Research On Highway Surveillance Video Anomaly Detection Technology

Video surveillance systems are the important part of the freeway management system. Clearing and normal is the important indicator of monitoring screen. However,there are some abnormalities in real surveillance video such as a missing signal, cast,blur, and camera interference. These abnormalities affect the results of video surveillance systems, and they can’t conducive to detect abnormal events follow the highway. Complex environment impacts existing video anomaly detection method which Real-time performance is bad and difficult to meet the complex scenes and real-time highway monitoring testing requirements. For these reasons, the digital image processing technology to achieve automatic detection of abnormalities highway video surveillance has important academic significance and application value.This paper focuses on the techniques about video camera cast and interference detection, and automatic identification technology about does not meeting the quality standards of the highway video screening.For often appear in expressway video monitoring system, video signal loss and monitor screen freezes signal failure problem, this paper gives the two signal failure detection of the problem based on the variance of image pixel gray value, frame difference. For image color cast problem, we put a cast factor calculation method and give color cast detection method based on three-dimensional RGB color space Cartesian coordinate system. In the RGB color space, we quantify processes each color component and calculate its cast factor to achieve the detection of video surveillance systems cast abnormal events. Experimental results show that the detection method can effectively detect the cast of the surveillance video.For the interference problem often occurs freeway camera pan, deflection, shelter,etc., this paper presents a template matching highway camera detection methods based on dynamic interference Harris corner. Angular point has advantages in describing image location. So, we Use three tectonic grayscale corner to extract Angular point in specific detection area. According to the angular point number and location information,we put forward the dynamic angular point matching template matching algorithm calculation factor to implementation of the interference detection of cameras.Comparing with Evan Ribnick detection methods, the detection algorithm is not only less time-consuming, and has strong adaptability and anti-jamming capability.For diming, whitish, blurring highway surveillance video, this paper presents the Video image sharpness evaluation method based on the angular point neighborhood pixels standard deviation to achieve clarity anomaly detection on highway video image quality. Firstly, this method exclude slants dark or partial white video through calculating the average energy intensity and analyze the video image structure features,evaluate the sharpness of the video image based on the proposed Video image sharpness evaluation method to exclude fuzzy surveillance video. Measured highway video verify the consistency and stability of the method.Finally, we design and complete highway surveillance video anomaly detection system using the above video anomaly detection algorithm, and Write the test software to tested the performance of the algorithm modules, and detect highway tunnels, key sections of the toll plaza scenarios surveillance video. Experimental results show that the system can more accurately detect abnormal highway video surveillance, and can meet the requirements of real-time detection.