3D facial recognition and Silicon-to-System technology

The 3D AI facial recognition access control system uses the Kneron KL520 AI chip, which has the characteristics of strong computing, low power consumption, and low cost. Combined with Kneron’s exclusive research and development of reconfigurable AI models and self-developed NPU, it can support a variety of machine learning The framework and network model can be widely applied to various devices such as AIoT. In addition, the use of dual infrared lenses achieves a compact size, flexible use of cameras, and high security and high precision face recognition solutions. More importantly, the cost of this system is only half of the traditional recognition system one. The system’s facial recognition algorithm is also among the best among NIST participants.

Liu Juncheng, founder and CEO of Kneron, said: “Kneron is committed to developing the application of artificial intelligence on terminal devices. The 3D AI solution combines 3D sensing technology and terminal artificial intelligence to greatly improve the accuracy of image recognition and use security. We It is expected that it will bring more new applications to the market. We are also pleased to announce that we will add a new AI SoC product line and launch an AI SoC specially designed for smart home applications. This product is the embodiment of Kneron’s product development strategy TIny AI, providing A lightweight terminal AI solution with small size and low power consumption.”

Featuring dual ARM Cortex M4 CPUs, the KL520 can operate as a host or as an AI co-processor, enabling edge AI on devices such as smart locks, cameras, drones, smart appliances, and robots.

Kineng self-developed algorithm

Support various optical modules


HD Mask Discrimination

twin identification

HD photo detection


Face mask recognition


apply to

Access Control

visitor management

smart door lock

meeting room management

Kneron’s 3D AI solution can support structured light (Structured Light), stereo vision (Stereo Vision), time-of-flight (TIme of Flight, ToF) 3D sensing technology for recognition of faces and body objects, and can be applied In security monitoring, smart home, smart phones, various IoT devices, etc.

Combining 2D image analysis and identification with in-depth information analysis can not only improve the recognition accuracy, eliminate the risk of unlocking with photos and videos, but also more accurately identify objects and behaviors, and provide other 3D image capture applications. It can provide accurate 3D facial recognition unlocking, 3D modeling and other applications, and is suitable for various terminal devices. In the face of increasingly stringent security standards for facial recognition systems, the facial recognition algorithm of this system is not only among the best among NIST participants, but also passed the 3D mask test and passed the anti-counterfeiting certification of BCTC payment level, which can effectively prevent system security vulnerabilities. fraud and identity theft.

Kner Energy Smart Access Control

For the security and modern management needs of residences, offices, and smart communities, the above-mentioned modules can be used to launch 3D face + fingerprint + password + remote snapshot photo door locks. It adopts advanced non-contact 3D face recognition as the key to open the door, and has a variety of door opening modes to choose from, which is more flexible, more convenient and safer!

Steps to use the program:

Enrollment Stage: Enrollment of family members requires multi-pose face collection instead of photo registration like access control. The so-called multi-posture means that during the registration process, it is necessary to design a guiding process, so that people’s faces can be swayed and there are obvious signs indicating that the collection is successful.

Recognition Stage (RecogniTIon Stage): Use or newly configured imaging equipment as an image capture device, and compare it with the database after de-identification.

The results are displayed and related information is bound and output: the comparison results need to bring out a code and time, and it needs to correspond to whether the database of the entrance and exit allows the person to enter and exit, and whether the person is blacklisted and belongs to the control personnel.

Core technical advantages

KL520 product features and technical advantages:

1. Low power consumption, support battery device.

2. Support CNN, TDNN.

3. Support the operation of multiple AI model combinations.

4. Support reconfigurable design, and the model execution efficiency is high.

5. Support multiple KL520 clusters, and the computing power can be superimposed.

Program Features

Program features:

1. Make full use of infrared face information and wider spectral information of visible light for face recognition, face comparison, and liveness detection. 2. At the same time, the depth information of the face is calculated through the feature point parallax obtained by the binocular camera, and finally all the information is matched with the original data through the fusion algorithm. Only after all the information is successfully matched can the authentication pass, and the security is greatly improved. The misrecognition rate is only one in hundreds of thousands.

3. It can adapt well to both indoor and outdoor light environments, and can effectively prevent attacks from photos, videos and 3D simulated face models of various materials.


1. Security

Kneron’s RLC-MS3F face recognition module is equipped with Kneron’s self-developed 3D living body algorithm and face recognition algorithm, which can effectively guarantee the security of user information and unlocking. It has passed the face recognition technology test of the Bank Card Testing Center (BCTC) (Live detection) Enhanced authentication, which can effectively prevent attacks from various types of printed photos, electronic photos or videos, 3D head models, 3D masks and other prosthetics, and its security is better than traditional monocular or binocular face recognition door lock solutions .

2. Ease of use

a) The RLC-MS3F smart door lock can provide users with a non-contact, non-inductive unlocking and opening experience, which saves the cumbersome process of pressing fingerprints or entering passwords, and also solves the problem of

For some people, such as children or the elderly, the fingerprint characteristics are not obvious or the fingerprints are severely worn and difficult to unlock, which also eliminates the potential safety hazard of contact fingerprint residues being copied;

b) Kneron’s 3D living body algorithm and face recognition algorithm have been deeply optimized and repeatedly calculated on the customized AI SoC platform, and the single unlocking time (calculated from waking up) is as fast as 1.0s, providing a smooth door opening experience;

c) With a field of view angle of 71° in the vertical direction, the height range that the door lock can support ranges from 1.3 meters to 2.0 meters, and children and tall people can be covered;

d) The horizontal direction supports a field of view of 58°, and the distance can support face unlocking within the range of 0.3-1.0m. When the user approaches the door, there is no need for the head to face the camera. Facing the door lock, you can realize dynamic snapshot unlocking, even if the door lock is close to the wall, it can be unlocked smoothly.

3. Adapt to complex lighting environment

It can adapt to a wide range of lighting environments such as night, indoor lighting, outdoor sunlight, corridor lighting or sunlight.

4. Ultra-low power consumption

The typical power consumption of a single unlock is 5V / 0.035mAh (unlock time 1.0s), which is equivalent to 0.18mWh; it supports lithium battery and dry battery power supply.

5. Ultra-fast speed

From the trigger to wake up timing, the fastest single unlock is about 1.0s.

6. Stability

Mature and stable hardware solutions and strict product quality inspection.

7. The interface is simple and easy to integrate

The Kneron face module only needs a 4PIN interface to connect with the lock control, and realize power supply and serial port communication at the same time.

Reviewing editor: Peng Jing

Leave a Reply

Your email address will not be published.