Intelligent analysis technology can be divided in two parts of the underlying modeling and high-level modeling. The underlying modeling including background modeling, target detection, target tracking, feature extraction, such as the basic algorithm, high-level modeling is to be judged according to the preset rules to generate analysis results. Now the intelligent video surveillance system has just started, there is no relevant standards can be based on, coupled with intelligent analysis would have been no standard algorithm, there is no standard indicators, so most of the implementation of the project according to the designer's experience and needs of the owners, the system implementation It is a matter of opinion.
Intelligent Video Surveillance System with respect to the traditional monitoring the most essential difference is the addition of information to understand and analyze aspects of theoretically understanding and analysis can occur at all stages of the data flow, it formed a different intelligent analysis solution. Understanding and analysis of information collected in the distal end unit, do give intelligence to a surveillance camera, to form a smart camera, and includes analog cameras and surveillance camera; understanding and analysis of information is performed in the intermediate unit is formed intelligent DVR devices; understanding and analysis of course, can also be in the post- end server to perform a centralized intelligent monitoring system. Different solutions have different characteristics, system design requirements are not the same. With respect to the the backend centralized deployment, we think that the front-end processing, distributed processing more competitive, the following reasons: First, the problem of the processing capability. Intelligent analysis requires the image pixel-by-point analysis, a huge amount of computation, when the input video channels more centralized processing processor bandwidth can not support. Distributed processing processor capacity of a single surveillance camera is not strong, but only handle all the way bandwidth is enough; followed by network bandwidth, centralized processing needs of all video data are transmitted to the server, and distributed processing can be based on an analysis Results selective transmission; Finally, there is the issue of cost distributed processing on the basis of the original hardware without increasing the cost or only very small increase in cost can be realized intelligent analysis of the large ones of the same video is clearly more centralized processing increased costs. To sum up the surveillance ip camera is the best equipped of intelligent video analysis platform.
An the ideal intelligent ip camera
hardware platform should meet the following requirements: computing power, intelligence analysis process includes background modeling, connectivity calculation, morphological processing, are basically on a pixel-by-point operations, but also to ensure processing capacity of more than 10 frames per second, the amount of computation is very large. So powerful processing capability is the basic requirements of the intelligent ip camera hardware platform. At this point, the DSP is significantly better than the SOC processor. SOC is the ARM architecture, ARM processor is characterized by low power consumption and small size, the instruction optimization limited capacity. DSP processors generally have stronger instruction parallel processing capabilities for both the pixel byte operations having specialized instruction. Take full advantage of the parallel processing capability of DSP, the dedicated video instruction and high-speed on-chip memory and reasonable arrangements DMA operation, can greatly improve the computational efficiency of the algorithm; programming flexibility, at this stage intelligent video analysis technology in the process of development, at the same time intelligent video surveillance system implementation depends on the specific scene. Intelligent video analysis there is no standard algorithm, is the integration of multi-algorithm and the adjustment of a particular scene. Programmable processors must have the condition; support advanced video encoding, H.264 video encoding its code rate, the standard uniform becoming the standard in the field of security monitoring. D1 resolution encoder is currently the mainstream configuration. Dual-stream support will become the development direction of the intelligent ip camera; lower cost, cost control is always the need to focus on. Intelligent video analysis functions to achieve the best increase in the original platform based on the few hardware costs even without additional hardware costs.
Although the development direction of the video surveillance system is digital and intelligent, however, no doubt that analog surveillance camera and digital surveillance systems coexist for a long time. Therefore, how to promote the intelligent analog surveillance camera is an unavoidable topic. Can already see in GDSF2009 Conference & Exhibition on such products. Panasonic exhibited an embedded intelligent analysis of analog surveillance camera, it can be set through the menu operating parameters the alarm results achieved through the action of the switch. Analog surveillance camera due to the lack of user-friendly interface, can not be a complex set of rules, more suitable for the realization of regional intrusion detection, the tripwire detection basic intelligence analysis capabilities. On the other hand, the existing video surveillance system including simulation systems, mixed analog-digital systems, digital systems, intelligent transformation of the large number of existing systems, and test of the wisdom of the designers work.