Abstract: A computerized method of extracting and recognizing characters on beer bottle is proposed in the paper. It replaces handwork and reduces production costs and improves efficiency. The image of beer bottle characters is firstly obtained with a linear area CCD camera, and the beer bottle image is preprocessed. Then the raised character is extracted using an improved active contour model. Finally, the characters are recognized, by using multi-layer artificial neural network. Experimental results indicate that the method can extract beer bottle characters and recognize "B" and the date efficiently. The marking accuracy rate is over 95 percent.