Cylinder-Trachea Matching Detection Algorithm Based on YOLOv7 and Image Processing

There are a lot of hot work on the construction site, and ensuring the correct matching of gas cylinders and air pipes is an important guarantee for the safety of their operations. At this stage, in addition to manual inspection, there is no perfect intelligent detection method that can accurately detect and judge this problem. In response to the above problems, an intelligent detection algorithm for gas cylinder trachea matching based on deep learning combined with image processing technology is proposed. We embed the SA attention mechanism module in the head network extraction part of the YOLOv7 network, enhance the target weight file in the dual channel, and improve the network's recognition and positioning accuracy of the gas cylinder and pressure gauge area; we also utilize the connections between the gas cylinder, pressure gauge and trachea Features, assist in locating the trachea area, ensuring the positioning accuracy of the trachea corresponding to the cylinder; this paper uses HSV color conversion to complete spatial mapping of the target area, and obtains trachea color information through a dual feature target extraction algorithm to complete matching detection. After experimental testing, the average detection time of this algorithm is 65.30 ms, and the matching detection accuracy is 87.51%. It can meet the matching detection requirements of gas cylinders and air pipes in hot operations in actual construction sites, and provides a reference for practical applications.