Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research on Intelligent Learning-Based Multi-Sensor Target Recognition and Tracking System

Author FuYanSheng
Tutor ZhangZuo
School Harbin Institute of Technology
Course Information and Communication Engineering
Keywords object recognition and tracking multi-sensor cooperation intelligent learning multi-thread programming DirectShow
CLC TP391.41
Type Master's thesis
Year 2008
Downloads 186
Quotes 0
Download Dissertation

Object recognition and tracking on video is a challenging subject, in which multi-sensor cooperation detection technique attracts more and more attention. While single sensor provides limited information, multi-sensor cooperation could integrate each information. In this thesis, infrared and visible sensor cooperation-based object recognition and tracking system is mainly researched, which is significant in theory and practical in application to research on how to make use of the redundant and complement information of the sensors, and realize object recognition and tracking continually, accurately and robustly.In this thesis, based on the theory, to solve the problem that object loses under the interferential environment in traditional visible sensor recognition and tracking system, an intelligent learning-based infrared and visible sensor object recognition and tracking system is proposed. Besides the arithmetic, a software experiment simulation platform is established. The main aspects of the thesis are listed below:Firstly, the basic theory of object recognition and tracking is analyzed, emphasizing on the particular filter arithmetic. Based on the theory, an object recognition and tracking system on single sensor is established. Apply weighted color histogram as the object character, combining with the Bhattacharya distance criteria to judge the recognition. Then the arithmetic of object status prediction and tracking is researched. The experiment result shows that under ideal environment, the system is with high recognition accuracy and good tracking robustness. While under interferential environment, the system is with bad robustness and appears lost tracking problem.Then, to solve the problem of single sensor recognition and tracking system, an intelligent learning-based multi-sensor object recognition and tracking system is proposed, of which the key algorithms are particular filter arithmetic on visible sensor and hybrid gauss background modeling arithmetic on infrared sensor, combining with the online learning and updating module and strategy based searching and processing module to judge the final state. Applying infrared and visible sensor for cooperation recognition and tracking according to cooperation criteria for interaction. The experiment result shows that under interferential environment, the system is also with the high recognition accuracy and good tracking robustness.Based on the research of arithmetic, the realization method and structure of system software is presented. Based on the platform of DirectShow technique, combining with multi-thread programming method, the arithmetic of multi-sensor object recognition and tracking system is realized on Visual C++. The platform is established to set up the foundation for the arithmetic updating in the future.

Related Dissertations
More Dissertations