The Research Of Target State Recognition Technology Based On High-Speed Video

The target state recognition technology is widely used in military,traffic, sports,intelligent monitoring and other fields.It can help the research personnel get the accurate position and attitude information for the weapon test, athletic training and traffic management application field by analyzing the state information of moving objects in real time.It is also helpful for the researchers to analyze the performance index of weapon system in time and accurately and helpful to develop scientific training programs for the sports scientific research personnel. Therefore, the research of attitude recognition technology based on high speed video has been highly valued by researchers in different fields.At present, the attitude sensor and the ground photogrammetry method are mainly used for the attitude recognition of the moving objects. The attitude sensor are mainly used in the environment of airborne, vehicle and so on. The ground photographic method needs to be set with multiple cameras, which can be taken form different angles to deal with it, the system is complex, and the real-time performance is poor. In the weapon test or sports training, it is urgent to need a kind of equipment which is convenient to carry, cheap, and can obtain the target attitude information in real time. Therefore, it has important application value to carry out the research of target pose recognition based in video.In this paper, based on the flying attitude acquisition of a flying saucer as the research object, the paper studies the attitude recognition technology of moving object based on the feature template matching algorithm. The research contents of this paper mainly include the following aspects:Firstly,The overall scheme of motion target state recognition system based on video is studied. Using the CMOS high speed camera shooting moving target image,using the background subtraction method to extract moving targets,extracting the characteristic value of the targets,using the extracted feature vector matching factor and carrying attitude information database matching template,then,the attitude information carried by success of the template is to be tested object attitude information.Secondly,Studying the feature extraction algorithm for moving objects. According to the motion of the target image as the center, the image after binarization processing of the row and column pixel values as feature vectors, feature information of moving object value construction.Thirdly,The control variables used in the 0° to 30° range in the 5° step, in order to change the flying saucer pitch angle, roll angle, yaw angle, and through the CMOS camera under different poses shot the flying saucer image,based on the image acquired by around the three axis of rotation and translation,expansion method, establish a standard temple database.Fourthly,Studying the motion state recognition algorithm. Adopting the template matching technology, taking the test object traversal of eigenvalue difference with the template database feature values,and according to the principle of the nearest neighbor searching, realize the recognition of moving flying saucer in real time.Finally, the performance of each module was tested and verified by the experimental platform. The results show that the technology can realize the real time recognition of flying saucer motion posture in 100fps, and the recognition accuracy is 95%.