Research On Detection Algorithms Of Human Abnormal Behavior In Outdoor Security Monitoring

Abnormal behavior detection is one of the key technologies of intelligent securety monitoring.In the current mainstream abnormal behavior detection methods,Vibe foreground detection algorithm and SURF feature extraction algorithm are widely used.However,the Vibe foreground detection algorithm has ghost areas in the detection process.And it is easy to misdetect the dynamic background as the foreground,that will interferes with the feature extraction.The computational complexity of SURF feature extraction algorithm results in a long time of feature extraction.In order to improve the accuracy and real-time performance of abnormal behavior detection in intelligent security system,the following research is carried out:The three-frame difference method is used to extract accurate background images and establish the initial background model of Vibe algorithm to solve the “ghost” problem.The adaptive threshold based on the dynamic background complexity is used to replace the fixed threshold of pixel classification in the traditional algorithm,which solves the problem that the dynamic background is misdetected as the foreground,and a more accurate foreground image is obtained.The experimental results shown that the improved Vibe algorithm effectively eliminates “ghost” areas,does not appear false foreground targets,and is more adaptable under dynamic background conditions.It can quickly adjust the changes of dynamic background and reduce the false detection caused by the influence of dynamic background.In the Harris corner detection,corner response function correction factor based on similar pixel function is added,which reduces the useless corners and improves the detection efficiency.The improved Harris corner is used to replace the traditional SURF algorithm to extract feature points of interest and reduce the extraction time of feature points.The experimental results show that the improved SURF algorithm is 37.42% of the of the original algorithm in feature extraction runtime,that improved feature extraction efficiency.The test results show that the detection accuracy of abnormal behavior is 91.78%.Compared with traditional abnormal behavior detection methods,the detection accuracy is improved by 2.83%,and the average processing time of each frame is reduced by 1.29%.