printer friendly version

This is the future. This is what we do.

Visible light facial recognition with deep learning from ZKTeco.

ZKTeco has created a unique range of visible light facial recognition products combined with a flexible Android platform. The unique range of products is based on the latest artificial intelligence (AI) and neural network technologies. ZKTeco’s visible light facial recognition technology has been specifically developed for purposes where high-speed facial recognition at an extremely high level of accuracy is vital.

Deep learning

As artificial neural networks become more efficient, graphics processing units (GPUs) get more powerful and as training data have seen explosive growth, deep learning applications have increased dramatically. From object recognition to feature extraction, to data management and business analysis, etc., deep learning brings a new level of video surveillance analytics to the security industry.

The ZKTeco facial recognition system is based on deep learning techniques, and adopts sophisticated algorithms to automatically identify or verify the identity of objects in images or videos. Its non-intrusive nature provides a wide range of potential applications. And with recent advancements in deep learning, accuracy has improved to the point that nowadays face recognition is often preferred over other traditional biometric modalities such as fingerprint or iris recognition.

Unlike traditional facial recognition devices in the market today, where face alignment poses a problem, the ZKTeco visible light facial recognition products use deep learning algorithms to provide true cognition of faces, which enables larger pose angles. By applying 3D rendering through machine learning technology, its products provide pose angle tolerance up to 30 degrees in all yaw, pitch and roll axes.

ZKTeco products make use of a proactive facial recognition technology which enables users to be automatically recognised by the device while walking past, unlike having to stand steadily in front of a camera. Thus, it is way faster than the previous near-IR facial recognition and the recognition distance has been greatly extended to up to three metres, which significantly improves maximum traffic rate. Detection of up to five people at a time can be recognised per second if the facial template is stored locally on the device.

Anti-spoof functionality

With the applied convolutional neural network (CNN) algorithm, anti-spoof functionality has been greatly enhanced within visible light facial recognition products and is able to for check fake faces, photos and video attacks. Face feature extraction is based on detecting differences in lighting, texture and resolution, and includes two types of tests. Anti-fake photo attack is tested by comparing the near detected face’s background pixel movement and consistency of the detected face’s resolution verses a video stream background. With the anti-fake video attack test, the device can detect a face’s texture and resolution, as well as whether the detected face is surrounded by a frame or not.

The visible light facial recognition technology is a touchless biometric solution, which is much more hygienic than other biometric technologies. The products use an Android operating system that brings clients the unlimited potential to formulate their own solutions, based on the needs of their end users. Designed in various structures and sizes, the facial recognition technology is very flexible and can be customised to fit any solution.

ZKTeco’s visible light facial recognition technology can either be utilised as an access control terminal, time and attendance terminal or any other solution can be derived from this platform. For example, the device can be utilised as an advertisement display panel with biometric analysis on age, gender, even behaviour, making it easier for clients to identify their target market. With this technology the ‘Know your Client’ concept can be perfected – staff can immediately recognise their regular customers and easily assist them.

The visible light facial recognition system is suitable for various applications including offices, warehouses, clinics, manufacturing plants, catering premises and constructions sites, etc.

Share this article:

Further reading: