In traffic scenes, in order to obtain clear images under low illumination at night, two methods are often used. One is high-brightness fill light, but this also causes serious light pollution. Citizens are dissatisfied and complained from time to time. Bright and dazzling flashing lights It is easy to cause driving safety accidents; the second is the use of infrared strobe lights. Although there is no light pollution, the color cast of the images captured by this scheme is serious. It can be said that the nighttime imaging quality of cameras has become one of the shortcomings of security development.
In order to solve the problems of light pollution and poor image quality, Huawei launched the self-developed AI ultra-low-light camera X2391-20-T. Using DNN ISP image optimization technology, even in low-light conditions, it can still clearly capture the license plate number, Body color and images of occupants provide high-quality forensic records for traffic management.
Huawei’s AI ultra-low-light camera follows the technical route of “computing power for images”. It has two built-in AI chips with super large computing power. Whether it is in the case of low illumination at night or in scenes with strong reflection of glass during the day, it can obtain full Colorful, clean and noise-free images. With LED intelligent fill light, the directional precision fill light technology is adopted, which effectively cuts off long-distance light, the light is soft and not dazzling, reduces the impact on drivers and passengers, and eliminates light pollution. At the same time, it supports the three-in-one of electronic police, vehicle bayonet, and flow monitoring.
The algorithm mall combined with super computing power is the cornerstone of efficient and intelligent
The Huawei AI ultra-low-light camera has two built-in AI chips with super computing power.
The camera supports the open-architecture SDC OS, and can quickly load third-party algorithms through the Huawei Algorithm Store (HoloSens Store). Relying on the algorithm store system and super computing power, a variety of intelligent algorithms can be quickly launched to meet the needs of multi-scenario intelligence. The software-defined concept proposed by Huawei reshapes the camera architecture and front-end intelligent system. Based on the standardization of the existing hardware platform, the camera has continuous vitality and brings a new intelligent experience to customers. The Huawei AI ultra-low-light camera tested this time fully utilizes its dual advantages in computing power chips and software definition to achieve the ultimate application of computing power and algorithms.
AI image enhancement, clear all-weather images inside and outside the car
Huawei AI ultra-low-light camera, using DNN ISP technology, can provide full-color images even in the dark night without color cast by noise reduction in the temporal and spatial domain of the image.
In a strong reflective environment, it can resist color pattern interference, and the capture rate of in-vehicle images is increased by 50%.
The measured results at night and during the day are as follows. In the case of glass rainbow film interference during the day, the optimization effect is obvious.
Intelligent fill light, effectively reduce light pollution
Huawei AI ultra-low-light camera, with LED intelligent fill light, 30Lux for soft strobe, 150Lux for non-perceptual strobe, effectively cut off long-distance beams of more than 50 meters, with little impact on drivers.
During the actual measurement, in the urban night traffic environment, the matching LED intelligent fill light is soft and not dazzling, which is in sharp contrast with the dazzling light of street lights. The aircraft’s fill light has little impact on the driver and passengers, but with the blessing of high-definition image recognition technology, the image characteristics of vehicles at night are still clear.
Adaptive tuning of image parameters, automatic marking of lane lines
Huawei AI ultra-low-light camera, the minimum color illumination is 0.005Lux, and the black and white is 0.0025Lux. It supports strong light suppression, digital noise reduction, and super wide dynamic, which can greatly improve the quality of vehicle monitoring images at night. It also supports parameter adaptation and automatic marking of lane lines, which provides great convenience for engineering installation.
In the actual measurement, the fill light is turned on at night, and the light is soft and not dazzling. Even in the dark night, the license plate number, model, body color and occupants can be clearly identified. The fill light function is uniform and can effectively Identifying illegal behaviors such as the main and co-pilots not wearing seat belts and the main driver making phone calls while driving, the portraits of the drivers and passengers in the car and whether there are illegal operations are clearly visible, providing strong evidence for traffic police management.
Waterproof and dustproof to adapt to outdoor complex environment
Huawei AI ultra-low-light camera, the main body of which is made of metal, the protection level reaches IP66, anti-riot IK10, is not afraid of wind, frost, rain and snow, and can adapt to complex outdoor environments.
1. Using a 9MP image sensor with a resolution of 4096 (H) × 2160 (V).
2. The size is 515.9mm×215mm×160mm, the weight of the whole machine is 6.36kg, the maximum power consumption of the power supply is 57.7W, the working temperature is -40℃~60℃, the working humidity is 5%~95% (non-condensing), 4kV Lightning protection and surge protection. Complete interface, there are 2 RJ45 10M/100M/1000M adaptive Ethernet ports, 1 RS485 interface, 1 alarm input/5 output interfaces, audio input/output interfaces, etc., built-in MicroSD card slot, support MicroSDHC/MicroSDXC .
feature of product:
1. Built-in AI chip with super computing power independently developed by Huawei, with a total computing power of 20TOPS.
2. Support DNN ISP image optimization technology, the image at night is clear without color cast, and it is resistant to glass color pattern interference in the strong reflective environment during the day, and the image in the car can be seen clearly.
3. Support intelligent fill light, cut off stray beams, and effectively reduce light pollution.
4. Three-in-one electronic police, vehicle bayonet, and flow detection, one machine is multi-purpose.
It can be said that this camera has super computing power and super intelligence. Under the blessing of DNN ISP image optimization technology, it adopts intelligent supplementary light, which not only solves the problem of light pollution, but also ensures that the low-light image at night is clear without color cast, and there is no glass color during the day. It is an excellent solution for the smart traffic control bayonet system.
Responsible editor: gt