Research On Human Detection And Tracking Algorithm Of Infrared Images Series

In terms of safety monitoring,personnel search and rescue,assisted driving,etc.,infrared human target detection and tracking is not affected by the external environment and can work around the clock.However,the further development of infrared human target detection and tracking is restricted by the single infrared imaging information,low signal-to-noise ratio and the non-rigid body characteristics of human body.This paper studies the problem of low detection rate and inaccurate tracking of human targets in various scenes in a static background to improve the detection rate and tracking robustness of human targets.By analyzing the existing infrared human target detection algorithms,it is concluded that the three-frame difference method has a small calculation amount,high real-time performance,and is suitable for the fast motion of the target;the background difference method has a wide range of applications,high accuracy and is suitable for the slow motion of the target.Therefore,in this paper,the above two algorithms are fused to achieve infrared human target detection at any speed.Firstly,obtain the binary image detected by the three-frame difference method and the GMM-based background difference method,and then use the logical operation to fuse.In each connected region,if the results after fusion don't coincide with the results of GMM-based background difference method,it is judged as a false target and eliminated.Experimental results show that,compared with the single three-frame difference method and the background difference method,the method proposed in this paper has a high detection rate and is not limited by the moving target speed.By analyzing the existing infrared human target tracking algorithms,it is concluded that Mean Shift algorithm and the improved inter-frame difference method both have the characteristics of small computation and high real-time performance.In this paper,an improved Mean Shift algorithm is proposed to solve the deviation problem in the tracking process.Firstly,the tracking box containing the target is determined,the outer three layers are divided into four parts: top,bottom,left and right and determine whether to reset the initial tracking box based on the relationship between the pixel value of each part and the target mean,Then,The mean value of the target is used to replace the pixel value of the outer three layers to ensure that the target is surrounded by similar backgrounds.The renderings and error curves under the situation of size change and interference by highlighted objects verify the feasibility of the improved Mean Shift algorithm,but it can not adaptively adjust the tracking box or even tracking failure.Considering the fusion of other algorithms and with the help of a variety of target characteristics will increase the amount of calculation and decrease the real-time performance.In view of this situation,this paper chooses an improved inter-frame difference method that is not restricted by the target features,and adds adaptive updating and template matching parts.Based on the target template,first compare its highlight pixel ratio with the next frame to determine the size change;then obtain the current frames that is enlarged and reduced by the same multiple,and compare the fixed aspect ratio of the two to update the tracking box component;Finally,the minimum variance value is selected from the two matches to determine the true position of the target.The renderings and error curves under the situation of size change,obstruction by trees or interference of highlighted objects show that the robustness,adaptability and anti-occlusion of the proposed algorithm are greatly improved the traditional Mean Shift algorithm,the improved Mean Shift algorithm,and the improved inter-frame difference method.In summary,the human target detection and tracking algorithm proposed in this paper has achieved the expected results in a variety of situations in a static background,which can lay the foundation for subsequent practical applications.