Research And Application Of Object Recognition And Tracking Algorithm

With the innovations of information and intelligence technology, computer vision technology was widely used all over the world, not only used in security monitoring, military reconnaissance, but also in medical health-care and industrial control. It has played a great role in modern life. The object recognition and tracking technology which based on computer vision acts as an important branch of pattern recognition, and it used in face & fingerprint recognition, also used in many aspects in our life, such as gesture recognition, intelligent security monitoring, intelligent robot, even remote sensing and medical diagnosis.Firstly, this paper introduces the background and meaning of this research, and discusses the research situation of the object recognition and tracking algorithm.Secondly moving target detection and tracking algorithm are introduced and summarized. Some common related algorithms are also introduced in this part.Thirdly, this paper aims to study the feature reduction algorithm in object recognition and describes the classical rough set theory and neighborhood rough set theory. The neighborhood rough set theory is applied in feature reduction of gesture recognition and in this study it can select features which can get better result. In the study of gesture recognition, Hu moments feature is used, the neighborhood rough set theory is employed for feature selection and some unnecessary features are reduced. This can improve the efficiency of the classification and save lots of time. The experiment results confirms that feature selection based on neighborhood rough set theory can restore more useful information of features, reduce redundant information, so it can improve the recognition rate of the gesture recognition.Fourthly, the integrated use of gesture segmentation, gesture recognition and tracking technology implements a human-computer interaction system based on gesture recognition. A method of combining the skin color detection algorithm and motion detection algorithm is put forward to detect gesture area. Using the template matching method to realize the gesture recognition, with the feature of Hu moments after reduction. The Mean Shift algorithm is used to track the hand area, and control of the cursor is realized.Fifthly, we made a research on road monitoring system based on video analysis technology and realized the road monitoring system using object recognition and tracking algorithm. The main research is the traffic parameters and the analysis of traffic incident detection. The moving object detection algorithm is used to identify the vehicle in the road scene, and then the tracking algorithm is used to track the vehicle. Then we can obtain the vehicle trajectory and finally analyze the behavior of the vehicle and the road condition. System detects the illegal parking of vehicles, retrograde, traffic congestion and speeding automatically and rapidly. And at the same time it automatically records and real-time processes the relevant data and feedback the road information to the traffic management department. A vehicle speed measurement algorithm based on lane line is proposed, and the automatic detection of traffic incidents, such as speeding, retrograde, illegal parking, road occupancy rate, traffic volume and so on, are studied.