Moving Detection Research Based On Web Image Surveillance

Visual analysis of objects motion is one of the most important research topics in the domain of computer vision; it is also an active field which has interested many researchers in recent years. Objects motion analysis aims at detecting, identifying and tracking objects from image sequences, further, describing and understanding their behaviors. Motion detection belongs to the low-level vision problems, is the basis of other object motion analysis. The technique of motion detection has already obtained a substantial progress through several years' research, but the complexity and multiplicity of the supervising background makes missed detection and false alarms still present, so motion detection can't satisfy people completely and so calls for further research.The motion detection is important processing step to movement image analysis, visible monitoring and visible man-machine. Its goal is withdraws the movement region in the image sequence from the background image. We can find the image movement information through the motion detection. It is significance to simply difficulties of the image analysis and process. However, as the influence of the background image dynamic change, such as the weather, the illumination, the shadow and the chaotic disturbance and so on, motion detection faces to quite difficult work.At present, there are two areas to the motion detection: the artificial neural networks and the image processing method. The majority experts use the image process method to carry on the moving goal detection, the existing quite effective image motion detection method mainly is the flow and frame differencing. Generally speaking, the flow expenses much time, its timeliness and the usability are bad. On the contrary, frame differencing is simple and real-time, so it becomes to apply the most successful motion detection method.Some common approaches of motion detection are introduced in this paper, including the approach based on character, the approach based on optic flow, the approach based on frame subtraction and the approach based on background modeling. Their advantage and disadvantage also with the field they fit are pointed out too. We should adopt the proper motion detection algorithms for specified applications. In this paper, the application is an indoor surveillance. In this case, the video is captured with a fixed camera, and the background's change is slow. The algorithm present in this paper has high real-time quality and availability, can detect motion objects, and is fit for real-time detecting and tracking system.This article improves the algorithm the basic thought is: in the video image sequence, Based on the background differencing and coterminous frame differencing, we operate"and"to the continuously two frame difference images and the background difference image. Then we will carry on binaryzation processing to the result. Finally we can separate the moving goal from the background image, and find a binaryzation image of the movement object in the video frequency sequence image.The steps of correlation testing algorithm are as follows:(1) carries on 3×3 median filter to the sequence image.(2)selects the background image B_k(x,y) from the video frequentcy image sequence, it only to contain the fixed background image;(3)selects the continual two frames in the video frequency image sequence, in which preceding image P_k-1(x,y), current frame image P_k(x,y); (4) computes current frame and the background frame difference FD(x,y), withdraws the goal from the image;(5) computes current frame and the preceding difference FG(x,y), obtains the goal change quantity;(6) to be mixed the frame difference FD(x,y) with FG(x,y) withdraws the movement goal rough movement region of movement goal;(7) mathematic morphology operation, the movement region seal, continuously, the integrity, and removes the noise in the background, and so on.At the same time, its fixed background cannot be irrevocable in the background differencing. But in actual scene, it also has the disturbance by the light changes and so on .Therefore, at the same time we must establish the background updating model to guarantee light change in the background image.This paper uses based on the Karmann filtering background updating model.After this kind of improvement, we can not only protect background model but also rapid response background change. Thus, the background model can update in each time, Through this method, we may obtain the binaryzation prospect image. As a result of the noise and the empty influence, such binaryzation image is not good, therefore we need to carry on the processing to the binaryzation image, discover the omission the spot and rid of the unusual spot.According to the new algorithm, we construct a motion detection model, and act it on a series of testing sequences and some video segments which imitate the input of camera to test this algorithm .The results show that this model can accomplish the pre-establish target of the experiment successfully: Detect moving objects presence or leave availably, in the meantime fully considered the validity incase of complicated background and not rigid body; compare the experimental results to other methods at the same time, the contrast results show that, the method present in this paper is highly robust in detecting small moving object than other methods, and also can effectively and exactly detect moving foreground objects, this method has lower computational complexity, so is fit for real-time transport. But the improvement algorithm requests background and the mobile contrast gradient is bigger than above 5%, some shortcomings in the method of difference have still not been solved.