Research On Segmentation And Recognition Of Drug Images In Smart Pharmacy

With the continuous advancement of computer technology,the continuous deepening of the integration of artificial intelligence with industry and manufacturing,and the continuous improvement of people’s requirements for medical service experience,as of now,more and more countries have focused their attention on further improving the automation of pharmacies Questions come up.However,most of the completed smart pharmacies have the problems of low automation level,poor real-time performance,and inseparable cooperation of pharmacists.The degree of "wisdom" is not high enough,which not only greatly wastes medical resources,but also exists in the process of manual delivery of drugs Issues such as attrition and security risks.In order to improve the automation of smart pharmacy equipment,this article focuses on the research of drug image segmentation and recognition algorithms,and is committed to adding visual assistance functions based on image processing technology to pharmacy equipment.In this paper,based on the field environment of the Fifth Affiliated Hospital of Sun Yat-sen University,the image sample collection of inpatient pharmacy is completed.In order to solve the problem of poor quality of the original data and subsequent poor image segmentation and recognition performance,image preprocessing techniques are used to improve the sample image quality.To this end,through the experimental testing of a variety of image preprocessing technologies,and comparative analysis of relevant experimental results,two optimal preprocessing algorithms were selected to complete the construction of the drug image database.In order to improve the accuracy and real-time performance of drug image segmentation,this paper improves on the basis of the traditional region growth algorithm.In the case where the existing segmentation algorithm is difficult to segment the capsule medicine accurately,this paper uses the characteristics of the upper and lower parts of the same capsule to be different from other connected domains in distance and angle by calculating the geometric moment of the capsule medicine to achieve the algorithm merge Scheme design.By comparing the processing results of this algorithm with other segmentation algorithms,the improved algorithm proposed in this paper has greatly improved the segmentation accuracy and real-time performance compared with other segmentation algorithms.In order to improve the accuracy of drug image recognition,the article is based on the drug image database constructed in Chapter 2,extracting some drug image samples for feature extraction of color,shape,and texture,and selecting the strongest expressing ability after analyzing the results after normalizing the features.One group performs classifier identification.However,the recognition accuracy of a single basic classifier is low.For this situation,this article optimizes the classifier based on integrated learning,and extracts sample images of drugs for testing and verification.Through experimental comparison with each classifier,it is found that the recognition accuracy of the integrated classifier Optimal,and finally,the computer verification function was implemented by the host computer in combination with the operation of the equipment.