Research And Implementation Of Peach Grade Classification Algorithm Based On Ensemble Learnin

The grading of peaches is a fundamental part of peach processing.The grading methods are of two types: manual grading and machine grading.Both have their disadvantages.Manual grading is inefficient;machine grading is likely to cause secondary damage,and machine grading often uses a single feature and a single classifier for grading,with low grading accuracy.The development of image recognition technology is becoming increasingly mature,and the grade classification method based on the appearance of the peach can not only improve production efficiency but also avoid secondary damage to the peach so that the grading quality of the peach can be guaranteed.The grade classification algorithm based on integrated learning proposed in this paper can improve the recognition accuracy of peaches,which is of great significance for the grading of peaches.The primary work of this paper is as follows:Firstly,the datasets for the appearance of peaches were built.Relatively high-end honey peaches were selected as the research object.Peaches of different sizes were purchased in batches,and photos of peaches were taken and preprocessed by filtering techniques to remove gaussian noise and salt and pepper noise.The effects of image segmentation of OTSU threshold segmentation,TRIANGLE threshold segmentation and color segmentation were compared.Secondly,the four characteristics of peach maturity,roundness,diameter and texture were extracted,and the accuracy of different algorithms for peach diameter extraction was compared.When extracting defects,the peach image was first pretreated,and then Local Binary Pattern(LBP)operator was introduced.LBP operator and gray level Co-occurrence Matrices(GLCM)were combined to extract the four parameters of peach entropy,energy,contrast ratio and inverses of square matrices,and a support vector machine(SVM)was used to distinguish normal fruit from defective fruit.The experimental results showed that the accuracy of peach defect grading with the LBP operator was 90%,higher than that without the LBP operator.Finally,by analyzing the existing single classifier algorithm and introducing a voting polling mechanism to integrate the single classifier algorithm.This algorithm can effectively improve classification accuracy compared to the traditional single classifier.It first used a feature extraction algorithm for peach feature information,combined these features with SVM,decision tree and k-nearest neighbors(KNN)algorithm,and used a soft voting method for algorithm integration.The classification accuracy of95% was achieved through algorithm integration and improved compared with a single classifier.At the same time,a peach classification system was designed and implemented based on the above work,and the effect was also shown.