In this paper, an image recognition technique for dot-matrix character on piston surface based on Hopfield neural network method is proposed. The overall structure presentation of the measuring device, the difficulties of recognition technique about the sparse dot-matrix character and the spray-ink character, the actualization of the training and the classification with Hopfield neural network method, the operation control process of automatic detection etc., are all discussed. Meanwhile, the experiment results in production line are presented as well.