In order to solve the problems of fuzzy image, different measuring range and recognition accuracy of pointer instrument in the process of power inspection, propose a pointer-type instrument recognition method based on Yolov3. Firstly, super-resolution technology is introduced to reconstruct the instrument image. Then, a combined Yolov3 and OCR- based dial feature extraction method is used to detect the dial in the instrument image, effectively extract the digital features in the region, and recognize the instrument by combining the digital features pointer-type angle. Next, to solve the instrument image’s fuzzy problem, a super-resolution image reconstruction technology is designed to reconstruct the fuzzy instrument image and get a clear instrument image. Finally, a method based on a Yolov3 and OCR combination is proposed to detect the instrument image and recognize the digital features of the dial to target the insufficient use of dial features. Comparison of the effect of traditional machine learning algorithm and Canny edge detection in the pointer-type instrument image recognition shows that the accuracy and practicability of this method have been significantly improved in the recognition of pointer-type instruments in complex environment images.