Intelligent analysis industry is imperative

Intelligent video surveillance technology originates from computer vision technology. As a branch of artificial intelligence research, it is an emerging security technology and has broad prospects for development. With the increasing demand for intelligent analysis applications, smart video applications are showing their way in the security field and the future is boundless.

With the increasing demand for intelligent analysis applications entering the 21st century, the issue of public safety has become a major hot spot in the world. The security monitoring system has been applied to all corners of society, affecting and changing people's lives. The rapid development of Internet, 3G wireless communications, and intelligent multimedia processing technologies provide strong support for the development of networked intelligent video surveillance systems. Today, video surveillance systems are increasingly used in various aspects of society and play an extremely important role in various fields such as security, government, banking, education, and transportation. In China, the video surveillance market is rapidly developing. Various monitoring needs have been increasing year by year. The monitoring equipment has also become more and more abundant. People have constantly put forward higher and newer requirements for the monitoring system.

With the technological upgrading of the security industry, intelligent video analysis technology as one of the weapon of market application, manufacturers compete vigorously research and development, improvement, as the core technology for future development. Intelligent analysis technology has been gradually tried out in many fields. However, it has to be said that the current intelligent analysis technology with serious homogeneity and immaturity is underwhelmed by the needs of the gradually subdivided application industry.

Smart analytics security applications Daxian Dao Intelligent video analytics applications can be roughly divided into public security law enforcement and security, Wenbo, and tobacco industry. Law enforcement will focus more on pattern recognition, such as license plate recognition and face recognition. Security will focus on retrograde monitoring, illegal intrusion, crowd gathering, slipping, and the detection of relics. Wenbo class focuses on the protection of objects and personnel. Surveillance can prevent objects from being lost, take away, pick up alarms, and carry out retrograde monitoring on special channels. It can also monitor the number of people and personnel density; tobacco industry applications are biased in behavior analysis, such as detection of remnants in prohibited areas. , combined with the special performance of production equipment failure to detect faults. HD camera smart and non-HD smart use the same principle in the basic algorithm. HD algorithms need to provide more computing power for analysis. At present, the behavior analysis class adopts CIF resolution. When the input is a high-definition video, the pixels are cropped and then analyzed. For the pattern recognition class, high-definition video is used for analysis to obtain more accurate results.

As far as the intelligent video analysis function can be realized, almost all applications are mainly targeted at three aspects: people, vehicles, and objects. Beijing Wenan divides it into three categories according to the specific applications: 1. Public safety prevention. It includes tracking of target movement trajectory, target movement range, target movement direction, special human behavior, monitoring of special vehicle behavior, and so on. Its prominent feature is that it can provide timely warning for suspicious abnormal events. 2, statistical analysis of data. Typical examples are traffic statistics and traffic statistics. This type of application is relatively independent. It uses data output as the final result, and provides multiple types of data reports to assist management decisions. 3, intelligent traffic monitoring. Typical applications such as license plate recognition, illegal monitoring of red lights, and comprehensive surveillance of traffic violations are mainly applied to the analysis of vehicles. They are relatively mature and application cases are relatively common.

In addition, Beijing Monsteel Technologies combines the current status of domestic video surveillance, and specifically developed a class of applications that specifically address the quality of video images in order to monitor the signal quality of each video image. Video video loss, snowflake, scrolling, blurring, color cast, picture freeze, gain imbalance, and Other common camera failures make accurate judgments, helping users to find the front-end camera's video quality faults in time and effectively control the operation of the front-end devices. The normal operation of the monitoring system.

The requirements for video surveillance in different industries generally have very obvious differences, especially for the application requirements of intelligent video analysis technology, which also determines the specificity of the types of detection behavior and abnormal events among different industries. For example, in a safe city, it is possible to develop behavior analysis functions such as fights, robbery, and climb over in response to urban security emergencies. In the bank's ATM self-service area, behavior analysis functions such as illegally pasting small paper strips, installing fake keyboards, masking, and violent robbery can be realized through the analysis of current crimes in ATM self-service areas. In the transportation industry, more functions can be implemented to alert traffic incidents such as retrograde, illegal parking, and traffic jams. Only by combining the practical applications of the industry and in-depth understanding of the specific issues of different industries, can we better grasp the needs of users and put the functions of intelligent video analysis technology into practice. This is also the ultimate reflection of the future industrial value of intelligent video analysis technology. .

For smart video analytics solutions, different industries have different focuses. Taking prisons as an example, in order to prevent the prisoners from escaping from prisons and gang fights, the demand for the perimeter and crowds is relatively prominent, mainly to solve the problems in these two areas, and for some airports and other public places where mobility is relatively large. The demand for the detection and detection of abandoned objects is relatively outstanding. To prevent explosions of dangerous goods and suspicious individuals from committing crimes, such places as nuclear power stations, oil fields, and power grids are generally located in remote areas with limited traffic. However, the perimeter of its security needs are relatively high. Once someone approaches, it needs to cause warnings. The defense needs of the perimeter are more prominent.

For video surveillance, the clearer the image, the more detailed the details, the better the viewing experience, and the higher the accuracy of application services such as smart, so image clarity is the eternal pursuit of video surveillance. In the past, the low definition of video has made it difficult for surveillance personnel to find valuable clues. The use of high-definition video technology has provided us with high-definition, high-quality video sources that contain rich and complete information, thereby improving the intelligent analysis of video. The accuracy rate avoids the loss of information due to scene problems. From this perspective, the intelligent analysis of video is based on high-definition, so the requirements of intelligent video analysis for front-end camera resolution are relatively high. High-definition means that the storage pressure increases, and accordingly, the back-end storage device The stability and capacity also have certain requirements.

There is not much difference between the intelligent video analysis of HD surveillance cameras and the non-HD intelligent video analysis technologies, but the results are completely different. HD monitoring can improve the efficiency of intelligent video analysis, and can obtain more and more effective information from the video. For example, face and license plate recognition require more detail in the image, which can increase the recognition rate while providing more convincing pictures and videos.

Intelligent analysis advantage highlights fast response time: millisecond-level alarm triggering reaction time; more effective monitoring: security personnel only need to pay attention to relevant information; powerful data retrieval and analysis function: can provide fast response time and investigation time.

Motion detection is the foundation: Most intelligent video analysis is based on moving target detection technology, that is, the first intelligent analysis system can accurately detect the moving target, effectively separate the moving object from the background of the image, and extract the moving target information. From the practical application of computer vision, the main challenges and problems that need to be solved in moving target detection, recognition, and analysis can be summed up in three aspects, namely, the robustness, accuracy, and real-time performance of the algorithm.

Robustness: Robustness is the robustness of the system to characterize the insensitivity of the control system to perturbation of characteristics or parameters. The robustness of the moving object detection algorithm is able to achieve continuous, stable detection, analysis and recognition of moving objects under various environmental conditions. The most important reasons that affect the robustness of the algorithm are the following: changes in the state of the target, changes in the ambient light, partial irregularities caused by the target occlusion, and temporary disappearance of the target caused by all occlusion.

Accuracy: Detection and recognition of moving targets For different applications, the detection and recognition rates are different, and it is almost impossible to achieve 100% detection success, that is, there are cases of misdetection and missed detection. Because the actual monitoring scene environment is complex and ever-changing, there are a lot of noise and interference, optimization through the algorithm can improve the detection accuracy, and often can only be based on actual needs, in the false detection rate (false alarm rate) and missed detection. The balance between rates (missing rates) seeks balance.

Real-time: A practical intelligent video surveillance system must have the ability to process video image sequences in real time. Because the processing method of video dynamic image is based on the processing of two-dimensional digital signals, the object to be processed contains a huge amount of data and information, requiring that the algorithm cannot be calculated too complicated and must be fast and real-time. For real-time analysis and warning tasks, the computational complexity is critical so that more resources can be allocated to more advanced tasks. However, real-time and robustness are often contradictory. How to seek balanced development is the key to technology.

The development of intelligent analysis industry is imperative At present, the industrial characteristics of intelligent analysis technology are not obvious, and the actual application effect is naturally not color. Therefore, it is imperative to carry out industrialized development of intelligent analysis technology. The main reasons are as follows:

Illumination changes in the actual environment, complexity of the target motion, occlusion, similarity of the target and background colors, and cluttered background all increase the difficulty of designing the intelligent analysis algorithm. When the environment of the application environment is complicated, and the change of the illumination causes the change of the target color and the background color, the analysis software may cause spurious detection and error tracking. The influence of this illumination change on the algorithm cannot be completely eliminated. In addition, when the moving object in the video image is partially or completely obstructed, or when multiple objects occlude each other, the absence of the target information will affect the stability of the intelligent analysis software in analyzing and tracking. In addition, on the one hand, the current intelligent analysis system must ensure the real-time performance of a large amount of information analysis and tracking, select an analysis algorithm with a small amount of calculation, and at the same time make the analysis algorithm more adaptable to complex backgrounds, changes in illumination, and occlusion, etc. To select a complex analysis and calculation method, it is difficult to satisfy both. Thus, when intelligent analysis technology is applied in various industries, if it is possible to distinguish between application environments and simplification of computing methods, to achieve a single application, specific development for each industry, and embed a special algorithm, or only for a certain kind Or simple analysis of several events, such as personnel tracking of important entrances and exits, the system only needs to embed analysis and tracking algorithms, etc., it will simplify the operation mode of intelligent analysis technology, and the intelligent analysis technology will be more in line with the characteristics of the industry demand, and more For accurate analysis and calculation.

The demand for industrialized development of intelligent analysis technology comes from industrial development and technical limitations on the one hand, and, at the same time, depends more on the actual feedback of actual application effects. At present, intelligent analysis technology has behavior analysis, feature recognition, video diagnosis, classification statistics, etc. The application of intelligent video analysis technology in different industries also has different focuses. For example, the intelligent analysis system in prisons is mainly for cross-border detection, event detection within the region, abnormal behavior identification, etc. The main purpose of this system is to prevent prisoners from being jailed and gang fights, and to maintain perimeter security. For the application needs of road transportation highways and other industries, it is mainly violation detection, traffic statistics, reverse driving, license plate recognition, and traffic incident detection. For public places such as airports and stations where the mobility of personnel is large, the need for detection and detection of abandoned objects is more prominent in order to prevent the occurrence of dangerous goods explosions and suspicious person crimes.

Therefore, it is necessary to carry out effective and reasonable intelligent analysis technology development and application according to the needs of various industries.

Intelligent video surveillance technology originates from computer vision technology. As a branch of artificial intelligence research, it is an emerging security technology and has broad prospects for development. After several years of technological research and market development, intelligent video analysis is no longer unfamiliar to people. The application of smart analysis industry is gradually on the right track.