Research On Infrared And Visible Video Moving Target Detection Methods

As an essential part in intelligent video surveillance system, moving object detection is the foundation of follow-up target identification and behavior analysis. There are different optical property between visible light and infrared light: visible light has more detail and color information, infrared light display the heat effect of scene. Therefore, we can get better moving object detection result by fusing them. In the paper, we focuses on the question that how to extract moving area in static background video, which contain visible light and infrared light. In order to solve this question, we combine the independent multimodal background subtraction(IMBS) and the ellipsoid color model which fuse visible light and infrared light. Otherwise, we design the moving object detection system for visible video and infrared video which can get the visible video and infrared video synchronously and detect moving object in the video.At first, we discuss the existing moving object detection methods, and focus on the background modeling method. By discussing some classical background modeling methods, we summarize the question and solution in background modeling methods, and propose the independent multimodal background subtraction that we use.Secondly, we discuss the independent multimodal background subtraction(IMBS) method in this paper. As a non-parametric background modeling method, IMBS method can match background color distribution better. By analyzing the IMBS method, we find that this method can't follow the changing of background in some case. To solve this problem, we propose the continuously updating IMBS method. In the test result, we can prove that our method can follow the changing of background better.Next, we propose the solution that using the ellipsoid color model to fuse visible light and infrared light. By analyzing the distribution of infrared light intensity and visible light intensity, we use the ellipsoid color model to match this distribution better. Furthermore, we propose use the ellipsoid color model in the IMBS method to complete the target that detecting moving object in visible video and infrared video. Otherwise, we design a morphology method to remove the potential foreground in the result of IMBS.Finally, this paper introduces the moving object detection system using visible video and infrared video. By using four channels camera, we can get the visible image and infrared image which are matching in pixel. Then, we design the visible video and infrared video recording software, which can get data from camera, display the image and record video which has synchronous visible image and infrared image. Finally, we use this record file to test our method which fusing visible light and infrared light to detect moving object.