Research On Target Tracking Algorithm Under Complex Illumination Conditions

With the wide application of target tracking in civil and military fields,such as airborne warning,automatic target tracking of drones,intelligent video surveillance,human-computer interaction,etc.The target tracking technology has received extensive attention in the field of computer vision.However,in the application of real-world scenarios,the target tracking technology has also encountered many problems.For example,when the target's motion trajectory suddenly changes,the motion speed is too fast,illumination changes,similar object interference,complex background,occlusion,noise,etc.,the tracking is affected to a certain extent,and the ideal tracking result cannot be obtained.This paper focuses on the related illumination problems,to solve the problem of poor tracking accuracy under complex illumination conditions.Main research content:A target tracking algorithm based on real-time compression.Based on the real-time compressed tracking algorithm,the algorithm uses the multi-scale Weber face method to reduce the influence of illumination and improve the acquisition accuracy of the target features.The non-adaptive random projection is used to form the appearance model.A very sparse measurement matrix is used for feature extraction and dimension reduction of the extracted multi-scale image features.The tracking task is implemented by the naive Bayes classifier for binary classification.Combining the Weber face method with multi-scale space,the influence factors of illumination are reduced,and the accuracy of algorithm tracking is improved.The experimental results show that the proposed target tracking algorithm based on multi-scale Weber face and real-time compression can effectively deal with complex illumination and achieve good tracking results.A target tracking algorithm based on LBP and kernel correlation filter.In order to further improve the overall performance of the kernel correlation filter algorithm and improve the tracking accuracy in real-time scenarios.Using LBP robust to illumination changes,the input image is first processed using the LBP operator,which greatly reduces the effects of illumination changes during target motion,and using the kernel correlation filter algorithm can obtain faster computational speed and higher computation accuracy.After experimental verification,the proposed method can reduce the loss of moving targets caused by illumination changes.Compared with many classical tracking algorithms,the proposed algorithm can achieve higher precision under complex illumination conditions and faster computing speeds.