Study On Automatic Object Detection And Tracking Based On Image

Automatic object detection and tracking in image is the key technology of the precisely guided weapons, which need to be solved urgently in the world. It is also an active field in computer vision. As an important component of the production of national defense basic project named "The automatic target recognition technology based on the images in the battlefield", the theory, algorithm, and experiment of automatic object detection and tracking are studied in this paper. The template-matching method is respectively realized with two different tools. By doing tests on static photos and motive video, the advantages and disadvantages of template-matching method are analyzed. The method based on Haar feature which could detect object automatically is introduced into millitary fields. The AdaBoost algorithm is developed. Based on the developed AdaBoost algorithm, the training method for military target classifier is established. The precision of classifier training is improved and has been verified by experiments. Speed and anti-jamming ability of automatic object tracking method is improved by merging Kalman filter into CamShift algorithm. Applying these methods built in this thesis, a system with hardware and software, which can detect and track object quickly and automatically, is built. The real-time and efficiency of this system is verified by some object detection and tracking experiments in different scenes such as simple background, complex background, jamming and obstruct. This system could be applied to many fields such as video surveillance, and navigation. In order to transfer the algorithm to DSP and apply to homing head on the missile at last, the design idea of software and the transfer process of algorithm from PC to DSP are proposed. Important theory and technical support are provided for the precisely guided weapons system based on automatic object detection and tracking.