A New Algorithm Of Moving Object Detection In Complex Environment

With the development of modern electronic monitoring technology, various types ofmonitoring such as the human tracking, regional invasion, monitoring of the special objectshave been widely used in airports, train stations, shopping malls and other large spaces, it alsoshows that the intelligent monitoring has gradually into people’s daily lives, and it isincreasingly valued. Moving target detection technology is one of the core technologies in theintelligent monitoring. And it is also the basis of intelligence analysis. Efficient moving targetdetection technology enables accurate extraction of moving objects in video. And this objectscan be effectively analyze by subsequent image processing technology. In summary, movingtarget detection technology is a hot issue in the current electronic monitoring technology.Majority of the target detection technology currently available is based on a simpleapplication environment or a specific application environment, It has a very narrowapplication environment, although such cases have been many effective ways, but it alsohighlights the complex environment of the moving target detection technical difficulties, howto identify a robust motion detection technology is still very difficult. But we should notbelieve that one single algorithm can solve the motion detection in all environments, weshould study one idea for a common environment, but the idea is not necessarily same inevery environment. That the algorithm has broad applicability.Based on the summary and analysis of the relevant research works home and abroad, wemake research on how to detect moving object in complex environment. An algorithm basedon digital image processing technology to detect the moving object in complex environmentis proposed in this paper.In this article, firstly we make a extensive analysis of motion detection, and Focus onthe algorithm with low time complexity. In complex situations, using a simple backgroundsubtraction to extract the prospects with the existence of noise, and then raised prospects for aparticular environment, we find a better algorithm to further extraction for specialcircumstances. We do a lot of tests in different algorithms for different applicationenvironments, and finally we select the algorithm in this paper, you can better cope with the many different application environments.This paper first introduces the basics of graphics image processing technology. In order toimprove the processing efficiency, grayscale images and narrow image is used to determinethe specific moving target object. In order to determine the position of the moving target, thebinarization of image is need. Meanwhile, in order to reduce the noise in the image, two kindsof filtering algorithm, such as median filter and mean filter are introduced. Inaddition, in order to further suppress the noise in the moving target to introduce corrosionmorphological processing, the operation of expansion and opening operation and closingoperation. By pretreatment of the original image and the moving target image, not only toimprove the processing efficiency, and improve the results of the moving object extractionclearly.Then, on the basis of research moving target detection algorithm, this paper brieflydescribes the three basic methods of motion detection technology: the frame-differencemethod, background subtraction method, optical flow method, and we analyze theirapplication. Principle of the three commonly used motion detection algorithm is described indetail and summarize the advantages and disadvantages of the various algorithms and theirscope of application. By contrast, a moving target detection algorithm based on opticalflow method is proposed in this paper.Then this paper proposed an algorithm based on optical flow method and Lucas kanade.Meanwhile, in order to improve the moving object extraction results, the optical flowmethod and the background integration deduction wears. Computing optical flow basedon the pyramid Lucas-Kanade method used in this article, the first to use the harris cornerdetection algorithm to detect corners in the current frame, and then detected corner points asinput to calculate the optical flow of feature points obtained elected after the threshold opticalflow, according to the feature points obtained by optical flow, optical flow on a movingtarget, before the first attractions from time to calculate the bounding rectangle of the opticalflow and background subtraction attractions comprehensive moving target.By analyzing the algorithm and comparison of experimental results, the result showsthat this algorithm in this paper can effectively extract the moving objects which are thebasis for follow-up algorithm to detect. Meanwhile, it is real-time and has goodanti-interference capability.