Video Structured Technology Application Content

Video surveillance can be divided into security applications that are used to assist the relevant agencies and institutions to improve the protection of the public environment, and as an aid to management and services. Non-security applications are used to improve service levels. The application of video structuring technology is mainly reflected in the identification of vehicle features, portrait features, and behavioral events.

1. Vehicle identification

With the extensive construction and application of intelligent traffic electronic police, traffic bayonet, security bayonet, and parking lot, vehicle-related structured information obtained by using intelligent recognition algorithms for locations such as electronic police, bayonet, parking lot entrances and exits is stored in the vehicle theme bank. By identifying and combing vehicle features, useful information can be extracted to facilitate video tracking, identification, comparison, and ultimately verification of vehicles, thereby realizing intelligent transportation services and management. Vehicle description information includes vehicle number plate, license plate color, vehicle brand, vehicle type, vehicle body color, and the like. The information on the secondary analysis of vehicles includes the identification of seat belts, annual inspection targets, sun visors, pendants, ornaments, and crews.

Models: Determining models will greatly help in narrowing the scope of the search, clarifying the direction of movement, and saving human resources. Judgment of models is mainly based on differences in the external characteristics of different models, such as vehicle nameplates, rear wings, doors, windows, colors, taillight shapes, and brake light positions.

Vehicle license plate: The vehicle number plate is the external identification of the vehicle. It has uniqueness and legitimacy and is an important basis for distinguishing different vehicles. According to the license plate number of the vehicle in the video image, the number can be input to the vehicle information management system, the traffic violation information system, the road toll information system, etc. to inquire, and the vehicle feature information can be obtained to the greatest extent, and the tracking verification.

Individual characteristics: In the analysis of individual vehicle characteristics, first identify the vehicle type based on the vehicle's large characteristics, and then identify the individual vehicle based on its detailed characteristics. The individual identification of the vehicle is mainly based on the painting of the text on the vehicle, the decorative color strip of the body decoration, the hanging of the interior accessories, the placing of the articles, the perfume bottle, the license plate bracket, the anti-static strip on the rear of the vehicle, and the annual inspection on the front windshield. The position of the logo, the payment sign, etc., the advertisement pattern on the rear windshield, and the location of the attachment.

Based on the key vehicle feature information, the vehicle record data is formed, which promotes the development and application of backstage big data analysis services and industrial data mining, forming stealth vehicle excavation, screening of deck vehicles, initial entry into the city, multiple vehicle licenses, and multiple vehicle licenses. , Frequent passing cars, similar vehicles in series, peer cars, high-risk vehicle integration models, vehicle trajectory analysis, space-time collisions and other practical combat methods. On this basis, the search for big data of vehicle feature data can quickly find all the vehicle information that meets the conditions, including driving time and direction, driving speed, vehicle logo, license plate, and year. You can also use the search method of map search to retrieve relevant video and image files on the actual combat platform, quickly query the relevant suspect vehicle information, restore the trajectory history information of the vehicle, and realize the entire journey track of the suspect vehicle in the entire city. Or combined with video surveillance information, to achieve the entire process of visualized track playback of the vehicle, as well as the precise deployment and control of vehicles involved, can also be combined with the public security vehicle management information base to achieve vehicle-to-person correlation.

2. Portrait recognition

The recognition of portrait features is based on video surveillance, supported by computer and network information, and supported by modern information processing technology. It has opened up new patterns of recognition from “image to image” and “from image to person”. The current video surveillance system can structure the pedestrians and their attributes on the screen, allowing the machine to identify some basic attributes such as gender, age range, etc., as well as clothing, sports, backpacks, bags, and umbrellas. , whether to ride a bike and other information for a structured description. If you capture from the front of the target in the video image, you can identify some specific features of the face, such as wearing a mask, glasses, beard type and so on.

Face recognition uses the depth recognition of portrait recognition and intelligent search technology to accurately locate the face of the person in the video and extract facial features. After the face image in the video is automatically detected, feature matching is performed with the background portrait model data. As a result, the identity of the personnel is identified and alerted.

Can feature recognition replace face recognition? There are two characteristics that must be explained: first, feature recognition provides more “false” intelligence than face recognition; second, when people’s characteristics change, their work efficiency decreases. Therefore, feature recognition is more suitable for real-time tracking within a short period of time in the crime scene, but not for long-term tracking. The recognition of portrait features can be used to mobilize the camera to keep track of the corresponding features in situations where the camera cannot capture the human face. This is a way to obtain information when the camera is short-lived and requires low resolution.

Search queries on the basis of human structuring can solve the problem of fast target search. For example, the screenshots of event people are input into the investigation system. Using the function of humanoid retrieval, the system quickly finds the target person’s clothing, color distribution, and posture characteristics. A global search was performed on the multi-cameras near the incident site to find out similar targets, and the results were output in the form of snapshots. The GIS maps were used to conduct space-time analysis and analysis to depict the actor's action trajectory.

3. Identification of behavioral events

Descriptive information on behavioral events includes: behaviors such as violations of traffic light signals, speeding, violations of traffic markings, courtesy of pedestrians, traffic flow statistics, and other behavioral characterizations; human behaviors such as cross-border, regional, obscene, legacy, aggregation, A variety of behavioral characterizations, such as traffic statistics, describe behavioral characteristics such as legacy, loss, and displacement.