Abstract A digit recognition method based on machine vision is provided for the frequently-used 7-segment LED digital meter. First, images of the meter display value were acquired real-timely by computer-controlled camera. Then the pre-processes of images were done including selection of region of interest and image binarization. Digit images segmentation and coding were further done by projection transformation and then the displayed values of the meter were obtained automatically. Results show that the digit recognition algorithm has great robustness and could meet long period and large amount of data reading and recording.