A system based on computer vision is designed to realize the automation of intelligent electric meter quality test. It can effectively detect the defect of Chinese characters, nixie tube, alphabet and symbols on the LCD panel ,caused in automatic welding; Main algorithm of the system includes three parts: orientation of the interested part, characters segmentation and defect judgment; Back projection of color image histogram is used to locate general LCD area. Canny edge detection algorithm and line detection are applied to obtain the precise coordinate of featured area; Horizontal projection and vertical projection are respectively used to split the symbols in the featured area; At last, template matching is used to judge the existence of the defect. The system is implemented to test the meter DL/T645-2007, and good precision and high efficiency are acquired.