Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects
Citrus Black Spot (CBS) causes considerable damage to the Florida citrus industry. Early detection of CBS, especially in the presence of other peel blemishes, would enable better mapping and control of CBS spread, reduce wasted fruit, and permit early removal of culls from the packing stream. Orange...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-12-01
|
| Series: | Smart Agricultural Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772375523001934 |
| _version_ | 1797390342095372288 |
|---|---|
| author | Quentin Frederick Thomas Burks Adam Watson Pappu Kumar Yadav Jianwei Qin Moon Kim Mark A. Ritenour |
| author_facet | Quentin Frederick Thomas Burks Adam Watson Pappu Kumar Yadav Jianwei Qin Moon Kim Mark A. Ritenour |
| author_sort | Quentin Frederick |
| collection | DOAJ |
| description | Citrus Black Spot (CBS) causes considerable damage to the Florida citrus industry. Early detection of CBS, especially in the presence of other peel blemishes, would enable better mapping and control of CBS spread, reduce wasted fruit, and permit early removal of culls from the packing stream. Oranges whose peels bore the symptoms of four defects/disease (CBS, greasy spot, melanose, and wind scar), as well as a normal control group, were imaged with a hyperspectral imaging system. Principal Component Analysis- (PCA) and Linear Discriminant Analysis (LDA) -based methods were employed to select bands from these images, and a custom convolutional neural network (CNN) for feature extraction was trained with these bands. The extracted features permitted classification of the peel conditions with four classifiers: SoftMax, Support Vector Machines (SVM), Random Forest Classifier (RFC), and K-Nearest Neighbors (KNN). A mean overall accuracy of 94.9 % was achieved using an SVM classifier on five bands selected with PCA, and 90.2 % with LDA-selected bands. These results show the potential of CNNs to extract features for automated postharvest citrus inspection. |
| first_indexed | 2024-03-08T23:09:23Z |
| format | Article |
| id | doaj.art-5702abd5d6fd44b19a9bd011fdd5b33f |
| institution | Directory Open Access Journal |
| issn | 2772-3755 |
| language | English |
| last_indexed | 2024-03-08T23:09:23Z |
| publishDate | 2023-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Smart Agricultural Technology |
| spelling | doaj.art-5702abd5d6fd44b19a9bd011fdd5b33f2023-12-15T07:27:16ZengElsevierSmart Agricultural Technology2772-37552023-12-016100365Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defectsQuentin Frederick0Thomas Burks1Adam Watson2Pappu Kumar Yadav3Jianwei Qin4Moon Kim5Mark A. Ritenour6Department of Agricultural and Biological Engineering, PO Box 110570, University of Florida, Gainesville, FL 32611-0570, USADepartment of Agricultural and Biological Engineering, PO Box 110570, University of Florida, Gainesville, FL 32611-0570, USA; Corresponding author.Department of Agricultural and Biological Engineering, PO Box 110570, University of Florida, Gainesville, FL 32611-0570, USADepartment of Agricultural and Biological Engineering, PO Box 110570, University of Florida, Gainesville, FL 32611-0570, USAUSDA/ARS Environmental Microbial and Food Safety Laboratory, Beltsville Agricultural Research Center, Beltsville, MD 20705, USAUSDA/ARS Environmental Microbial and Food Safety Laboratory, Beltsville Agricultural Research Center, Beltsville, MD 20705, USADepartment of Horticultural Sciences, 2199 South Rock Road, University of Florida, Fort Pierce, FL 34945-3138, USACitrus Black Spot (CBS) causes considerable damage to the Florida citrus industry. Early detection of CBS, especially in the presence of other peel blemishes, would enable better mapping and control of CBS spread, reduce wasted fruit, and permit early removal of culls from the packing stream. Oranges whose peels bore the symptoms of four defects/disease (CBS, greasy spot, melanose, and wind scar), as well as a normal control group, were imaged with a hyperspectral imaging system. Principal Component Analysis- (PCA) and Linear Discriminant Analysis (LDA) -based methods were employed to select bands from these images, and a custom convolutional neural network (CNN) for feature extraction was trained with these bands. The extracted features permitted classification of the peel conditions with four classifiers: SoftMax, Support Vector Machines (SVM), Random Forest Classifier (RFC), and K-Nearest Neighbors (KNN). A mean overall accuracy of 94.9 % was achieved using an SVM classifier on five bands selected with PCA, and 90.2 % with LDA-selected bands. These results show the potential of CNNs to extract features for automated postharvest citrus inspection.http://www.sciencedirect.com/science/article/pii/S2772375523001934Citrus inspectionCitrus black spotConvolutional neural networkFeature extractionFruit inspectionImage classification |
| spellingShingle | Quentin Frederick Thomas Burks Adam Watson Pappu Kumar Yadav Jianwei Qin Moon Kim Mark A. Ritenour Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects Smart Agricultural Technology Citrus inspection Citrus black spot Convolutional neural network Feature extraction Fruit inspection Image classification |
| title | Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| title_full | Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| title_fullStr | Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| title_full_unstemmed | Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| title_short | Selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| title_sort | selecting hyperspectral bands and extracting features with a custom shallow convolutional neural network to classify citrus peel defects |
| topic | Citrus inspection Citrus black spot Convolutional neural network Feature extraction Fruit inspection Image classification |
| url | http://www.sciencedirect.com/science/article/pii/S2772375523001934 |
| work_keys_str_mv | AT quentinfrederick selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT thomasburks selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT adamwatson selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT pappukumaryadav selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT jianweiqin selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT moonkim selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects AT markaritenour selectinghyperspectralbandsandextractingfeatureswithacustomshallowconvolutionalneuralnetworktoclassifycitruspeeldefects |