Foreword: It is irresponsible . The participation of artificial intelligence has ushered in a top-down change in the entire security industry. The traditional security industry has gradually become intelligent. However, in this new era of advocating artificial intelligence, the security industry pays more attention to the deep integration of technology and application. Artificial intelligence is not just a supplement to camera technology. In the era of new security, the industry will fully utilize artificial intelligence technology to enable security products to realize active identification of dangerous molecules, and security behaviors will be transformed from passive to active. Wang Feng, Product Director, Beijing Bitland Technology Co., Ltd. What is the fifth generation video surveillance? This is the best era, and it is also the worst era. Artificial intelligence has rapidly moved from soft to hard in the security industry, and has turned from a concept into hardware, products, and solutions. Tracing back to the history of security video surveillance industry, along with the advancement of science and technology, the security video surveillance industry has developed all the way through the four eras of the analog era, the digital age, the high-definition era and the intelligent age. In the era of security video surveillance intelligence, the video surveillance industry has completed the city's intelligent monitoring, dynamic face control , face recognition and capture and other related links, but at the same time, how to quickly extract key information from massive video data has been Be the key issue for the next era. In the era of intelligence, users have many pain points, including many videos, large videos, long time, difficult to find, associated, and no statistics. Future video surveillance will not be reduced by the emergence of these problems, but will show explosive growth, and these problems will become more serious. But there is only one solution, and that is video structuring. In other words, what is the era of the fifth generation video surveillance? The answer should be video structuring. Multi-level structuring: realizing video full-target structuring system At present, artificial intelligence combined with high-definition video, cloud platform, big data, image processing, and deep learning technology can be arbitrarily embedded in acquisition, front-end equipment, storage devices, or independently form a rich intelligent product line for analysis equipment, and gradually realize The primary form of video data structuring. Bit-continent's multi-level structuring technology realizes three levels of structure through one-dimensional point, two-dimensional time and four-dimensional space-time, thus realizing the structure of video full target, helping users to experience smarter AI technology application. L1: One-dimensional point structure, including target detection, target deduplication, image selection, image attribute, and object tracking; L2: Two-dimensional time structure, including target attributeization, target classification, target comparison, target retrieval, can be applied to map search, black and white list library; L3: Four-dimensional space-time structuring, including target space-time classification, target trajectory calculation, and regional target analysis, can be applied to space-time deployment control, trajectory tracking, security relay and situational awareness. In addition, multi-level structuring puts higher demands on algorithms and computing power, while Bitland is committed to providing high-performance, low-power dedicated custom chips, and has launched two mass-produced cloud AI chips BM1680 since 2017. And BM1682, based on two chips, developed boards, servers and other products, providing higher computing power for the security video structure. Bitcoin hopes to promote the development of security video intelligence through AI-specific chips, and in the future, it can apply video data like text data.