Nowadays, with the digital transformation of the power grid, the method of manually inspecting and recording the status of equipment in power substations is gradually transitioning to intelligent and automated monitoring. Firstly, an improved YOLOv4 algorithm is utilized to extract the dial region of the meters. Subsequently, an enhanced DeepLabv3+ algorithm is used to segment the dial region. Based on the segmented results, the pointer and edges of the dial are effectively extracted, and the angle between the pointer and the edge is calculated. Finally, the meter reading is determined based on this angle. Experimental results demonstrate that the proposed algorithm can accurately recognize meter readings, with the majority of errors falling within a 1.5% margin. Comparative experiments have also validated the effectiveness of each module in the proposed algorithm. Finally, the algorithm is deployed on the Atlas development board at the edge of the substation for the purpose of meter reading recognition.