Heshan Janno Kitchen and Bath Technology Co.,Ltd , https://www.janno-ks.com
From the product form of the intelligent analysis system, it is divided into two types. One is implemented by intelligent algorithms + DSP, which is commonly used in front-end intelligent analysis cameras and intelligent analysis video servers. At present, there are many systems that use this method. The software and hardware with the intelligent analysis function are placed on the video capture side. In a conventional video surveillance system, video occupies a large amount of storage space and transmission bandwidth, and how to solve these problems is the first major challenge. A lot of useless video information is stored and transmitted, which both wastes storage space and consumes bandwidth. The purpose of intelligent analysis is to ease the space required for video storage and the bandwidth pressure required for transmission, or to use low for some unimportant videos. Streaming is compressed and transmitted. In this way, it will help increase the application value of the monitoring system. The algorithm processing is implemented by the front-end, and the back-end service pressure is very small, thus a large number of intelligent analysis cameras can be configured in one system.
The other type is the operation mode of the back-end PC server plus intelligent analysis software, such as Aimetis and iOmniscient. Because this method is used by the back-end PC server for processing, the performance of the processing is better than that of the front-end intelligent analysis camera. Since the algorithm occupies a large amount of hardware resources, when processing multiple analyses at the same time, the system The lack of processing capacity is manifested. Because the back-end PC server has powerful analysis and processing capabilities (compared with the front-end DSP + software approach), the PC server processing method is usually applied to very important intelligent analysis occasions.
Smart Analytics Application Benefits
From the main application of intelligent analysis, there are two major development directions. One is the intelligent identification technology based on license plate recognition and face recognition, which is mainly applied to electronic, mechanical and customs. The other is the behavioral analysis technology represented by rules such as perimeter prevention, population statistics, automatic tracking, retrograde, and prohibition, and is mainly applied to perimeter perimeter perimeters, shopping malls, traffic, and scenic spot flow statistics, and roads are prohibited from being parked. Violation of retrograde, scene tracking and other aspects.
1, dual-track automatic tracking: intelligent analysis camera plus ordinary fast ball. It can be applied to urban police emergency plans. Incident object tracking.
2. Flow statistics: The number of people entering and exiting the statistical box is used to analyze and count customer flow in supermarket shopping malls and help merchants formulate corresponding sales strategies. Applied to attractions, subways, and provide traffic data for personnel control applications.
3. Cross the alert area: Set up a virtual fence to detect the perimeter. When suspicious persons or objects are found crossing the fence, an alarm is triggered and an alarm signal is uploaded to the monitoring and management center. At the same time, the alarm screen can be uploaded to remote monitoring users via the network. It is applied to traffic crossing pedestrian crossings or zebra crossings, plant wall areas, schools, and detention centers.
4. Lost analysis: Draw an area where important items are placed on the monitor screen as an alert area. As soon as the item leaves the alert area, the alarm rule will be triggered immediately. Applied to key protected areas such as museums, exhibition halls, auctions, gold and silver stores, etc.
5, direction analysis: In the actual monitoring, people may be concerned about the direction of the flow of people and the direction of traffic flow, through the identification of the direction can determine whether the target is not legal walking or driving, if there is reverse behavior, the target will be automatically locked, And alarm at the same time.
Roads that apply to one-way travel; important entrances and exits.
6. Intelligent tracking: Targeting a suspicious person or object and recording the trajectory of the target, and the camera will follow the target to turn and alarm. Applicable to high-grade residential areas, personnel banned into the area, confidential areas, important protection areas. And can be used as the incident after the analysis of the trajectory of the case playback process. To quickly solve the case.
The intelligent video analysis system solves the problem that security personnel are relieved from the complicated and boring “eye-screen†task, and the device completes this part of the work; the other one is to quickly search from the massive video data to find the desired image. Studies have shown that when operators stare at the video wall for more than 10 minutes, 90% of video information will be missed, making this work meaningless, and often the monitoring system becomes the basis for subsequent investigations. However, the intelligent video analysis system turns the ex post evidence into an active defense, enabling the project to achieve the most effective security with the most cost-savings. Intelligent video analysis system has played a decisive role in successfully applied to all walks of life.
Intelligent Analysis Technology Application Bottlenecks
Isn't video analytics a panacea? What are the shortcomings? In the actual environment, the change of illumination, the complexity of the target motion, the occlusion, the similarity of the target and the background color, and the disorderly background will increase the difficulty of designing the target detection and tracking algorithm. We can specifically look at several aspects that affect smart analytics applications:
The complexity of the background: changes in the light caused by changes in the target color and background color may cause spurious detection and error tracking. Using different color spaces can reduce the impact of changes in lighting on the algorithm, but it cannot completely eliminate its effects; the conversion of foreground objects and backgrounds in the scene, and the dropping and picking up of baggage, starting and stopping of the vehicle; the color of the target is similar to the background color. It affects the effect of target detection and tracking; the difference between the target shadow and the background color is usually detected as the foreground, which makes it difficult to segment the moving target and extract features.
The trade-off of target characteristics: The sequence image contains a large number of feature information that can be used for target tracking, such as the motion, color, edge, and texture of the target. However, the characteristic information of the target will generally change at any time. Selecting the appropriate characteristic information to ensure the effectiveness of the tracking is difficult.
Occlusion problem: Occlusion is a difficult problem that must be solved in target tracking. When the moving target is partially or completely blocked, or when multiple targets are occluded from each other, the invisible part of the target will cause the loss of target information and affect the stability of the tracking. In order to reduce the ambiguity caused by occlusion, the correspondence between features and targets must be handled correctly. Most systems generally use statistical methods to predict the location, scale, etc. of targets, and they are not able to deal with the more serious occlusion problems.
Considering real-time and robustness: Sequence images contain a large amount of information. To ensure the real-time requirements of target tracking, you must choose an algorithm with a small amount of computation. Robustness is another important performance of target tracking. Enhancing the robustness of the algorithm is to make the algorithm more adaptable to complex backgrounds, changes in lighting, and occlusion, which in turn comes at the cost of complex operations.
To solve the above problems, we can improve or enhance the effectiveness and practicability of the video intelligence analysis system from the following aspects.
1. Optimize the algorithm and develop an analysis strategy for the scenario so that the accuracy of the algorithm can be increased.
2. Reasonably choose the camera installation angle, the accuracy of the analysis, most of it depends on the rationality of the angle.
3. Add auxiliary supplementary equipment, such as adding a fill light or infrared light source, so that it can be applied under various complicated conditions.
Video intelligent analysis system is an inevitable trend in the development of future video surveillance systems. As the most convenient means for obtaining useful information from large amounts of data, it will become the main force of video surveillance systems, and with the organic combination of high-definition and intelligent, With the integrated development of cameras and intelligent analysis, its role and influence will increase, transforming traditional human defense into technical defense.
Intelligent Analysis Technology Application Advantages and Bottleneck Analysis
Intelligent video analysis refers to the use of computer image analysis technology to understand the content of the video screen. By separating the background and the target in the scene, the target that appears in the scene is analyzed and tracked. By presetting different analysis rules in the camera scene, once the target has violated the predefined analysis rules in the scene, the system triggers the pre-set linkage rules to achieve the active alarm function.