SF6 Pointer Pressure Meter Reading Method Based on Fusion Attention Feature UNet

Aiming at the problem of decreasing the accuracy of meter reading caused by manual reading, an SF6 pointer-type pressure meter reading method based on the fusion attention feature UNet is proposed to improve the accuracy of pointer-type pressure meter reading. Firstly, design the fusion attention feature UNet neural network to segment the pointer and dense scale of the SF6 pressure meter. The Ghost convolutional module used in the coding layer of the neural network can reasonably utilize redundant features to strengthen the inference ability of the network. Deep semantic information feature fusion module built to extract deep potential feature information. Also, introduce the pyramid split attention mechanism to strengthen the information interaction between the network coding and decoding layers. Then use the minimum outer rectangle algorithm and K-means clustering algorithm to determine the circle’s center of the SF6 pressure meter for the segmented data. Finally, use the circle’s center to fit the initial scale and the pointer in two straight lines. It calculates the angle between the two straight lines. The angle conversion formula obtains the accurate reading of the SF6 pressure meter. It is proved by experiment that the proposed intelligent reading algorithm will not be affected by environmental factors and can better divide the pointer and dense scale in the SF6 pointer pressure meter to complete the accurate reading of the meter.