A Method Of Allergy Test Sheet Recognition Based On OCR Technology

As one of the most common diseases in the world,allergic diseases affect people’s life and health.To treat allergies,the allergen must be identified first.At present,most medical institutions detect allergens by prick skin test.The detection method of prick skin test requires injecting multiple allergens outside the body,and medical staff observe the patient’s response to each allergen.,Fill in the electronic allergy test form,and then manually enter the medical system,and finally form the allergy electronic report form.Due to the large number of allergen testing items,the manual entry of medical allergy checklists currently used has the problems of low work efficiency and error-prone methods in the entry process.With the increasing number of medical allergy checklists that need to be processed,there are The need for identification is becoming more and more urgent.In view of the above situation,this paper designs a system for automatic recognition of allergy checklist document images.The main work of the paper is as follows:First,a solution is proposed for image quality problems such as shadow interference and unbalanced illumination that may exist in the uploaded document pictures on the patient’s mobile phone.First of all,for the shadow interference that appears in the document image,this paper introduces the two-dimensional OTSU algorithm to calculate the optimal threshold segmentation point based on the sliding window multi-threshold binarization,and adaptively modifies the current pixel segmentation threshold through the threshold of neighboring pixels,Successfully reduced the shadow interference in the binarization process;for the problem of low document contrast caused by printing equipment,shooting equipment,external light,etc.,based on the adaptive gamma transformation,this paper uses the variance weighted guide filter Retinex The algorithm replaces the traditional Gaussian filter to extract the illuminated image,and then uses the adaptive gamma function to correct the image to achieve document image enhancement.Second,for the recognition of handwritten digits in the allergy checklist,this paper introduces a residual module based on the convolutional network to address the problem of gradient disappearance,and improves the expression of the model by increasing the width of the convolutional layer of the residual block Ability to realize an improved residual network that can be used for handwritten digit recognition in allergy checklists,so that the accuracy of the classification results reaches 98.1%;for the recognition of handwritten allergens in English words in the allergy checklist,the convolutional network of the CRNN neural network Above,the introduction of the residual module uses SELU activation function and batch normalization to accelerate the network convergence and improve the generalization ability of the model.Among them,BiGRU is used in the cyclic network model to replace BiLSTM,which further increases the convergence speed of the network.In the processing part,dictionary lookup and N-gram post-processing strategies are added to further improve the accuracy of model recognition and finally realize the identification of manually filled allergens.Experiments show that the accuracy of this method for handwritten English vocabulary recognition reaches 64.9%.Third,on the basis of the above algorithm,this paper designs a document image recognition system for allergy checklists.This system can automatically preprocess document images affected by shadows and light to improve the quality of the documents;and for allergy checklists The system realizes the automatic layout analysis of the allergy checklist,so that key information in the document can be extracted;based on the handwritten number recognition and English word recognition algorithms implemented in this article,the system can effectively implement the allergy checklist The identification of fixed and non-fixed items improves the efficiency of generating allergy checklists.