Amiel Joseph M. Lozada, Nigel L. Monsanto, Glenn B. Pepito
{"title":"基于异常的卷积神经网络图像滤波在草药植物识别中的比较研究","authors":"Amiel Joseph M. Lozada, Nigel L. Monsanto, Glenn B. Pepito","doi":"10.1109/iSemantic55962.2022.9920424","DOIUrl":null,"url":null,"abstract":"Plant identification, through the use of Convolutional Neural Networks (CNNs), has been utilized in several studies over recent years. With CNNs being almost the default approach when dealing with image processing, the researchers shifted their focus on image filtering techniques. This study determined to investigate the most effective image filter for herbal plant identification. An image dataset of eleven medicinal plants was used by the researchers, made into four copies for image processing. Three image filters were then applied to three different copies of the dataset, namely: Canny Edge Detection filter, Color Saturation filter, and Contrast Enhancement and Thresholding filter; none were applied to the fourth copy since it served as the control group of the study. The Xception model was trained using each of the processed datasets. Afterwards, the researchers discerned which CNN and image filter yielded the most accurate results during testing through the confusion matrix. It was calculated and concluded that the Color Saturation filter was the best image filtering technique to use for identifying herbal plants, achieving 100% in the metrics used during the study. The results of this study can be applied in works and systems that focus on plant identification and image processing in general.","PeriodicalId":360042,"journal":{"name":"2022 International Seminar on Application for Technology of Information and Communication (iSemantic)","volume":"271 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Comparative Study on Image Filtering For Herbal Plant Identification Using Xception Based Convolutional Neural Network\",\"authors\":\"Amiel Joseph M. Lozada, Nigel L. Monsanto, Glenn B. Pepito\",\"doi\":\"10.1109/iSemantic55962.2022.9920424\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Plant identification, through the use of Convolutional Neural Networks (CNNs), has been utilized in several studies over recent years. With CNNs being almost the default approach when dealing with image processing, the researchers shifted their focus on image filtering techniques. This study determined to investigate the most effective image filter for herbal plant identification. An image dataset of eleven medicinal plants was used by the researchers, made into four copies for image processing. Three image filters were then applied to three different copies of the dataset, namely: Canny Edge Detection filter, Color Saturation filter, and Contrast Enhancement and Thresholding filter; none were applied to the fourth copy since it served as the control group of the study. The Xception model was trained using each of the processed datasets. Afterwards, the researchers discerned which CNN and image filter yielded the most accurate results during testing through the confusion matrix. It was calculated and concluded that the Color Saturation filter was the best image filtering technique to use for identifying herbal plants, achieving 100% in the metrics used during the study. The results of this study can be applied in works and systems that focus on plant identification and image processing in general.\",\"PeriodicalId\":360042,\"journal\":{\"name\":\"2022 International Seminar on Application for Technology of Information and Communication (iSemantic)\",\"volume\":\"271 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-09-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2022 International Seminar on Application for Technology of Information and Communication (iSemantic)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/iSemantic55962.2022.9920424\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 International Seminar on Application for Technology of Information and Communication (iSemantic)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/iSemantic55962.2022.9920424","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Comparative Study on Image Filtering For Herbal Plant Identification Using Xception Based Convolutional Neural Network
Plant identification, through the use of Convolutional Neural Networks (CNNs), has been utilized in several studies over recent years. With CNNs being almost the default approach when dealing with image processing, the researchers shifted their focus on image filtering techniques. This study determined to investigate the most effective image filter for herbal plant identification. An image dataset of eleven medicinal plants was used by the researchers, made into four copies for image processing. Three image filters were then applied to three different copies of the dataset, namely: Canny Edge Detection filter, Color Saturation filter, and Contrast Enhancement and Thresholding filter; none were applied to the fourth copy since it served as the control group of the study. The Xception model was trained using each of the processed datasets. Afterwards, the researchers discerned which CNN and image filter yielded the most accurate results during testing through the confusion matrix. It was calculated and concluded that the Color Saturation filter was the best image filtering technique to use for identifying herbal plants, achieving 100% in the metrics used during the study. The results of this study can be applied in works and systems that focus on plant identification and image processing in general.