The Key Technique Research For Automatic Recognition Of City Appearance Violation Behavior Based On Image Recognition

As an important part of the management of the city, city appearance violation behavior management gives the city a good appearance and environmental. It has a very important significance. This paper researches the intelligent surveillance video analysis which is used in the city appearance violation behavior.Firstly, the development and research of intelligent video surveillance system as well as the commonly used method in the field of intelligent video analysis are introduced. Then the color space and the denoising techniques which are used in the next step are also introduced.Secondly, a three background subtraction method is proposed. This method constructs two dynamical background and combine the standard background to detect violation behavior. The short time window background model use to remove the foreground information. And a new Gaussian mixture model is used to construct the long time window background. The violation behavior is detected by the two background subtraction. The improved background model is updated by the occurred frequency of different Gaussian models and applicable to the slowly changing scene. Experimental results show the effectiveness of this method.Thirdly, an image adjustment method based on SIFT transform is introduced. This method is used to establish the relationship between the two images through SIFT transform and mark the violation behavior detection region. Then the light and shadow detection algorithm based on HSV color space is obtained. The light and shadow discriminate function is used to detect the light and shadow region. Experimental results show that these methods are fast and accurate, with broad practicability.Finally, the image segmentation method is used in violation behavior detection. A detection method based on image segmentation is proposed to detect garbage hoarding. By using the difference between the background image color histogram, the clustering segmentation is used to detect the garbage hoarding region. Then an image processing method based on region growing and edge detection is applied to completing background subtraction results. And the region growing is limited by the edge distribution. Experimental results are presented to verify the effectiveness of the proposed method.