the camera module for face recognition
Face has certain immutability and uniqueness, which is the most common way for human beings to confirm their identity. The “face brushing payment” for shopping in supermarkets and the “face brushing attendance machine” for commuting to and from work every day have been widely used. There are more applications that are gradually connected to face recognition technology. However, apart from relying on big data computing, what is the relationship between face recognition and cameras? Today, Xiao Jin will take you to understand.
The product interaction logic of general face recognition needs to go through four steps.
● Step 1: face detection
The camera detects whether there is a face in the current image. If a facial portrait is recognized, it will analyze the portrait attribute of the face. (facial feature point location, facial occlusion judgment, facial angle analysis, etc.)
● Step 2: Live face detection
Live detection is mainly used to check whether the facial recognition of the camera is attacked by a prosthesis. For example, put a face image, video facial dynamics, or face masks. This link is an alternative according to different product logic. In the face payment scenario, in order to reduce the payment risk, or fake clocking and face brushing, live detection is a necessary product link.
● Step 3: facial feature extraction
In this link, the abstract features of the target face are extracted through neural networks, and a set of X dimension feature values are output to describe the abstract features of the face.
● Step 4: Face feature search and recognition
Through the quick matching and comparison between the current face recognition influence and the facial data features stored in the face recognition server, the identity judgment can be recognized.
Through the above interaction logic of face recognition products, let’s look again at the relationship between face recognition and cameras.
It is not difficult to see that in the first and second links of the application scenario, face recognition has certain requirements for the camera hardware module. Xiaojin gives you some professional references:
01. Components and parameters
The working principle of the camera is to convert the light signal into an image signal, so the complex light environment such as strong light, backlight and dark light will directly affect the image quality of the camera. The quality of camera imaging will also directly affect the use effect and experience of face recognition and live detection. Therefore, when selecting products, developers should pay attention to the camera imaging core components and key parameters to ensure that the image quality collected by the front end meets the face recognition requirements.
02. Safety and Cost
The monocular camera with appropriate live detection algorithm can achieve monocular live detection. For example, Jinshikang has introduced ArcFace face recognition algorithm, which is free and open on the Hongruan vision open platform, to achieve silent live recognition without user action cooperation. In addition, ArcFace algorithm can also be used to quickly realize face recognition, age detection, gender detection, face recognition under large area occlusion and other full functions.
The binocular camera adds an IR infrared lens on the basis of RGB camera, which has the video capture function of near-infrared light and visible light. Based on the infrared imaging principle, the screen type cannot be imaged, so it is inherently capable of resisting the false face attack of screen imaging.
Therefore, compared with monocular cameras, binocular cameras are better at defending against fake face attacks. For example, the 5 million binocular wide dynamic face recognition camera sold by Jinshikang uses the IR binocular living technology of Hongruan free ArcFace algorithm. Based on the infrared imaging principle and the high robustness of depth learning algorithm, it has excellent defense against paper photos and screen photos.
However, from the perspective of cost, binocular cameras are larger than monocular cameras. Therefore, in the selection process, specific considerations should be made according to the actual application scenario, the living effect to be achieved, and the final cost to balance the needs of all parties.
03. Rigorous test
To ensure the smooth use of the product after landing, testing is an essential link. As far as the camera module is concerned, its basic performance tests include analytical force test, screen test, line test, color cast test in normal shooting test environment and dark angle test, dirt test and bad point test in whiteboard test environment. In all these links, testing is not only to raise bugs, but also an important step for development and product design improvement. Therefore, testing is often highly professional and rigorous to product effect.
Summary: In the face recognition application scenario, in addition to relying on the big data computing center, the camera module is also a very important product hardware. Its parameters, costs, testing and other aspects need to be carefully compared.
Through the cooperation with Hongruan, Jinshikang has made a variety of camera modules with different demand schemes and different costs simultaneously, mainly to cope with different choices of different customers, facilitate customers’ simple and low-cost use, and also reduce the barriers to entry for more customers who want to be in the field of artificial intelligence. Shenzhen Jinshikang Technology Co., Ltd. welcomes you to visit and investigate cooperation.
camera module for face recognition camera module for face recognition camera module for face recognition camera module for face recognition camera module for face recognition camera module for face recognition