Study On Detection Method Of Germinated Potato Based On SVM And Weighted Euclidean Distance

The storage space of potatoes needs to have proper temperature and humidity conditions,otherwise it will cause potato germination.For germinated potatoes,a lot of solanine is produced,and solanine is a toxic substance that will cause certain harm to the human’s sports system and respiratory system.Therefore,before the potato enters the market,it must be tested to remove germinated individuals to ensure the quality of agricultural products and human health and safety.Aiming at the shortcomings of germinated potato detection technology,this paper proposes a germinated potato detection method based on SVM(Support Vector Machine)and weighted Euclidean distance to improve the detection accuracy of germinated potato.First,the original image of the potato sample was obtained based on an industrial camera,and grayscale and median filtering were used to ensure the image quality.Then,the B component and the H component were extracted in the RGB color space and the HSV color space to train the SVM classifier.The SVM classifier is used to segment potato images from the background.Finally,weighted Euclidean distance and morphological methods are used to detect and label potato germination sites.Based on Matlab R2014 a software platform,weighted Euclidean distance,traditional Euclidean distance method and manual detection method are used to detect potato samples.The test results show that the average recognition rate of the weighted Euclidean distance method is 94.6%,and the average recognition rate of the traditional Euclidean distance is 89.2%.The recognition rate of the artificial detection method is 88.3%.Compared with the traditional European distance detection method and the artificial detection method,it shows that the method has a higher recognition rate and a better detection effect for germinated potatoes.The main research work of this paper is as follows:(1)Review the research status and application prospects of germinated potato and the research status of detection models mainly by consulting a large amount of literature;(2)Based on the research content of this article,the technical roadmap,image acquisition system design,image acquisition equipment experimental materials,and software platform and equipment construction were developed in detail.(3)An introduction to the key technologies used in the article,including image enhancement algorithms,SVM image segmentation,weighted Euclidean distance detection methods,and related technologies used in experiments.(4)SVM segmentation principleSVM image segmentation is realized by using SVM model to classify each pixel.First select pixels that represent two types,extract feature components,and generate a training set;then select appropriate parameters such as kernel parameters and penalty coefficients to train the classifier to obtain a support vector model;and then extract the feature components of each pixel one by one,Generate a sample to be classified sample set,and find the distance between the corresponding sample of each pixel and the hyperplane;then classify each pixel into two different classes according to the decision function to complete the segmentation of the image.SVM-based potato image segmentation method,extracting B and H components respectively in RGB color space and HSV color space to train the SVM classifier,and use the trained SVM classifier to segment the potato image and background to better discover potato sprouts s position.(5)Research on detection methods of germinated potatoes The traditional Euclidean distance algorithm is improved,and a weighted Euclidean distance algorithm is proposed.The algorithm improves the following two problems:(1)The traditional Euclidean distance is avoided,which only represents the cumulative difference between two space vectors,and ignores the effect of the difference between the corresponding single elements.(2)The similarity between eigenvectors is used to measure the similarity between the eigenvectors.For the detection of germinated potatoes,there will be a large error in the detection accuracy.The weighted Euclidean distance algorithm proposed in this paper can effectively improve the detection accuracy and can better perform potato detection.(6)Add morphological treatment to the detection of germinated potatoes.As the potato image is susceptible to noise interference during processing,which results in a lack of detection accuracy,morphological methods need to be added in subsequent processing to remove mislabeled pixels and defect locations.(7)Test results and analysisBased on Matlab R2014 a software platform,weighted Euclidean distance and traditional Euclidean distance method were used to detect 480 potato samples.The test results showed that the recognition rate of germinated potato weighted Euclidean distance was 94.6%,and the traditional Euclidean distance recognition rate was 89.2%.The detection recognition rate was 88.3%.That is,the detection accuracy of the method in this paper has been improved by 5.4% and 6.3%,respectively.For normal potatoes,the recognition rate of weighted Euclidean distance is 93.3%,the recognition rate of traditional Euclidean distance is 86.7%,and the recognition rate of manual detection is 89.6%.The detection accuracy of this method is improved by 6.6% and 3.7%,respectively.The results show that the method in this paper has a high recognition rate for germinated potatoes,and the detection accuracy of normal potatoes reaches the standard.