Research On Segmentation And Recognition Methods Of Dot Matrix Characters

With the progress of society and the development of science and technology,people are paying more and more attention to the safety of commodities,especially food.In order to improve production efficiency,enterprises usually use inkjet printers to print the production date label on the products,making the production date label as an important evidence for judging the quality and safety,and it must be tested before leaving the factory.However,the dot matrix characters contained in the production date label are affected by production environment and production equipment.The above effects cause character bending,redundant ink spots,uneven illumination and incomplete characters.The existing character recognition system has poor performance on the segmentation and detection of dot matrix characters.Therefore,this paper studies the character segmentation and recognition detection methods to improve the detection accuracy of dot matrix characters.The main research are as follows:(1)Aiming at the interference information of dot matrix characters such as uneven illumination and noise pollution,an adaptive binarization method based on image gray information was proposed.This method obtained the average gray value of the dot matrix character image and applied it to adaptive threshold function for binarization.The experimental results shown that the information of the region of interest of the dot matrix characters had been enhanced,and redundant interference information had been filtered out.Theoretical analysis and experiments shown that the method proposed in this paper can effectively reduce the influence of interference information on dot matrix characters,and had strong adaptability to illumination.(2)Aiming at the problem that the traditional character segmentation method was cannot accurate in segmenting dot matrix characters,a dot matrix character segmentation method based on spray dot fusion features was proposed.Firstly,the jump feature and projection feature of each column of character block were obtained;secondly,the two features were cascaded to get fusion feature;thirdly,the segmentation model based on SVM is established to segment dot matrix characters;finally,the single dot matrix character was sent to ANN for recognition.This method can effectively improved the robustness of the segmentation method by obtaining the two characteristics of the spray dots as the basis for segmentation,and establishing a segmentation model based on SVM can improve the segmentation efficiency of dot matrix characters.Experiments were carried out on the image processed by adaptive binarization,and the results shown that the accuracy of dot matrix character segmentation and recognition reached 99.92% and 98.97% respectively.Theoretical analysis and experiments shown that the method proposed in this paper can effectively segment the characters,and the segmentation result greatly improves the recognition accuracy.(3)Aiming at the defects of dot matrix characters such as incomplete characters and ink dot pollution,a detection method of dot matrix characters based on cosine similarity was proposed.Firstly,the 19 layer network structure with residual block was constructed to prevent the model from fitting;secondly,the center loss function and softmax loss function were calculated respectively in the last two layer full connection layer to enhance the robustness of classification detection method and to distinguish the category of samples;thirdly,calculate the average feature vector of each type of samples;finally,obtained the feature direction of the new input samples The cosine similarity between the two vectors was calculated,and the character defects of dot matrix were judged according to cosine similarity.The results shown that the accuracy of the sample set composed of defect sample and normal sample is 99.867%,and that of the sample set composed of only defect samples is 99.4%.Theoretical analysis and experiment shown that the method proposed in this paper can not only detect the defective dot matrix characters,but also distinguish the types of the defective dot matrix characters,and effectively improve the recognition accuracy of dot matrix characters.