Raise your hand, aim it at the scanner, and with a "beep," access is granted.
Recently, at the Jinao Mountain subway station on Chongqing's Rail Transit Line 18, a reporter from Science and Technology Daily experienced a new way to enter the station using palm vein recognition technology.
Since last year, the adoption of "palm scanning" recognition technology, supported by palm vein identification, has been accelerating. In cafeterias and sports venues of universities such as Shenzhen University, a simple palm scan now facilitates offline payments and access control. Now, palm vein recognition technology has made its debut on the subway turnstiles, making Chongqing Rail Transit Line 18 the first subway line in China to fully adopt palm vein recognition technology.
Compared to card swiping and facial recognition, what makes "palm scanning" a smarter way to travel? What are its unique features and advantages? And what challenges and difficulties must be overcome to apply it in the field of rail transit? To find out, a reporter from Science and Technology Daily interviewed the team behind turning the idea of "entering the subway with a wave of your hand" into reality.
"The origin of our development of palm vein recognition technology can be traced back to 2015," explains Wang Tianming, founder and Chief Marketing Officer of Foshan Yumai Technology Co., Ltd. That year, Amazon introduced "palm scanning" technology in its retail stores. Inspired by a chance encounter with this technology, the Yumai team conducted further research and decided to develop palm vein recognition technology. They believed that, compared to fingerprint and facial recognition, "palm scanning" would be more convenient, secure, and easy to use, representing a future trend.
Wang Tianming says that vein biometric recognition technology identifies individuals based on the pattern of veins under the skin. The technology uses near-infrared light to illuminate the palm. Hemoglobin in the veins absorbs this light at a wavelength of 760 nanometers, resulting in a vein pattern image. The non-contact, hygienic nature of this technology is seen as its greatest advantage, along with its respect for user privacy through active identification.
Starting in 2018, Yumai assembled a dedicated team and spent six years developing the first generation of their product. "After several years of development, we clearly felt that the time had come to promote palm vein recognition technology and enter the market," Wang Tianming stated.
However, as Xie Qingshan, Director of Industry Sales for Intel's Network and Edge Group in China, points out, "Turning an imaginative scenario into reality requires solid technological support. Every simple technological upgrade in our daily lives is backed by a vast ecosystem and a group of cooperating partners."
The system applied to Chongqing Rail Transit Line 18 was created by Intel in collaboration with Shanghai Huaming Intelligent Terminal Equipment Co., Ltd., Shenzhen Ruibao Intelligent Information Co., Ltd., and Yumai. This integrated Automated Fare Collection (AFC) solution for smart city rail transit combines card swiping, facial recognition, and palm vein recognition, further "upgrading" the rail transit automatic ticket inspection system.
Xie Qingshan told reporters that bringing palm vein recognition technology from the lab to everyday life requires extensive detailed work, such as sufficient computational power, high device stability, and robust software and hardware support for vein image collection, preprocessing, feature extraction, matching, storage, and secure operation. "Therefore, we are particularly pleased to have such excellent upstream and downstream partners in China to promote this initiative." In his view, palm vein recognition technology offers significant advantages in security and uniqueness compared to fingerprint and facial recognition.
Fu Qiang, R&D Director of Shanghai Huaming Intelligent Terminal Equipment Co., Ltd., also expressed confidence in promoting this technology due to the uniqueness and security of palm vein information. For example, palm vein images can better protect user privacy. "As a practitioner, I cannot accurately identify my own image among 100 palm vein images."
According to statistics from the Ministry of Transport, as of December 31, 2023, 55 cities in 31 provinces across China had launched 306 urban rail transit lines, with a total operating length of over 10,000 kilometers and 5,897 stations. Globally, as of the end of 2022, China's urban rail transit leads the world in operating mileage, number of stations, and line count, significantly outpacing other countries.
For this reason, Intel's Network and Edge Group, after deep engagement in the retail, education, and healthcare sectors, has entered the transportation field. Xie Qingshan explained, "The future-oriented intelligent city rail transit system is an important component of smart cities. Intel is deepening the ecosystem for urban rail transit, connecting upstream and downstream ecosystem participants, building integrated smart city rail transit solutions, promoting the deep integration of cloud computing, edge computing, the Internet of Things, artificial intelligence, and other technologies, and fostering more innovative outcomes."
As an essential part of the urban transportation system, urban rail transit plays a significant role in improving traffic efficiency and driving urban economic development. Urban rail stations, as vital transport hubs for passenger entry, exit, and transfer, require the introduction of "automatic ticket gates" to manage and regulate passenger access. With the development of mobile internet and near-field communication technologies, urban rail automatic ticket gates are integrating more digital functions to expedite the ticket checking process for passengers and provide convenience, leading to the emergence of the AFC system. Introduction by Fu Qiang: "Today, AFC systems play a significant role in accelerating traffic operation efficiency and facilitating passenger convenience. The digital transformation of AFC systems relies on the application of innovative technologies such as palm vein recognition."
However, the development of AFC systems also faces various challenges. For example, insufficient performance to support the load of new digital applications like biometric recognition, factors like wide temperature ranges and vibrations affecting equipment stability and availability, and cost pressures impacting rail transit operational revenue.
To address these challenges, Huaming, which has been focusing on the development of automatic ticketing systems and terminal equipment, has introduced an integrated system solution for ticket machine registration, identification, and verification using palm vein recognition technology. This solution, with the Yumai palm vein recognition algorithm as its core, utilizes the Ruibao Zhilian next-generation high-speed and highly reliable edge computing box based on Intel architecture, jointly applied in the Huaming intelligent ticket machine system. This solution enables rapid and precise retrieval of passengers' palm vein features, achieving fast palm recognition for users to pass through gates, effectively alleviating peak-hour crowding issues, and avoiding the instability and fraud issues associated with QR code verification.
Before applying palm vein recognition technology to the transportation sector, Yumai had experimented with it in other industries. But why did they ultimately focus on rail transit? Wang Tianming told reporters that this field best showcases the technology's advantages. Palm vein recognition technology can achieve precise recognition within a range of 1:1,000,000 to 1:10,000,000 or even larger, making it more suitable for large-scale applications compared to fingerprint recognition technology.
What about the next generation?
According to Wang Tianming, the current AFC system incorporating palm vein recognition technology has an error rate of one in ten million and a rejection rate of one in a hundred thousand, with extremely high accuracy. In terms of recognition time, the Chongqing AFC solution is designed to meet demands in the millions, with a single recognition speed of 100 to 200 milliseconds.
Fu Qiang told reporters that the actual single recognition speed of palm vein recognition in the laboratory can already reach 30 to 50 milliseconds, but it has not been achieved in practical experience yet, requiring further improvements in various aspects of the industry chain. For example, improving the speed of palm vein feature recognition and matching features with images in the database.
It is reported that the Yumai palm vein recognition consists of two parts: vein feature extraction, encryption, compression, and high-speed matching of massive features. In this application, vein feature extraction, encryption, and compression run on the Ruibao Zhilian edge computing box based on Intel Atom/Celeron processors. For feature extraction technology, Yumai utilizes a deep learning-based palm vein feature extraction algorithm and employs the OpenVINO toolkit as the inference framework for deployment based on the edge computing box.
Additionally, the Yumai palm vein feature extraction algorithm is optimized using the Intel oneAPI tool suite to enhance performance on Intel architecture servers. Meanwhile, the high-speed matching of massive features runs on edge servers based on the third-generation Intel Xeon Scalable processors to better cope with the performance pressure brought by matching massive features.
Fu Qiang said they are collaborating with Intel to further enhance performance in the next generation of products.
Wang Tianming told reporters that in the process of applying palm vein technology, it is noticeable that the pass rate of male users is significantly higher than that of female users. This is because female users tend to have lighter skin tones, which may lead to less clear vein recognition when the device lacks polarizers and infrared extinction technology, posing a hardware challenge. He revealed that they have developed targeted solutions for this phenomenon and applied the polarizer scheme to Yumai's second-generation palm vein recognition products in the first quarter of this year.
Furthermore, the second-generation products from Yumai incorporate a function to determine the distance between the palm and the sensor, with indications from ambient lights to ensure the palm is in the best position for data collection. For instance, if the palm is out of the recognition range, the indicator light may be red, or when the hand is held particularly low, the indicator color may be blue or orange.
In the past two years, the development of large-scale model technology has driven the iteration of artificial intelligence applications. Xie Qingshan hopes that in the near future, large-scale models can also be applied to the traffic control machines developed by Ruibao Zhilian, allowing the palm vein recognition industry chain to benefit from the "dividends" of large-scale models. "Of course, this requires negotiation and effort from partners."
"Patting hands" at the entrance. Photo by Ji Hongmei.
