|
In digital video processing and computer vision applications, the target detection and tracking is an important, but also the most basic tasks. Currently in target detection and tracking some of the more popular aspects of the application of autonomous vehicle navigation, robot control, motion-based recognition, video compression, vision-based control, man-machine interfaces, medical imaging, augmented reality and video scenes monitoring. In the field of computer vision, object detection and tracking technology has been studied despite more than a decade, but still is a hot research field. Due to the complex background, illumination changes, occlusion and initialize other issues, there is no one universal, accurate, high-performance and real-time target detection and tracking algorithm, moving target detection and tracking effect is still very good, you need to further improved. In this paper, do the following tasks: (a) in the moving target detection research, for complex dynamic scenes of infrared target detection problem, we propose a cross-entropy-based transition region extraction method of infrared moving target detection. This method first uses frame difference and background difference method fusion detection methods, differential processing of infrared images, and then cross-entropy-based transition region binary image segmentation algorithm, and finally morphological filtering to detect the complete infrared target. (2) the moving target detection and tracking process, many factors have led to the result of detecting and tracking deviation, and the shadow is one of the main factors. In this paper, the traditional HSI color model in the current background color close to a moving target and appears when the detection rate is not high drawback has been improved to a degree of density function stepped into HSI color model, using a sequence of images collected experimental The experimental results show that the model based on HSI shadow detection algorithm is effective. (3) In the moving target tracking, this paper, the traditional color histogram target tracking loss situations occur, we propose a Kalman filter based on the weighted color distribution and color target tracking method. The method uses Kalman filter predict the target position and angle, by predicting the direction histogram is calculated adjusting the weighted able to quickly find the target tracking. New method in a certain extent, improve the measurement accuracy and stability, improved tracking results. The above method is used for video tracking and obtain good experimental results further demonstrate the feasibility of the method has some practical value.
|