Abstract: In the study of the intelligent ink-jetted code character recognition, it is difficult to locate and recognize the ink-jetted code characters accurately since the ink-jetted code characters are printed on the package of the product with complicated background and other standard characters. During the printing process, the font size of the ink-jetted code characters, the amount of the ink ejection and the influence of illumination are different. In view of the above difficulties, an ink-jetted code character recognition method based on Probabilistic Neural Network (PNN) is proposed in this paper. Firstly, the original image is converted to a grayscale image, and the Gaussian filter is used to remove the noise. Secondly, the improved FAST corner detection algorithm is used to quickly locate the ink-jetted code character. The HOG features and the mesh features of the ink-jetted code characters are extracted, and the two features are merged. Finally, the features are sent to the PNN to establish a classification model, which are used to recognize ink-jetted code characters. The experimental results show that the proposed method is very fast and the accuracy of location is high. Compared with BP neural network, when the characters are subject to different light and have different fonts, the recognition accuracy of the model trained by PNN is improved.

Key words: FAST corner detection algorithm, probabilistic neural network (PNN), ink-jetted code character recognition, character location, feature extraction