The Detection of Kiwifruit Sunscald Using Spectral Reflectance Data Combined with Machine Learning and CNNs

Sunscald in kiwifruit, an environmental stress caused by solar radiation during the summer, reduces fruit quality and yields causes economic losses. The efficient timely detection of sunscald similar diseases is a challenging task but helps to implement measures control stress. This study provides high-precision models relevant spectral information on kiwifruit physiology for statuses, including early-stage sunscald, late-stage anthracnose, healthy. Primarily, laboratory, 429 groups reflectance data leaves four statuses were collected analyzed using hyperspectral reflection acquisition system. Then, multiple modeling approaches, combined preprocessing methods, feature extraction algorithms, classification designed extract bands evaluate performance detect kiwifruit. Finally, different stages under anthracnose interference was accomplished. As influential bands, 694–713 nm, 758–777 780–799 1303–1322 nm extracted. overall accuracy, precision, recall, F1-score values reached 100%, demonstrating ability all with 100% accuracy. It concluded that processing moving average standard normal variable transformations (MS) could significantly improve data; near-infrared support vector machine visible convolutional neural network MS (NIR-MS-SVM VIS-MS-CNN) established as techniques 25.58% higher accuracy than single machine. VIS-MS-CNN model convergence stable cross-entropy loss 0.75 training 0.77 validation. developed this will orchard management efficiency increase researchers’ understanding physiology.