Aiming at recognizing the double-pointer temperature and humidity meter and without angle information, a method based on image processing is proposed．Firstly, binary images of temperature meter and humidity meter are extracted by adopting the color characteristics in HSV color space model; and image shadow is reduced by single scale Retinex. Secondly, scale circle is detected by Hough transform, main scale and pointer images are segmented by K-means clustering. Then, the least squares method is used to fit scale circle by the centroid of main scales, least squares method and Hough transform is chosen to detect the pointer according 3σ rule. Finally, the pointer readings are identified by the angle method. The experiment results show that the proposed recognition scheme is accurate and reliable, has actual application value.