Research On Infrared Target Tracking Method Based On Graph Model

Infrared target detection is a very important research direction in the field of computer vision.The purpose of significant infrared target detection is to find the most informative and interesting area in the image.Infrared target detection has been effectively applied in image retrieval,image segmentation and target tracking.Infrared target tracking plays an important role in military missile guidance,infrared video instrument monitoring and medical treatment.The traditional target tracking method based on boundary box is prone to drift in the process of non-rigid target tracking,while the tracking method based on segmentation can effectively solve this problem.In this paper,the significant infrared target detection based on absorption markov chain is adopted,and on this basis,the infrared target tracking method based on graph model is completed,which is a tracking method based on segmentation.Its main content is as follows:(1)An infrared target detection method based on absorption markov chain is proposed.This is a significant bottom-up infrared image targets detection model,this method will be super pixels infrared image segmentation,will build super pixels into a absorbing markov chain,the maximum entropy of random walk with absorbing markov chain,established the significance,the framework of infrared target detection to absorb the significant test calculation for time.The target of traditional significance detection is usually located in the middle region of the image.In order to solve this problem,this paper USES k-means clustering for pre-segmentation to roughly estimate the target position and remove the target pixel located at the boundary from the absorption node.In this method,multi-scale super pixels are used to fuse detection results in order to increase the effectiveness of significance target detection.The method is tested on three independent infrared data sets,and the experimental results prove the effectiveness of the proposed scheme.(2)An infrared target tracking method based on graph model is proposed.This is a segmentation and tracking algorithm based on absorbing markov chain,in which the target state is estimated by combining top-down and bottom-up methods,and the target segmentation is transmitted to the subsequent frames in a recursive way.In this method,infrared video sequence is preprocessed for super pixel segmentation,and a graph model based on absorption markov chain is constructed for two consecutive frames of infrared images.The background super pixel of the previous frame corresponds to the absorption node,and the weight of each side depends on the features between super pixels and is obtained due to the support of regression vector.The region of the interest of the last frame can be obtained through the foreground region of the previous frame by optical flow method.Once the construction of the graph model is completed,the absorption time of each super pixel is used to segment the target.This method uses two independent data sets to carry on the experiment,the experiment result proves this method is superior to the present method.