Several major difficulties faced by video surveillance

At present, in public security work, the most practical application is to find images with the same clue characteristics from massive video data. The biggest technical obstacle is the structure of the video. From the perspective of video analysis and intelligent vision, we can use some intelligent analysis methods to automatically collect some commonly used 'metadata' information from a large number of pure videos, and use the data TAG method to re-process the video information. Organize and store for long-term use in 'wartime' to complete the patchwork of the criminal's behavior. ? The rapid growth of video surveillance data makes traditional video surveillance architecture, data management methods, and data analysis applications face new difficulties. ? Dilemma 1: Contradiction between the rapid expansion of data volume and IT investment ? According to the laws of the IT industry: under the premise of meeting customer needs, the lower the cost of technology, the stronger its vitality. Due to the rapid expansion of the data volume and the increasing demand for large-scale computing, the blind use of high-profile hardware has made hardware investment an unbearable burden for customers, and customers are increasingly looking to meet the requirements. Replace high-end hardware with low-end hardware. ? Dilemma 2: Contradiction between Massive Data and Effective Data ? The camera works 7X24 hours. It is true to record everything that happens in the lens coverage. It is not enough to just record the information, because for the customer, most of the information may be invalid. The effective information may only be distributed in a short period of time. According to mathematical statistics In terms of information, information is presented in a power-law distribution, also known as the density of information. Often, the higher the density of information, the greater the value to customers. ? Dilemma 3: The contradiction between resource utilization and efficiency, the contradiction between serial computing and parallel computing ? After the video surveillance service is networked and connected, more and more devices are on the network. The use of idle computing resources to achieve the maximum utilization of resources is related to the efficiency of computing. In the field of video surveillance, the efficiency of video analysis often determines the value, and lower latency and more accurate analysis are often the common needs of customers such as safe cities. As the amount of data increases, even if TB-level data is used for data analysis and retrieval of video content, the serial computing model may take hours of calculation, which is far from being able to meet the timeliness requirements. Video analysis and retrieval cannot rely on traditional methods. The efficiency of huge amounts of data is optimized. Parallel computing is the way to intelligent video analysis.