Intelligent video analytics encompass a wide range, and many intelligent video analytics technologies have played an important role in various industries. This paper discusses the structure, content, difficulty, application and development trend of intelligent video analysis technology, and focuses on the application of intelligent video analysis in the video surveillance industry. It is expected to provide reference for the application, research and development of intelligent video analysis technology.
This article refers to the address: http://
Intelligent Video Analysis Overview
What is intelligent video analysis?
Computer vision technology creates a mapping between images and image descriptions, enabling computers to understand the content of video images through digital image processing and analysis. IVS (Intelligent Video Surveillance) mainly refers to the computer "automatically extract and analyze key information in the video source", judge according to certain rules and decide whether to give an alarm. If the camera is seen as the human eye, the intelligent analysis system can be seen 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 a safe city, by 2010, according to the statistics of the Beijing Municipal Public Security Bureau, there are more than 2.7 million types of public security construction monitoring, sharing more than 3 million social resources. In the traditional video surveillance mode, real-time events are monitored by manual monitoring of limited video walls; by manually searching for events that have occurred based on time periods and approximate locations. With so many cameras, a very small number of video walls, manual real-time monitoring can not be considered. Statistics show that manuals can't effectively monitor multiple TV screens. Operators staring at the screen TV monitoring will miss 90% of the video information after more than 10 minutes. Other events will interfere with the monitoring effect (such as phone, chat, etc.). It is difficult and inefficient to retrieve events by hand after an incident, and most of the incidents will miss the past with a slight oversight. In the July 7 bombings in London, more than 100 security personnel spent more than 70 hours working to find the information they needed on a large number of tapes.
In addition, thousands of cameras bring trouble to the maintenance of the management department. How to determine whether each video is working properly becomes a problem; some videos will be shaken by strong winds or vibrations; some videos will be foggy. Become unclear and so on.
The emergence of intelligent video analytics technology is to solve the above problems, it can assist in the diagnosis of video quality, help determine which cameras may have problems; it can analyze the real-time video 24 hours a day without feeling tired or Being disturbed; it helps us intelligently search for content of interest rather than finding videos one frame at a time. However, it does not replace our work. It is necessary to clarify the role played by intelligent video analysis. It can improve the efficiency of work through scientific and technological strength, and make final judgment or artificial.
Intelligent video analysis technology analysis
System structure analysis
Based on the existing video surveillance system architecture, intelligent video analysis systems have different system structures.
For the traditional analog video surveillance system, the external DSP processing host (integrated software license) or the Industrial Computer + video capture card + software license is usually added. For the traditional IP video surveillance system, an external embedded host + software license or server + software license is usually added.
Video quality based analysis
Video quality-based analysis mainly consists of two aspects. Video quality diagnosis is mainly used for equipment operation management. The role in medium and large-scale video surveillance systems is very obvious. Video image enhancement is used for improved visual effects for certain specific occasion.
Video quality diagnosis
Functions: video signal missing, occlusion, sharpness, brightness, noise, snow, color cast, freeze, PTZ motion out of control.
The video quality anomaly is analyzed by a method based on video image comparison, a method of automatic machine learning, and a method of simulating motion instruction image analysis, and an alarm is given to an abnormal camera, which is manually checked and corrected. The automatic learning method of the machine should extract a large number of video clips in the actual video surveillance system, including normal video and video with various faults, form training samples, 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, and the movement of the dome or pan/tilt in the outdoor environment. The design of learning ability is different from the recognition of the human eye. The machine is identified by various parameters. The scene change is more sensitive to the machine. Therefore, the automatic learning adaptability is especially important for the video quality diagnosis system. Training to improve system performance is feasible.
Shenzhen Innovative Cloud Computer Co., Ltd. , https://www.xcycomputer.com