Research And Application Of Target Detection And Recognition Algorithm

The research of computer vision is to use computer to realize the human cognition, recognition and understanding the real world. Target detection and recognition, that can accurately identify the location of the target and as an important field of research in computer vision, has attracted wide attention in the field of theoretical research in computer vision, and has broad application prospects. However, the complexity of the scene, the attitude of target shape, and the uncertainty of the scale seriously affect the accuracy of target detection and recognition. How to accurately find the target object and ensure real-time, is the focus of attention of researchers at home and abroad.In this paper, the related issues of target detection and identification are studied, and the main work is as follows:1. Object proposes. This paper investigates object propose algorithm, analyzes the characteristics of the algorithm, and in-depth study of the Bing algorithm. In the paper, we modify the score in the first phase of the Bing algorithm and improve the recall rate.2. A fast detection algorithm for pedestrian detection is proposed by applying Bing algorithm to pedestrian detection. The algorithm can detect the pedestrian quickly based on the guarantee of a certain degree of recall rate.3. A hierarchical model based on sparse coding is proposed. The paper introduced the sparse coding histogram into target detection, builds a hierarchical model based on image block, and combines the Hough voting method in algorithm. The method reduce the number of Hough voting image block, improve the real-time performance, while guarantee the degree of recall rate.