Intelligent video analysis covers a wide range, and many intelligent video analysis technologies have played an important role in various industries. This article discusses the structure, content, difficulties, applications and development trends of intelligent video analysis technology, focusing on the application of intelligent video analysis in the video surveillance industry, and hopes to provide reference for the application, research and development of intelligent video analysis technology.
Smart video analysis overview
What is intelligent video analysis
Computer vision technology establishes a mapping relationship between images and image descriptions, so that computers can understand the contents of video frames through digital image processing and analysis. Intelligent video surveillance (IVS, Intelligent Video Surveillance) mainly refers to the computer "automatically extract and analyze the key information in the video source", according to certain rules to judge and decide whether to give an alarm. If the camera is regarded as the human eye, the intelligent analysis system can be regarded as the human brain.
demand
With the vigorous construction of video surveillance systems in recent years, more and more cameras have been built in various industries. Only in the construction of safe cities, by 2010, according to the statistical data of the Beijing Municipal Public Security Bureau, there were more than 2.7 million various types of monitoring of public security construction and more than 3 million shared social resources. In the traditional video surveillance mode, real-time events are monitored through limited manual monitoring of the TV wall; humans are used to find events that have occurred based on time periods and approximate locations. With so many cameras and a very small number of video walls, manual real-time monitoring simply cannot be taken into account. Statistics show that manual monitoring of multiple TV screens is not effective. Operators staring at the screen for more than 10 minutes will miss 90% of the video information. Other events will interfere with the monitoring effect (such as phone calls, chats, etc.) After an event occurs, it becomes more difficult and inefficient to manually retrieve the event, and most events will be missed if they are negligent. In the July 7 bombings in London, more than 100 security personnel spent more than 70 man-hours to find the required information in a large number of tapes.
In addition, thousands of cameras have caused trouble for the maintenance of the management department. How to determine whether each video is working is also a problem; some videos will be shaken due to strong winds or vibrations; some videos will be shaken due to heavy fog Become unclear, etc.
The emergence of intelligent video analysis technology is to solve the above problems. It can assist in the diagnosis of video quality and help determine which cameras may have problems; it can perform 24-hour uninterrupted analysis of real-time video without feeling tired or Being disturbed; it can help us intelligently search for interesting content instead of finding video frame by frame. However, it cannot replace our work. It must be clear that the role of intelligent video analysis is auxiliary. It can improve the efficiency of work through scientific and technological strength, and make the final judgment or manual.
Analysis of Intelligent Video Analysis Technology
System structure analysis
Based on the existing video surveillance system architecture, the intelligent video analysis system has different system structures.
For traditional analog video surveillance systems, usually add an external DSP processing host (integrated software license) or industrial computer + video capture card + software license. For traditional IP video surveillance systems, external embedded host + software license or server + software license are usually added.
Analysis based on video quality
The analysis based on video quality is mainly composed of two aspects. Video quality diagnosis is mainly used for equipment operation management, and its role in medium and large-scale video surveillance systems is very obvious; video image enhancement is used to improve visual effects and is used for certain specific occasion.
Video quality diagnosis
Functions: video signal loss, occlusion, abnormal sharpness, abnormal brightness, noise, snowflakes, color cast, frozen screen, PTZ motion out of control, etc.
Analyze video quality anomalies through methods based on video image comparison, machine automatic learning, simulated motion instruction image analysis, etc., and raise alarms for abnormal cameras, and manually check and correct them. The method of automatic machine learning should extract a large number of video clips in the actual video surveillance system, including normal videos and videos with various faults, form training samples, and simulate human visual characteristics, and extract a large number of video image feature parameters for different fault types To train the detection system.
In the actual operation scenario, the video quality diagnosis system should automatically adapt to the camera's light changes, scene changes, seasonal changes, various installation angles, dome or pan / tilt motion adaptation through outdoor learning, especially the need to strengthen the automatic The design of learning ability is different from the human eye recognition. The machine is recognized by various parameters. The scene changes are more sensitive to the machine, so the automatic learning adaptability is particularly important for the video quality diagnosis system. Training to improve the performance of the system is feasible.
UPS Accessories,Remote Monitor Card,Smart Card,Intelligent Card
Shenzhen Unitronic Power System Co., Ltd , https://www.unitronicpower.com