Integrated zoom movement, non-human classification detection, making perimeter deployment easier

As the “eyes” of cities, video surveillance is indispensable. Under the fierce construction of safe cities and smart cities, video surveillance is rapidly popularizing and upgrading. Many video surveillance applications urgently require non-cooperative and efficient integration and analysis in complex environments. . The deep intelligent perimeter function is based on deep learning technology and integrates a neural network acceleration engine and an intelligent calculation acceleration engine, which can accurately locate the position of the non-human target in the screen, and the target detection rate is over 90%, which greatly avoids filtering and misdetection. Errors caused by incorrect classification of alarm events, accurate alarms. Compared with the pan-intelligent perimeter, the deep perimeter not only supports cross-border detection, entering area, leaving area, area intrusion and other functions, but also supports motor vehicle/non-motor vehicle/pedestrian classification detection, and can display alarm pictures on the front-end intelligent Web interface , At the same time, the camera will lock the target and follow the target in conjunction with the gimbal. Of course, you can also set the follow time and whether to turn on the automatic tracking mode. It is widely used in rich scenes such as illegal parking areas and perimeter walls. Supports starlight level low illuminance, excellent monitoring effect under weak light conditions. In a starlight environment, it can display clear color images even without any external auxiliary light source. It uses a super-sensitive image sensor and unique electronic multiplication and noise control technology to greatly improve the sensitivity of the camera, and has a 24-hour full Color real-time effects, no tailing phenomenon of ordinary low-light cameras, to meet the needs of high-quality monitoring at night, and low power consumption, real image effects, and no color cast. In a complex environment such as a large flow of people and a mixture of people and vehicles, how to quickly identify and deal with it is a big problem. For this reason, the 3MP starlight-class 18 times intelligent network core of Ushilink supports machine non-human classification detection. For non-human targets on the machine, the lens automatically follows the target to rotate, pay close attention to the target in depth, and avoid monitoring blind spots. The front web screen appears humanoid, car model, motor vehicle, and the camera will link the pan/tilt to lock the target and follow the target. You can set the follow time or auto mode. It can be configured as an intelligent preview mode in the preview interface with NVR, and it can display real-time pictures of faces, inhumans, and alarms on the right. When high-brightness areas under strong light source (daylight, lamps or reflections, etc.) and areas with relatively low brightness such as shadows and backlighting exist in the image at the same time, the image output by ordinary cameras will appear bright areas due to overexposure. White, and dark areas become black due to underexposure, which seriously affects the image quality.