Study Of Artificial Object Recognition And Tracking Algorithms

This paper introduces the fundamental methods of object recognition and tracking, then airport runway is used as a sample to study the recognition and tracking of artificial objects in depth. A runway auto-recognition algorithm and two runway tracking algorithms are proposed. In the process of recognition, firstly a model is created according to the construction features of runway, then the knowledge-based recognition method from top to down is adopted to complete the runway auto-recognition through hypothesis and examination, in which the main steps are threshold, morphology, Hilditch thinning, Hough transform, parallel lines acquisition, lines fitting and parallel lines selection etc. Here, thresholding and Mean-Shift algorithm are combined together to get a better performance. At the meantime, a template-based lines fitting method is proposed, which has a good performance in the experiments. At last, some other artificial objects such as power-stations and bridges are studied, experiments indicates the effectiveness of the algorithms.In the process of tracking, gray-level correlation-based matching algorithm and feature matching algorithm are tried. A particular template acquisition method for runway is designed, and a template-rotation method is proposed to solve the runway rotation problem. Another tracking algorithm is based on SUSAN operator which has a good ability to detect corners, it uses multi-features detection method to track the four corners of runway to achieve real-time tracking of runway.