As the main business of security, video surveillance has experienced many years of development and changes. Video surveillance technology has provided a lot of help and support for public security departments in personnel search and control, vehicle control and illegal evidence collection. However, at present, video surveillance imaging is still urgently limited by ambient light, which can provide clear images with high color reproduction during the day, but the image quality is greatly reduced in low light and night conditions.
In order to solve this problem, manufacturers in the industry have focused on improving the performance of video surveillance products in low-light situations, and have given different solutions accordingly.
▲ Comparison of mainstream low-light schemes
Super Starlight Technology
It mainly uses a star-level image sensor, with a large aperture lens, and traditional ISP image modulation technology. The super starlight technology mainly relies on the sensitivity improvement of the image sensor to improve the conversion efficiency, because the large aperture lens can increase the amount of light entering. At present, the technology is very mature, and the implementation cost is relatively low.
Therefore, the super starlight camera has great advantages in general video surveillance, but the fly in the ointment is that it is not suitable for special application scenarios. For example: personnel bayonet, vehicle bayonet and electric police cameras.
Due to the lens with a large aperture, although the amount of light entering is enlarged, the brightness of the image is improved. But it also greatly reduces the depth of field of the imaging and expands the effect of the halo of the glare of the headlights.
Secondly, since it is necessary to capture clear moving objects such as motor vehicles, non-motor vehicles and pedestrians at night, it is necessary to increase the electronic shutter speed of the camera according to the speed of the objects, otherwise the pictures captured by the moving objects will be obviously blurred and trailed.
In general, the working shutter of ordinary video surveillance cameras at night is 1/25s, the shutter of the personnel bayonet is 1/100s, and the shutter of the vehicle bayonet is 1/400s faster. The increased shutter speed on these professionally applied cameras offsets the increase in brightness brought by the large aperture lens, resulting in a dark scene around the target object, losing a lot of effective information.
black light camera
The second option is a black light camera. Black light technology mainly uses two star-level image sensors through special optical components, one of which collects image brightness information and object outline through infrared supplementary light, and the other collects color information. Then, the two collected image information is fused through the image fusion algorithm, and the output is both bright and colorful.
▲Schematic diagram of the principle of black light technology
Black light technology does not mean that the camera can image clearly in a completely dark environment. It can also be seen from the technical principle that the first sensor that collects brightness relies on infrared supplementary light, while the other camera that collects color needs ambient light just like ordinary cameras. In fact, in a dark environment, the black light camera Also have to rely on LED fill light to capture the image.
In the application of traffic bayonet, the infrared transmittance of the explosion-proof film pasted on the front windshield of the car is very poor. Therefore, the environmental protection bayonet using black light technology and infrared strobe light, as shown in the figure, cannot collect the image information of the front row passengers.
When encountering some objects with infrared reflective materials, cameras using black light technology also have certain drawbacks in color reproduction. For example, in the following example, because the clothes worn by the driver are made of black anti-infrared light material, the image restored by the black light technology will restore the black clothes to white, which will cause certain troubles in subsequent illegal evidence collection.
Ultra-low light technology
In addition to the above two solutions, the third solution is “super low light”. Ultra-low-light technology is a new image enhancement algorithm based on basic ISP image modulation.
This technology needs to first collect a large number of image samples and simulation data of vehicle bayonet, vehicle electric police, personnel bayonet and fully structured cameras under low light conditions at night, and then carry out targeted mathematical modeling. End-to-end deep learning models for plotting, training, and model transformation.
In the case of low light, the algorithm model skips the ISP imaging modulation method of the traditional camera. Through the learning of a large number of scene snapshot pictures, the algorithm directly restores the image of the sensor input data, which can greatly reduce the camera’s dependence on the fill light. While improving the brightness of the image, it can fully restore the details such as the color and texture of the object.
Relying on the image restored by this algorithm not only greatly improves the human eye’s subjective experience of the captured image, but also improves the feature analysis of the image by many back-end intelligent algorithms. For example, vehicle feature analysis, non-motor vehicle feature analysis, driver and passenger feature analysis, etc.
As shown in the above example, under the low illumination of 1.5Lux without any supplementary light, the visibility of the image processed by the ultra-low-light algorithm has been greatly improved.
In order to obtain high-quality snapshot videos and pictures with good clarity and color fidelity as much as possible in low-light conditions at night, the past practice is usually to add fill lights next to the surveillance cameras. This method makes the monitoring catch up with the installation of too many LED fill lights. The disadvantages of this method are:
The fill light is dazzling;
Affect the line of sight and bring security risks;
Raise the vigilance of potential criminals;
Some color information is lost, and all valid information cannot be obtained;
Now, the emergence of super starlight technology, black light technology, and ultra-low light technology makes it possible to obtain clear and high-reduction pictures and videos at night without active supplementary light, and more comprehensively help video surveillance equipment to operate 24 hours a day .
At the same time, the built-in level of the front-end AI chip in the camera has begun to increase, and the improvement of the front-end computing power will help the further development of the technology, and will also further expand the working environment of the cameras and face recognition devices equipped with these technologies and reduce the lighting conditions. Constraints on performance make its video surveillance, identification and comparison capabilities more comprehensive and excellent.
Responsible editor: gt