Reserch And Application Of Tracking Algorithm In Photoelectric Detection System

With modern warfare demands on the stealth combat platform, full range of battlefield intelligence and real-time performance, photoelectric detection system as one of the most important part of the aviation electronic technology become the high-tech which many countries compete to develop. This paper proposed a photoelectric detection system including an omnidirectional infrared search and track system and an electro-optical sighting system. To realize reconnaissance at day and night, target search and tracking, 360-degree full-scene, our system using of ultraviolet, visible, infrared and laser imaging technology widely. This photoelectric detection system can provide data support for target indication and operational effectiveness evaluation of weapon delivery. Visual tracking is used more and more widely in photoelectric detection system. With the development of computing ability, tracking algorithm based on online learning is applied more widely instead of traditional tracking algorithm. But there is still no tracking algorithm which can overcome all problems in the complex environment such as illumination, deformation and occlusion in the tracking process. Based on the online learning of the target tracking algorithm and its application in photoelectric detection system, this paper proposed a tracking algorithm with multi feature fusion based on SVM, which can effectively solve object deformation, illumination changes and camera jitter and other issues during the process of tracking. It is applied to our photoelectric detection system, which provides important technical support for the battlefield situation awareness, missile warning, target search and tracking of airborne electro-optical detection system. Specifically includes the following:Firstly, further research on target tracking algorithm. Studying the current typical TLD tracking algorithm, CT tracking algorithm and Struck tracking algorithm. Realization and completion the simulation of these three tracking algorithms on our data sets from photoelectric detection system. Analysis the advantages and disadvantages through the tracking results.Secondly, this paper add the hog feature in Struck tracking algorithm. With the haar features, raw features and hist features singly or in combination, make up 10 different tracking algorithms for feature extraction. The tracking relationship with feature selection is analyzed by the center location error. We evaluate the Haar&Hog tracking algorithm on the VOT_Challenge2013 platform.Thirdly, we design of photoelectric detection system and data processing scheme. With studying of photoelectric detection system, based on the development of the modernization of operational requirements and domestic and international integrated avionics technology. This paper poposed a photoelectric detection system which can achieve the day and night reconnaissance, target search and tracking, wide field of view imaging, multi-source image fusion of photoelectric detection and data processing scheme. With image fusion, target detection and image stitching technology, this paper realized the application of the tracking algorithm in the photoelectric detection system.