Different product form and direction of development
Intelligent video analysis system is the inevitable trend of development in the future of video surveillance systems, as a means to obtain useful information from massive data most convenient, will become the main force in the video surveillance system, and with the high-definition, the combination of intelligent, as well as the integrated development of video surveillance cameras and intelligent analysis, its role and impact increasing the air defense of the traditional anti-technology. However, intelligence analysis is not a panacea, the mass analysis still suffered Chen technological bottlenecks.
The form of two types of products:
From the product form of intelligence analysis system is divided into two categories, one is achieved by the intelligent algorithm DSP, common in video server installed in the front-end intelligent analysis of surveillance cameras and intelligent analysis. At present, the more the system in this way, its intelligent analysis capabilities of hardware and software front end video capture. Conventional video surveillance systems, video takes up a lot of storage space and transmission bandwidth, how to solve these problems is the primary face difficulties. A lot of useless video information storage, transmission, it would be a waste of storage space occupied bandwidth, using intelligent analysis aims to alleviate the bandwidth required for video storage space and transmission pressure, or some unimportant video using low stream compression and transmission. In this way, it is more useful to enhance the value of the monitoring system. Algorithm processing is realized by the front end, the rear end of the service pressure is very small, which can be configured in a system a large number of intelligent analysis security surveillance cameras.
The other is the use of back-end PC servers and the operating mode of the intelligence analysis software, like Aimetis, iOmniscient are used in this way. This way because processed by the back-end PC servers, processing performance, superior camera processing front-end intelligent analysis algorithm hardware resources occupied at the same time handle multiple analysis system lack of processing power manifested. Because the back-end PC servers with powerful analytical processing capacity (compared to the front-end DSP software), so the PC server approach is usually very important intelligent analysis occasions.
Two major development direction:
From the main application of intelligent analysis point of view, the two large development direction. One is based on license plate recognition, face recognition as the core represented by intelligent recognition technology, mainly used in electronic police machine Yang, Customs.
The other is a perimeter to prevent the number of statistics, the automatic tracking, retrograde, Jinting rules as the representative of the behavior analysis technology, mainly used in the walled perimeter of the warning area, shopping malls, transportation, attractions, traffic statistics, road Jinting ban, violation retrograde, scene tracking.
1, dual automatic tracking: smart analysis of monitor cameras plus ordinary fastball. Can be applied to the city police emergency plan. Tracking emergencies objects.
2 people traffic statistics: statistics the number of marquee area and out of personnel, applied to supermarket shoppers traffic analysis statistics to help businesses develop appropriate marketing strategies. Applied to the attractions, subway and flow data for personnel control applications.
3, across the warning area: set up a virtual fence, perimeter detection. When found suspicious persons or objects across the fence, that triggered the alarm, and the alarm signal is uploaded to the monitoring and management center. Alarm screen at the same time can be uploaded via the network to remote monitoring users. Applies to traffic road, crosswalk, or zebra crossing, the factory focus area walls, schools, detention center fence and so on.
4, Lost: important items placed on the monitor screen to draw a region as a warning area, as long as the goods leave the warning area, it will immediately trigger an alarm rules. Used in key protected areas such as museums, exhibition halls, auction, gold and silver shop.
5, direction: in the actual monitoring, it may be concerned about the direction of the flow of people and traffic movement direction, can determine whether the target is not legally move around or traveling through the identification of direction, if the reverse behavior, the target will be automatically locked. and alarm at the same time. Applied to the one-way road traveling; important entrances.
6, The intelligent tracking: targeting of suspicious people or objects, to record the trajectory of the target, while security surveillance cameras will follow the target rotation and alarm. Used in the upscale district of the forbidden area, confidential area, important protected area. Can be used as after the incident, the analysis of the trajectory of the case during playback. Achieve rapid detection the role.
Intelligent video analysis system to solve the security personnel are freed from the complicated and boring task "stare at the screen, and this part of the work done by the device; another quick search from the vast amounts of video data to´re looking image. Studies have shown that the operator staring at the TV wall more than 10 minutes will miss 90% of the video information to link this work to lose its meaning, often monitoring system became afterwards the basis for tracing the events. Intelligent video analysis system put after forensics becomes active defense, making the project cost to most provinces, to achieve the most effective security. Intelligent video analysis system is successfully applied in all walks of life, has played a pivotal role.
Four major technical bottleneck:
Is not a video analysis system is a panacea? Which lack its existence? In the actual environment, light fickle, the complexity of the target motion, obscured goals similar to the background color, background clutter will increase the difficulty of target detection and tracking algorithm design. Concrete to look at what impact the application of intelligent analysis of several aspects:
The complexity of the background
Target color and background color changes caused by illumination changes, may cause false detection and error tracking. Different color space can be reduced illumination changes algorithm, but not completely eliminate its impact; conversion of foreground objects and background scene, with the target and background color luggage aside and picked up the start and stop of the vehicle; When will affect the effect of target detection and tracking; goal shadows and background color difference is usually detected as prospects to moving target segmentation and feature extraction difficult.
Target characteristics of the trade-offs
Contains a large number of the image sequence can be used for target tracking feature information, such as target movement, color, edge and texture. Target feature information generally will be subject to change, and select the appropriate features to ensure the effectiveness of the tracking more difficult.
Occlusion is difficult issues that must be addressed in the target tracking. When the moving target is partially or completely blocked, or multiple targets mutually block, the target portion is not visible to cause the target information is missing, to affect the tracking stability. In order to reduce the problem of occlusion ambiguity must correctly handle the corresponding relationship between the characteristics of the target when blocked. Most systems are generally statistical methods to predict the target location, scale, etc., are not very good to deal with more serious occlusion problems.
Both real-time and robustness
The sequence of images contain a lot of information, and to ensure that the real-time requirements of the target tracking, you must select a small amount of calculation algorithms. Robustness is another important target tracking performance, and improve the robustness of the algorithm is to make the algorithm has strong adaptability to complex background, illumination variations and occlusion, which is also the cost of complex calculations.
The above problems, how to improve the effectiveness and practicality of intelligent video analysis system, we can from the following aspects to improve or enhance.
Optimization algorithm to develop strategies for scene analysis, making the accuracy of the algorithm can be increased; reasonable choice of video surveillance cameras installed angle, the accuracy of the analysis, most reasonable depends on the angle; auxiliary supplementary equipment, such as increased fill light lamp or infrared light source, so that can be applied in a variety of complex conditions.
Intelligent video analysis using computer image analysis, to understand the content of the video screen, analyze and track the target within the scene by scene background and target separation. Appear by default in the camera scene analysis rules, once the target is in the scene violate predefined analysis rules, the system will trigger a pre-set linkage rules, so as to achieve the active alarm function.