Structural health displacement monitoring equipment based on machine vision in the market can be divided into manual fixed focal length and automatic zoom according to the different lens and algorithm design of the equipment. Fixed focal length equipment, because the focal length is adjusted in the field, it cannot be adjusted after being fixed with screws, so the field of view and depth of field are also fixed and limited. In order to accurately identify the target, the machine vision algorithm needs the minimum pixel area of the target in the image, and the target that is too small will not be identified. For example, if the equipment of some companies is to be able to recognize the minimum displacement change of 1mm, then the target image needs 1 pixel to be recognized, then the target needs 100 pixel. If an industrial camera with 8 million pixels is used, the horizontal and vertical pixels are 3840*2 160. Then the corresponding field of view in reality is 3.84 meters wide and 2. 16 meters vertical. If the diameter of the target is 300mm, the number of targets that can be accommodated theoretically is greater than 10. Considering that many target measuring points are usually not arranged in the range of about 3 meters, the actual number of measuring points that can be accommodated on the same section is usually 1. This can be seen from the actual installation drawing provided by the manufacturer. In order to increase the number of measuring points, some manufacturers set the measuring points on different sections. However, the depth of field of a camera with fixed focal length is extremely limited. If the section distance is too long, the image of the target will become blurred and difficult to identify. Other manufacturers use different cross-sectional distances and targets of different sizes to increase the number of measuring points. That is, as the distance increases, the size of the target increases accordingly. Some projects even have a target diameter as high as 1 m. And the targets with different sections are almost on the same straight line parallel to the optical axis of the camera. Because the target needs to be customized according to the actual situation on site, it is difficult to design a target that meets the requirements of on-site use because no manufacturer sends people to visit the site. The scope of use is greatly limited. Even if the alternative scheme is adopted, the number of targets that can be measured simultaneously on different sections is still extremely limited, and it is usually difficult to exceed 10. This is also one of the important reasons why machine vision structural displacement monitoring products can be practical.
With the development of science and technology, high-performance embedded hardware that can capture high-definition images in real time and complete real-time data calculation appears constantly. Artificial intelligence technology has also developed rapidly, such as image super-resolution based on deep learning and more efficient target detection and recognition algorithm, which can run in embedded NPU. At the same time, the GPU hardware with high computing power trained by neural network model is easier to obtain, which provides a material and technical basis for a new generation of structural health displacement monitoring products based on machine vision. There are very few manufacturers who follow the trend of the times, take different paths of the technology tree and choose a more difficult technology, but they can fundamentally solve the technical problems in the current engineering field application. For example, Shenzhen Anrui launched the fifth generation of self-zooming visual displacement monitor. Through highly integrated devices, the equipment can automatically focus on and identify the targets of all measuring points in scenes such as bridges, tunnels, dams, railways, mines and slopes. The whole process is intelligent. It can identify and calculate the two-dimensional displacement of the target in a large space of 0. 1 to 400 meters wide 12 meters high and 8 meters high, with an amazing measurement accuracy of 0.02 mm.