Digital Instrument Reading Detection and Recognition Based on Improved YOLO11

Aiming at the issues of traditional digital instrument recognition methods, such as difficulties in accurately segmenting individual characters, completely recognizing valid characters, and the diversity of digital instruments, this paper improves the character detection and recognition network based on YOLO11. To reduce computational load, the depthwise separable convolution module GSConv has been introduced into the backbone network of YOLO11 to decrease the number of parameters. Experimental results demonstrate that the proposed method can adapt to various digital instruments, achieve accurate recognition of readings and equipment operating states, and contribute to promoting the automation of operation inspection and calibration for digital instruments.