Moving Objects Tracking Algorithm For Infrared Image Sequences

Infrared imaging technology works for acquiring infared images from the Infared sencors witch made from the radiation of outside objects.Acording to different objects or different parts of the object generated different thermal radiation,and then gets the different gray scale images.At present,due to the lower and lower price of devices,infared image technology has a rapid development.Its military usage and civilian application has been more and more widely,such as infrared guidance,video surveillance,reconnaissance and security search and tracking as the like.Multi-pedestrain targets detection and tracking of infrared video sequences is an important technology for infrared technology.Unlike the video image sequence under visible light,the signal to noise radio of the image information in the infrared video sequence is low,the noise signal is complicated,and the contrast between the target and the background is small.These features make it difficult to track and detect pedestrian targets in the case of infrared video sequences.In addition,the video multi-target tracking itself is a difficult field in the field of computer vision.For the infrared video sequence,it needs a better algorithm.The main content of this article is to achieve automatic identification and tracking the multi-pedestrian target of infrared video.The multi-target tracking algorithm of infrared video sequence divides the whole process into two parts.The first part is the pedestrian detection algorithm.In this part,we use the cascaded Adaboost algorithm and EOH(edge orientation histogram)features to achieve a fast and robust pedestrian detection algorithm,greatly reducing the target detect time consumption,and get a good detection results.The second part is the tracking part.In this part,we first select the particle filter algorithm as the main algorithm of the tracking algorithm because of the nonlinear non-Gaussian feature of pedestrian movement.Then we improve the particle filter based on the gray image feature of the infrared image,and use the LBP texture feature to discribel the image.The gray scale feature used to establish the observation model,and the tracking algorithm for the multi-pedestrian target of the infrared video sequence designed.Then we can solve the problem of disappearing,occlusion and re-emergence of the targets.This Thesis is the combination of design detection and tracking algorithm to realize the multi-target detection and tracking of infrared video sequences.Considering the real-time performance of the whole tracking process,the algorithm of the detection and tracking improved.At the last,two parts combined,the purpose is to real-time,robust implementation of multi-pedestrian target tracking of infrared Image.