As one of the important core equipment of video surveillance, the camera has made great progress from the first generation of cameras to the present. From black and white to color, from ordinary cameras to current all-in-one and smart cameras, technical indicators such as wide dynamics, low illumination, resolution, and signal-to-noise ratio have improved rapidly, but the basic principles have not changed. At present, the mainstream technologies in cameras involve wide dynamic, intelligent analysis, high-definition, and low illumination.

wide dynamic
Digital wide dynamic does not achieve the real purpose of expanding the imaging dynamic range, but improves the contrast of the local area through the image post-processing algorithm of the software, which is generally realized by the camera ISP module. The range of gray levels that can be discerned by the naked eye is very limited, but in fact the computer can distinguish very faint gray-scale differences. Digital Wide Dynamic uses image processing algorithms to enhance these faint differences to the naked eye. Later, on the basis of CCD hardware technology, dual-frame synthesis wide dynamic appeared. The solution was to use a CCD, but each point above was exposed twice in a single time, one long exposure (low shutter) and one short exposure (high shutter). 

Therefore, each point has two data outputs, which is called "dual output CCD". Using DSP's unique image processing algorithm, the parts of the two images with appropriate brightness are respectively cut off, and finally superimposed and synthesized and output a light and dark area. Clearly visible image. Whether it is digital wide dynamic or dual shutter wide dynamic, the wide dynamic effect is not ideal.

With the evolution of DSP and CMOS technology, DPS uses individual exposure and control technology for each pixel, plus the use of multi-frame images collected by CMOS sensors to synthesize a linear superposition of a complete image. Compared with the double exposure imaging of CCD A higher dynamic range. From a numerical point of view, the CMOS camera using DPS technology can reach 120dB or even 140dB with the current processing technology. Wide dynamic technology has become an important indicator to measure the performance of a camera. From the current point of view, the standard wide dynamic function has become the consensus of all IPC manufacturers.

Insight
The camera has developed from two or three intelligent analysis functions such as motion detection and video occlusion introduced in 2011 to today. Almost all mainstream manufacturers’ cameras have more than 10 standard intelligent functions. Of course, these intelligent functions are now standard. Most are only limited to mid-to-high-end industry products.

1. Diagnostic intelligent analysis. The diagnostic intelligent analysis of the camera is mainly for the black screen, blur,
Pan out of control, picture freezing, and other common camera failures, video signal interference such as scene changes, leftover/disappearing items, etc., perform accurate analysis, judgment and alarm.

2. Intelligent analysis of recognition. This camera technology is biased towards the analysis and processing of static scenes. Through core technologies such as image recognition, image comparison and pattern matching, the extraction and analysis of relevant feature information such as people, vehicles, and objects are realized. The main application of vehicle identification and analysis is license plate recognition technology.

3. Behavioral intelligent analysis. The camera technology focuses on the analysis and processing of dynamic scenes. Typical functions include: vehicle retrograde, defense zone intrusion detection, personnel focus detection, trip wire crossing detection, rapid movement, personnel wandering detection and passenger flow statistics. With the increase of computing power, the front-end camera can analyze the video content in real time, detect moving objects, identify the attribute information of people and cars through the built-in artificial intelligence chip, and transmit it to the back-end artificial intelligence central database for storage through the network. At present, facial identity verification, vehicle recognition applications, video structuring, and behavior analysis are more common.

high pixel
Only 720P high-definition network cameras were launched in the first year of high-definition in 2010. Until 2012, mainstream security enterprise high-definition network cameras were still dominated by 1.3 million and 2 million, and there were few pixels above 300. With the introduction of CMOS technology, high-definition cameras have developed rapidly. Cameras with 3 million, 4 million, 5 million, 6 million, and 12 million pixels have sprung up like bamboo shoots after a rain. This is the innovation of technology that brings innovation in the product system. At present, Ultra HD, represented by 4K and 8K, has become the standard configuration of various manufacturers. For users, 4K is not just a visual experience and enjoyment, but the resolution of 4K is 4 times that of 1080P. If you use a 4K camera and a 1080P camera to shoot the same scene at the same angle of view, the 4K camera will use 4 It is twice the amount of information used by a 1080P camera to restore the scene, and the picture is naturally clearer and closer to reality. From the perspective of "use", since the information volume of 4K images is four times that of 1080P, based on more information, more accurate intelligent analysis can be achieved. Once 4K is deployed on a large scale, the accuracy of intelligent analysis can be Go up a step, and there will be richer and more surprising smart applications to be realized.

low illumination
It is recognized in the industry as 0.001Lux and below as starlight camera. It is currently widely used in safe cities, finance, hotel buildings, safe villages, ports, highways and other projects. It can be obtained without large-scale installation of supplementary lighting facilities. Better night high-definition color surveillance picture requirements. Starlight-level illuminance monitoring technology is mainly affected by factors such as lens, image sensor, back-end image processing technology and other factors. All security companies also use large-aperture lenses, select large-target sensors, and good image processing technologies. Wait for several aspects to improve. Currently, AI ultra-low-light technology is represented. Through the refinement of the target image requirements of the application scene, a large number of image samples of vehicle bayonet, vehicle electric police, personnel bayonet, and fully structured cameras under night low-light conditions have been collected. Simulated data, and carried out targeted mathematical modeling, and designed a set of end-to-end deep learning models from collection, plotting, training and model transformation. There are also super starlight, black light, aurora and other technologies.