Research On Key Technology Of Video Surveillance System In Scotopic Vision Environment

With the development of network technology and multimedia technology, the monitor efficiency of video surveillance system is greatly improved. In particular, next-generation video surveillance system proposed the technical requirements of all-weather real-time monitoring, video intelligent analysis and others. For scotopic vision environments, video imaging systems should have good adaptability to the environment and with the function of important information mining. The night-vision system based image intensifier tubes implemented by hardware is common in enhancement technology of scotopic vision; however, the software-implemented enhancement technology to low-light image is rarely reported.In the scotopic vision environment, contrast resolution existed in human vision system is very low. For low-light images, a compensation method based on no-reference image quality assessment system is presented. By using the tool of gradually flattening with grayscale spectrum to analyze grayscale characteristics of low-light images and analyzing the characteristics of the human visual system, low-light image enhancement method which is suited for the human eye's observing is proposed. In order to search for the optimum compensating parameter, no-reference image quality assessment system based on the three parameters of objective assessment is built for evaluating the quality of the compensated image. In order to achieve the rapid compensation in video system, a predicted compensation of functional model based statistical analysis of the optimum parameters in different images is built and it improves the real-time performance of the algorithm. This automatic optimization method is used in rapidly obtaining the best quality image in the scotopic vision environment, especially in real-time processing for compensating video image. For the video applications of scotopic vision environment, video processing platform based the core of TMS320DM642 is built. Using DSP /BIOS real-time operating kernel and RF5 reference frame to design the program can make the video system good real-time.Finally, the vision effects of compensated image achieved by different compensation parameters is subjectively evaluated and compared with other image enhancement methods, the performance of key algorithms proposed in this paper is contrastively analyzed. Experimental results show that the evaluation result of the proposed algorithm is consistent with human subjective perception of image quality. The predicted compensation model is put forward for video processing, the results show that the designed video system can achieve the real-time compensation and improve visual effect effectively. This real-time system can also sample 2-road D1 images and display the comparative effect of original image and compensated image.