With the continuous development of information technology, video information is more and more widely used in entertainment, education, security, life and other fields. This paper introduces the research direction, application fields and technical advantages of face recognition technology, and makes a beneficial discussion on the architecture, key technologies and algorithms of the application of face recognition technology in video surveillance system, especially the technology of correcting face image with rotation angle. Finally, it is concluded that face recognition technology can be applied to monitoring system. The intelligent video surveillance system based on face recognition technology should have a very wide application prospect.
1 application status of video monitoring system
The development of video surveillance system has experienced three stages: the first generation of full analog system, the second generation of partial digital system and the third generation of fully digital system (network camera and video server). The existing digital video surveillance system realizes the digitization, networking and integration of video surveillance means, but it has one main defect: it can only rely on people to judge the video content. At the same time, it is mostly used for “post-processing”, which can not give full play to the initiative of the video surveillance system. Based on advanced biometric technology
The emergence of intelligent video surveillance system for face recognition is another sign of the development of video surveillance system. Intelligent video surveillance system can recognize different objects, find abnormal conditions in the monitoring picture, and send out alarms and provide useful information in the fastest and best way, so as to more effectively assist security personnel in dealing with crises, And minimize false positives and omissions.
2 face recognition technology
2.1 research and application scope of face recognition technology.
Face recognition, also known as face recognition, is not only the basic function of human visual system, but also the most direct means of human mutual recognition. Therefore, it is an important research content in biometric recognition. As a new biometric recognition technology, face recognition technology is an automatic identification technology based on human facial features. Face recognition comprehensively uses many technologies, such as digital image / video processing, pattern recognition, computer vision and so on. Face recognition technology has a wide application prospect in the fields of public security, human-computer interaction and so on. At the same time, face recognition is also a major research topic in the field of artificial intelligence, so it has attracted a large number of researchers to carry out in-depth research. Up to now, it has a research history of more than 30 years. Since the 1990s (especially after the “9 / 11” terrorist attacks in the United States), face recognition technology has made great progress in research and application. The research scope of face recognition can be roughly divided into the following aspects:
(1) Face detection: that is to detect the presence of faces from various scenes and determine their location. In most cases, because the scene is complex, the position of the face is not known in advance, so we must first determine whether there is a face in the scene. If there is a face, then determine the position of the face in the image. Facial hair, cosmetics, light, noise, face tilt, face size change and various occlusion will make face detection more complex. The main purpose of face detection is to find the face region on the whole input image, and divide the image into two parts: 2 face region and non face region, so as to lay the foundation for subsequent processing.
(2) Face representation: that is, a representation is adopted to represent the detected face and the known face in the database. Common representations include geometric features (such as Euclidean distance, curvature, angle), algebraic features (such as matrix feature vector), fixed feature template, feature face, moire, etc.
(3) Face identification: that is, the detected face to be recognized is compared and matched with the known face in the database to obtain relevant information. The core of this process is to select the appropriate face representation method and matching strategy. The construction of the system is closely related to the face representation method. Usually, either the global method or the feature-based method is selected for matching. Obviously, the features selected based on the side image are very different from those based on the front image.
(4) Expression analysis: analyze and classify the expression information (happiness, sadness, fear, surprise, etc.) to be recognized.
(5) Physical classification: analyze the physiological characteristics of the face to be recognized to obtain its race, age, gender, occupation and other relevant information. Obviously, this operation requires a lot of knowledge and is usually very difficult and complex.
2.2 advantages of face recognition technology.
As a new biometric technology, face recognition has unique advantages in application compared with iris recognition, fingerprint scanning, palm scanning and other technologies:
(1) Easy to use, high user acceptance. (2) Outstanding intuitiveness. (3) High recognition accuracy and speed. (4) Not easy to counterfeit. (5) Use universal equipment. (6) Basic information is easy to obtain.
3 architecture of face recognition video surveillance system
Face recognition video surveillance system has four core parts: video processing / face capture workstation, face comparison workstation, blacklist database and alarm display workstation. Video processing / face capture: find the face in the video image, evaluate the image quality and submit it to the face recognition and comparison module; Face recognition comparison module: extract the feature template from the logged in photos and compare it with the blacklist database; Blacklist photo collection: establish a template and add the template data to the blacklist database; Alarm display: display the alarm results according to the comparison results, or transmit the alarm information to PDA or other portable terminals.
4 key problems of face recognition monitoring system
(1) Illumination in face recognition.
Illumination change is the most critical factor affecting the performance of face recognition. The solution of this problem is related to the success or failure of the practical process of face recognition. It is necessary to separate the inherent face attributes from the non-human face attributes such as light source, occlusion and highlight from the face image, and carry out targeted illumination compensation in the face image preprocessing or normalization stage, so as to eliminate the impact of shadows and highlights caused by non-uniform front illumination on the recognition performance;
(2) Face detection and tracking.
Face detection is the preliminary work of face recognition, and face tracking is to continuously track and detect the motion trajectory and contour changes of the target face in the subsequent frames of the motion sequence according to the results of face detection and location. A multi-level structure face detection and tracking system under complex background can adopt face detection technologies such as template matching, feature sub face and color information, which can detect the rotating face in the plane and track the moving face with any pose.
(3) De redundancy problem.
It is required that the face recognition and monitoring system can quickly detect single and multiple face images in the video capture, automatically remove redundancy, subtract duplicate portraits, extract corresponding face image features, realize rapid face comparison, and output corresponding result information.
(4) Pose problem in face recognition.
The pose problem involves the face change caused by the rotation of the head around three axes in the three-dimensional vertical coordinate system, in which the depth rotation in two directions perpendicular to the image plane will cause the partial loss of face information. One scheme is based on attitude invariant features, that is, to seek those features that do not change with the change of attitude. Another scheme is to use a statistical based visual model to correct the input attitude image into a front image, so that the features can be extracted and matched in a unified attitude space.
With the development of biometric technology, face recognition technology is gradually changing from the process of theoretical exploration to the stage of practical application. Professional face recognition products have appeared at home and abroad. Face recognition technology has broad application prospects, and has typical applications in public security, intelligent access control, intelligent video surveillance, public security control, customs identity verification and so on. Among them, the intelligent video monitoring system based on face recognition technology can effectively solve some problems existing in the current digital monitoring system, such as determining whether there are people in the monitoring scene, it is difficult to track the monitoring object, determining the identity of the current monitoring object and so on.
Responsible editor: CT