AI knows people and has learned "head to toe"

'Will face recognition not work after makeup and face modification?' 'It's ugly to brush your face and pay!' Although nowadays face recognition technology has been widely used, there are still many roast points for the face recognition scene itself from the perspective of industry and users.

According to the reporter, while face recognition technology continues to develop, using other ways to 'identify people' has also become one of the directions of AI development, among which AI technology based on gait, wearing, hair style, posture and other information recognition has also begun to emerge.

'Take two steps' AI will know who you are

Recently, the artificial intelligence enterprise Galaxy waterdrop released the world's first gait recognition interconnection system 'waterdrop insight'. It is reported that the system relies on gait recognition technology, integrates gait database building, gait recognition, gait retrieval, large-scale tracking and other functions, and realizes the real-time intelligent interconnection of gait recognition under a large number of cameras: it supports the real-time concurrency of tens of thousands of cameras, It supports instant retrieval and positioning of massive historical videos and real-time videos, pre-warning, in-process alarm and post tracking, map control and map track tracking.

Mark Nixon, an expert in gait recognition at the University of Southampton in the UK, once said: 'gait recognition is more complex than other biometric technologies, because gait recognition is a biometric feature that contains both physiology and behavior. In model training, gait recognition needs to be trained based on three-dimensional video, while face recognition only needs two-dimensional images.' According to Huang Yongzhen, founder and CEO of Galaxy waterdrop, gait recognition has many unique advantages, such as long-distance, full angle of view, and no need to cooperate. At present, the 'waterdrop insight' system has been applied in Guangdong, Shanghai and other places, and has received good feedback.

Compared with gait recognition, pedestrian re identification (Reid) technology, which integrates more recognition elements, has also attracted attention. According to the reporter of Nanfang Daily, Pengsi technology, a domestic AI enterprise, has made a breakthrough in pedestrian recognition algorithm before. In the test of three mainstream Reid data sets, the key index of the algorithm, the first hit rate (rank-1 accuracy), achieved the best result in the industry. It is understood that as the latest AI recognition technology, pedestrian re recognition refers to pedestrian matching under the non overlapping field of view multi camera network, that is, to confirm whether the pedestrian target captured by cameras at different positions at different times is the same person. As an important supplement to face recognition technology, its development core is that in different videos, under the premise of unable to obtain clear face feature information, the machine recognizes the same person through wearing, hair style, posture and other information, so as to enhance the spatial-temporal connectivity of data. Pedestrian re recognition involves computer vision, machine learning, pattern recognition and other disciplines, and can be widely used in intelligent video surveillance, security, criminal investigation and other fields. In the view of insiders, pedestrian re identification has high application value in the fields of public safety, intelligent retail, intelligent transportation and smart city. With the continuous development of AI technology and the promotion and supplement of new technologies, AI is also constantly learning.

AI technology application scenario diversification

From simple face recognition to gait recognition and pedestrian recognition, the continuous learning and evolution of AI technology in 'human recognition' has also greatly accelerated the application of AI technology in many industries.

According to the reporter of Nanfang Daily, judging from the commercial landing of each company in the past two years, 'ai+ security' can be said to be unparalleled. Major public security projects such as safe city and Xueliang project provide a broad market space for it. Based on face recognition, video structuring and other technologies, artificial intelligence took the lead in realizing the landing application in criminal investigation, public security, registered residence and other public security fields. The most well-known is the news that AI assisted the public security organs in catching fugitives at Jacky Cheung's concert. In the view of insiders, it is an urgent need for the current public security system to quickly find the suspect in a large number of videos. According to iResearch data, the average time for human led detection of a single case is 3000 hours of video. Although traditional video surveillance solves the problem of video storage and playback, it is often unable to quickly and accurately identify, locate and find suspects because of factors such as video quality problems or suspects' anti reconnaissance camouflage. According to the introduction of Galaxy waterdrop, the galaxy waterdrop gait recognition Internet system 'waterdrop insight' integrates eight functional modules, including gait capture, gait extraction, gait comparison, gait acquisition, which basically meet the basic needs of the field of public security, and can be widely used in safe cities, important infrastructure, customs and other scenes. It is reported that the gait recognition smart movement developed by Galaxy waterdrop has been applied to smart appliances of well-known brands.

In an interview, Tian ye, general manager of the Security Department of tulin technology, an AI company, also said that the use of AI technology has changed the detection logic of the case. Previously, it relied more on the experience of the case handlers. For example, through the location of the suspect's last appearance, the suspect's next escape direction was judged based on experience, while AI was based on face recognition. No matter which direction the suspect fled to, as long as the suspect appeared within the monitoring range, Through face recognition, we can locate immediately, so as to realize the transformation from post experience judgment to Pre-warning Technology. It is reported that this year, tulin technology has also extended AI technology to broader areas such as justice and communities, such as smart prisons and smart communities.

'In the field of smart retail, Pengsi technology's pedestrian recognition technology can help supermarkets collect the relationship data between 'people' and 'field' and reproduce it in a visual way.' According to the relevant person in charge of Pengsi technology, through pedestrian re identification technology to track and record the action trajectory of customers in the store, analyze hot areas and commodity attention, and analyze the passenger flow stopping rate in different areas, so as to optimize the commodity display and obtain the best passenger flow moving line; Perceive the peak and valley of passenger flow, the most popular area, reasonably allocate personnel and improve services. At the same time, it will also help to achieve accurate offline advertising based on user behavior portraits.

In the view of insiders, with the rapid development of Internet of things, new retail, big data and other technologies, AI is gradually penetrating into urban life through different technical directions and application scenarios, which also provides a broad market and prospect for the development of AI.