Smart camera key technology introduction

At present, the composition and hardware technology of smart cameras have been relatively stable and mature. To complete the monitoring tasks and intelligent technologies of smart cameras, software functions need to be closely coordinated. Efficient video codec technology and effective computer vision algorithms are the core of smart cameras. Technology provides an important technical guarantee for the camera to complete intelligent analysis tasks. As shown in Figure 1, the structured output from video capture to intelligent results mainly includes: moving target extraction, moving target tracking, moving target classification and moving target behavior analysis, and structured description.

Figure 1 Intelligent camera analysis process

Moving target extraction

The moving target extraction is the preparation of intelligent analysis. Based on this work, the camera can extract the change region from the image region from the image sequence. The effective extraction of the moving target will greatly reduce the computational complexity of the subsequent process, for the later target recognition and Behavior analysis is of great significance. At present, the most popular methods are background subtraction, time difference method and optical flow method. The most classic global optical flow field calculation methods are LK (Lueas & Kanada) method and HS (Hom & Schunck) method.

2. Moving target tracking

The tracking of the moving target, that is, the process of finding the position of the candidate target region most similar to the target template in the image sequence by the effective expression of the target. Simply put, it is the target location in the sequence image. Effective expression of moving targets In addition to modeling moving targets, the target feature expressions commonly used in target tracking mainly include: visual features (image edges, contours, shapes, textures, regions), statistical features (histograms, various moment features) ), transform coefficient features (Fourier descriptors, autoregressive models), algebraic features (singular value decomposition of image matrices), etc. In addition to using a single feature, the reliability of tracking can be improved by fusing multiple features. Currently, mainstream methods include: region matching tracking algorithm, contour matching tracking algorithm, and feature matching tracking algorithm.

3. Classification of moving targets

The classification of moving objects, as the name suggests, extracts specific types of objects from the detected motion areas, such as different objects such as people, motor vehicles, and people in the classified scene. At present, the more mainstream methods are classification based on motion characteristics and classification based on shape information.

4. Analysis of the behavior of moving targets

Behavior analysis is one of the key objectives of smart cameras, and it is also a key issue in video surveillance in maintaining public safety. Behavior analysis involves multiple fields such as computer vision, pattern recognition, and artificial intelligence. It is based on the low-level processing of the video image sequence, by analyzing and processing the image and video of the monitoring scene, acquiring the information of the monitoring scene or the information of the moving object in the scene, and further studying the properties of the objects in the image and the mutual Relationships, resulting in an explanation of objective scenarios and high-level semantic descriptions, often using neural networks and decision trees for behavioral analysis.