Classifying adaxial and abaxial sides of diseased citrus leaves with selected hyperspectral bands and YOLOv8

Authors: FREDERICK, QUENTIN - University Of Florida; BURKS, THOMAS - University Of Florida; YADAV, PAPPU - South Dakota State University; Qin, Jianwei - Tony Qin; Kim, Moon; DEWDNEY, MEGAN - University Of Florida

Submitted to: Smart Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/7/2024
Publication Date: N/A
Citation: N/A

Interpretive Summary: Citrus diseases cause financial losses as a result of smaller fruits, blemishes, premature fruit drop and eventual tree death. Since the symptoms generally first appear on the leaves, detection of the citrus diseases via leaf inspection can permit more effective mitigation tactics and more intelligent management of groves. This study developed an imaging method called hyperspectral reflectance imaging to detect and classify citrus leaf diseases. Both sides of leaves with the citrus diseases of canker, Huanglongbing (HLB), greasy spot, melanose, scab, zinc deficiency, and a control set were imaged using a portable hyperspectral imaging system. A machine learning model was trained to classify each side of the leaves and achieved the best accuracy of 87% for the front side using a selected subset of the image data. The hyperspectral reflectance imaging with the machine learning techniques would assist the citrus industry and regulatory agencies (e.g., FDA and USDA APHIS) in enforcing standards for the quality and safety of citrus-related food and beverage products.

Technical Abstract: Citrus greening disease (HLB) and citrus canker are diseases afflicting Florida citrus groves, causing financial losses through smaller fruits, blemishes, premature fruit drop and/or eventual tree death. Often, symptoms of these resemble those of other defects/infections. Early detection of HLB and canker via in-grove leaf inspection can permit more effective mitigation tactics and more intelligent management of groves. Autonomous, vision-based disease scouting in a grove offers a financial benefit to the Florida citrus industry. This study investigates the potential of hyperspectral reflectance imagery (HSI) for detecting and classifying these conditions in the presence of other, less consequential leaf defects. Both sides of leaves with visible symptoms of HLB, canker, zinc deficiency, scab, melanose, greasy spot, and a control set were collected and imaged with a line-scan hyperspectral camera. Spectral bands from this imagery were selected using two methods: an unsupervised method based on principal component analysis (PCA), a supervised method based on linear discriminant analysis (LDA), as well as randomly selected bands as a control. The YOLOv8 network architecture was trained to classify each side of these leaves with each band combination. LDA-selected bands from the front of the leaves yielded an overall classification accuracy of 87.09%, with higher recall of melanose and precision of control than any other model tested. Leaves with HLB and zinc deficiency were classified most accurately, with both band selection methods yielding F1 scores of at least 0.955 and 0.934, respectively. These findings favor the use of supervised band selection for HSI-based in-grove disease detection.