With the climax of AI in China, face recognition technology has developed rapidly. Relevant institutions predict that the market size of face recognition will reach 200 billion by 2020, including about 70 billion for customs clearance and security products and 50 billion for online payment.
Face recognition technology is an optical technology for biometric recognition of human face digital information features. Its principle is to use the scanning device to collect the image or video stream containing the face, and extract the features of the face, locate the face, and identify different people through comparison. The operation of the whole process is very huge, and the quality of the collected image has a crucial impact on the recognition efficiency. In this paper, we mainly explain the narrowband filter products used in the face recognition technology system to help the developers and users of the face recognition technology system better understand its role and use methods, so as to correctly select the technical indicators of the narrowband filter.
Determination of narrowband filter bandwidth for face recognition:
The bandwidth of narrow-band filter should be determined according to the environment and the light source used. It is not to say that the narrower the better, the higher the price of narrowing as soon as possible. For example, the bandwidth of 850nm infrared LED is about 50nm. When selecting narrow-band filter, the utilization rate of light energy should be taken into account, so the bandwidth of narrow-band filter cannot be set too narrow. For LED light sources, the bandwidth below 15nm is not appropriate. Because too narrow bandwidth blocks most of the strong signal light of LED, on the other hand, it will reduce the effective angle of use, resulting in the dark edge in the middle of the captured image. After a number of experiments, it is found that when the light intensity of the LED is controlled at about 70%, the contrast of the captured image is very good. In this way, the bandwidth of the narrowband filter can be selected at about 30 nm, while the bandwidth of the narrowband filter can be selected at 20 nm for those with higher anti-interference requirements. The 850nm narrowband filter produced by Suzhou Xixian Optoelectronics Co., Ltd. has a bandwidth of 20 nm and a bandwidth of 30 nm. There are many people who choose the bandwidth of 20 nm.
Determination of central wavelength of narrowband filter for face recognition:
Theoretically, the narrowband filter is matched with the central wavelength of LED. However, the two factors we mentioned earlier are the incident angle effect and the heating problem of LED itself. These two factors urge us to consider the changes in the use of narrowband filters when determining the central wavelength of narrowband filters. Therefore, the central wavelength of the narrowband filter should be predetermined to be about 5 nm longer than the central wavelength of the LED. In this way, the incident angle between 0 and 10 can be considered. At the same time, the case that the central wavelength of LED moves upward due to temperature rise is also considered.
Determination of cut-off range of narrowband filter for face recognition:
The cut-off range of the narrowband filter is determined mainly according to the response range of the receiver itself and the wavelength range of the interference source in the environment where the receiver is located. The general response range is 400-1100nm. In the case of face recognition, the interference sources are mainly scattered or scattered sunlight and artificial light sources around, which span a wide range of wavelengths, from ultraviolet to near-infrared. Combining these two reasons, it can be determined that the cut-off range of the narrowband filter for face recognition is 400-1100 nm.
Determination of cut-off depth of narrowband filter for face recognition:
In theory, the lower the transmittance is within the cut-off range, the better. However, considering the production cost and actual needs, the cut-off depth should be selected within a reasonable range. In face recognition system, when the cut-off transmittance of narrowband filter is less than 1%, the isolation effect of interference light can be clearly reflected. The cut-off transmittance of the narrowband filter for face recognition produced by Suzhou Xixian Optoelectronics is less than 0.5%. For applications with particularly high interference light intensity in the environment, we can also provide narrowband filters with higher cut-off depth to meet the needs of customers.
Determination of peak transmittance of narrowband filter for face recognition:
It is generally believed that the higher the peak transmittance of the narrowband filter, the better, but this is not always the case in face recognition applications. When the face recognition device is exposed to direct sunlight, the intensity of the interference light is large, and the interference light with the same wavelength as the signal light is also strong. This interference light cannot be eliminated by narrowband filter. In order to improve the anti-interference ability of LED light, it is necessary to further improve the incident intensity of LED light, so that the signal intensity is several times stronger than the interference intensity. It is relatively simple to increase the intensity of LED light source. As long as the number of LEDs is increased, when the energy of LED light source is strong enough to a certain value, plus the interference light energy with the same wavelength as LED, the response of CCD receiver is easily saturated. Even if the exposure is reduced by software, the image distortion will be serious. At this time, the narrowband filter is required to filter the interference light in the cut-off area. According to the actual situation, the peak transmission rate of the narrowband filter is required to be 40% or 60% or other values.
The face recognition filter produced by Suzhou Xixian Optoelectronics has successfully achieved the narrow-band filter on the glass substrate of more than 0.5mm by using the all-medium hard film coating technology, which can be built into the camera. Because the absorption color glass is not used as the substrate, the filter has good stability, is not easy to mildew and fog, and is suitable for working under strong light (too sunlight) interference; The transmittance is up to 92%, and the light attenuation coefficient of the instrument is small, which can effectively improve the working distance and light intensity, and make up for the instability of LED infrared light; The actual cut-off rate is high, and there is no penetration under the strong infrared light when it is vertically incident, and the cut-off value can reach above OD5.