In order to solve the problems of uneven illumination and low accuracy in the automatic recognition system of pointer instruments, an Otsu algorithm and an improved Hough algorithm are proposed to process the instrument image. Firstly, the instrument image is preprocessed to enhance the characteristics of the instrument pointer area, Otsu algorithm performs the image binarization to segment pixels that only belong to the pointer, and the improved Hough transform is performed to detect the pointer. By quickly extracting the center pixel point of the dial pointer's connected area, combined with the dial center constraint, the double threshold Hough is performed on the extracted pixel point transform straight line detection. Experimental results show that the proposed algorithm detects the meter reading error at about 4.76%, which meets the system accuracy requirements.