotation angle change involved in tracking.In the tracking process,feature point matching is introduced.FAST feature points and BRISK feature descriptors are used to describe the object.The feature points and feature descriptors are extracted from each frame of the image to match the object's feature points,and sparseness is combined during the matching process.The optical flow improves the matching,and uses the successfully matched points to estimate the scale and rotation angle of the object.Finally,the center positioning error,overlap rate,accuracy rate,and success rate are used as indicators to verify the advancement of the tracking algorithm in this paper.Under infrared and visible fusion images,the overall accuracy and success rate are higher than other algorithms,which can be very good.Respond to the scale change deformation and partial occlusion of the object.