Dahua Technology, which has long committed itself in the areas of AI and deep learning, has introduced new facial recognition solutions targeting different types of users.
Needless to say, artificial intelligence (AI) has become a hot topic these days. A major application of AI is facial recognition, which has become more accurate and effective thanks to better hardware, more data and deep learning-driven algorithms. Amid this trend,
Dahua Technology has rolled out its latest generation of AI-enabled facial recognition technologies to help end users meet their real-life needs, from protecting lives and assets to customer service enhancement.
Facial recognition is not a new technology per se, but with changing times and advances in hardware and software, it has reached a new level of maturity and accuracy, prompting stronger demands from different verticals and market segments. Dahua Technology, which has long committed itself in the areas of AI and deep learning, has introduced new facial recognition solutions targeting different types of users.
Solutions at different levels
According to James Wang, CTO of Overseas Business at Dahua Technology, the company has unveiled face recognition solutions on three levels — cloud, edge domain and edge — each with different applications and use cases. At the highest level is cloud, where feeds from hundreds if not thousands of cameras are processed by Dahua servers equipped with GPUs and advanced facial recognition algorithms. In this architecture, cameras on the front end are equipped with basic face capture and analysis capabilities, sending only metadata to the cloud to reduce the burdens and strains on the network. “Applications at this level include city surveillance where the system can recognize a blacklisted person instantly from a large number of people in a scene, and determine whether that individual is still in the city,” Wang said.
At the edge domain level are face recognition-enabled
IVSS-series NVRs which are capable of recognizing an individual’s gender, age, emotion or other features such as whether he’s wearing a hat or glasses. This, then, can help facilitate post-event investigation by pulling out relevant footage instantly. With 4 to 32 channels, these NVRs are ideal in medium-sized projects. “Examples include retail shops, where the shop owner needs to identify and recognize thieves. It’s also suitable for communities, hospitals and factories — any medium-sized project with security needs,” Wang said.
Finally, at the edge level, Dahua Technology has face recognition-enabled IP cameras, entrance control, and even time attendance terminals with face matching and recognition capability directly inside the camera, which represents an affordable, cost-effective solution for smaller user entities who can use these cameras for access control purposes, Wang said. “Of course the embedded database is smaller in scale. For our cameras we can support up to 10,000 face images and 5 image libraries for face images matching,” he said.
Key enablers
According to Wang, facial recognition has become more mature and advanced than before, due to three primary enablers. To start, today’s computing platforms have become increasingly powerful to run computing-intensive face recognition algorithms. “Companies such as HiSilicon, NVIDIA and Intel have launched chips whose performance is good enough,” he said. “We already have NVIDIA GPUs in our NVRs and Intel Movidius chips in our cameras. Now we’re working with HiSilicon to develop camera chips as well.”
Then, as mentioned, facial recognition algorithms have seen their accuracy significantly improved, especially after the introduction of AI and deep learning with which the facial expression, gender, age, hair color, accessories and emotion can all be better recognized. “Facial recognition includes three key parts: face detection, facial features alignment and feature extraction comparison. If deep learning technology was adopted, the performance of each part would be improved dramatically,” Wang said. “We develop deep learning-enabled facial recognition algorithms both by ourselves and through partnership with different developers. Our strategy is one characterized by openness and cooperation.” Finally, for deep learning to be run effectively, the system needs to be fed with large amounts of data which has become more available today than in the past. “Without good data, advanced hardware and algorithms are basically of no use. As a video surveillance solutions provider, we’ve amassed a good amount of data. We’ve also obtained data from our partners who demand a great user experience out of their system. We use these data to train systems so they can recognize faces with more accuracy and precision.”
Different applications
According to Wang, facial recognition applications encompass both security and non-security arenas. “As far as security is concerned, a primary application is city surveillance where operators need to find suspicious individuals from a large crowd. On a smaller scale, it can be applied in hospitals and schools where individuals seeking to access critical areas can instantly be detected. For retail, they can build a database for thieves and be alerted instantly if a thief has entered the premises,” he said.
“For non-security,” he added, “banks can use facial recognition for customer service enhancement, for example being alerted once a VIP guest has walked in. Schools can use it for time-attendance purposes. It can also be applied to cashless payment to ensure it is really that person who is paying. Finally, it can be used to match a person’s face against his ID card to make sure no one is using a fake ID to buy train tickets or register into a hotel.”
Abundant projects in Asia
Dahua Technology’s facial recognition solutions are deployed in different regions throughout the world. As for Asia, Dahua Technology has projects in Thailand, Singapore, Vietnam and Malaysia, among others. “Some are large projects and some are smaller. For example they have retail shops that need face recognition to catch thieves. Further, the smart city and smart community concepts popular in China are taking hold in those markets as well, triggering facial recognition needs,” Wang said.
In particular, Wang cited a recent case study involving Dahua Technology’s facial recognition deployment. It was the 9th BRICS Summit held in Xiamen, China, from September 3 to 15 last year. The challenge was that the summit demanded high-level protection for national leaders from Brazil, Russia, India, China and South Africa. As Xiamen’s city center is located on an island, the security system required comprehensive surveillance of all roads leading into the area, also known as the “Four Bridges and One Tunnel” network. A solution of this proportion would also require an advanced, automated, and centralized system to ensure efficient, coordinated responses.
To make sure all the areas were well-covered, Dahua provided well over 2,000 cameras, 116 of which were facial recognition cameras to make sure that suspicious or blacklisted individuals could be identified instantly and that no undesirable or unwanted people were able to get into restricted areas. For more details on the case study,
click here.
“Building on its success in providing a comprehensive security solution for the G20 Hangzhou Summit in 2016, the 9th BRICS Summit was yet another accomplishment for Dahua,” Wang said. “National leaders and their delegations were able to smoothly move about the city and attend the conference while Dahua products prevented threats through a vast network of smart video technologies. In addition, the large-scale project significantly upgraded Xiamen’s city surveillance system and provided a long-term foundation for not only public security-related operations, but also broader public projects such as the Xiamen Wireless Image Transmission System.”
Better than ever
Indeed, facial recognition has become a mainstream application that end users find helpful in meeting their security and non-security objectives. Having devoted itself in AI and deep learning research and development for a long time, Dahua Technology has seized this trend to roll out cutting-edge facial recognition solutions targeting a variety of users, from city municipalities to smaller-scale end user entities. With facial recognition technology evolving at a rapid pace, we can expect more cutting-edge products and solutions from Dahua Technology in the near future.