The Video Surveillance Research For Several Targets’ Abnormal Behavior

This paper studies typical human violation behavior and based on video monitoring technology typical human behavior and the smoke detection of marine collision. First introduce the knowledge of image recognition, computer vision, video monitor technology, and also the current research situation. As to lighting and shadow impacting the test results issue, we detailed discussed image denoising and enhancement algorithm, which provides a solid basis for video monitoring intelligent analysis.Then introduce the background of the current modeling algorithms combined with practical requirements of this article, and comparing select out the background modeling and region extraction algorithm for this scenario. This body’s abnormal behavior is divided into simple and complex behavior. The former’s detection is based on centroid and aspect ratio with minimum enclosing rectangle of the foreground area. The latter due to the motion complex, centroid detection and aspect ratio change are not competent, so introduces the pyramid optical flow method to calculate optical flow field.Because of the reflection of waves and often accompanied by wind on rivers or lakes, the smoke detection of marine collision is difficult. According to the static and dynamic characteristics of smoke, this thesis makes a rapid detection method based on smoke more than one feature. The suspected smoke area is detected by color model and contour irregularities, and finally through Two Dimensional Discrete Wavelet transformation to obtain the exact region.Based on the research of basic knowledge of image recognition algorithm, this thesis preliminarily realizes the typical abnormal behaviors’ detection of human and marine collision’s smoke, laying the foundation for future research.