Machined Parts Dot Matrix Character Recognition Research

Carving dot matrix characters on the machined parts, as part of the product, arethe product’s status symbol. Therefor,this kind of characters have the characteristicsof long-term storage, non-easy corrosion. However, low efficiency and high errorrate of artificial identification will affect the quality of the product. Therefore,according to the characteristics of machine parts, to develop the machine carvedcharacter recognition system with high recognition rate can greatly improve theproduction efficiency and product quality, also can improve the production level ofautomation.Therefore, the study of this subject has practical significance andpractical value.The character images of this study are obtained by a CCD camera. Because ofthe site environment’s influence,the image will be disturbed.First carries onpre-prosessing to it.To the general degraded image using histogram processing,tomore serious degradation blurred image using partially overlapping sub-blockhistogram equalization. As well as the binarization, morphological erosion, dilation,thinning process.Improved HOUGH algorithm is used to extract characters lattice.Then,to segmentation, the projection method is used,and to the single characterscarries on normalization processing.In recognition phase, the parameters of the neural network is designed.Per-pixel feature extraction method is used to extract the characters’ feature, torecognition the similar characters re-classification method is used.While thecharacters binary code as the network’s input.The results show that the fuzzy image processing algorithm can obtainsatisfactory results,and the proposed method of character recognition not only has afast recognition speed but also has a high recognition rate.