Transactions of the ASAE · 2004 · 16 citations · 0 references
ComputingComputer EngineeringComputer ScienceComputer VisionMachine LearningMachine VisionClassificationDefect DetectionFeature ExtractionImage AnalysisImage ClassificationUnsupervised LearningFruit DetectionFruit Image AnalysisImage-based Defect ClassificationSortingUnsupervised Defect ClassificationUnsupervised Defect DetectionUnsupervised Feature LearningUnsupervised Image Classification
Neural networks and unsupervised feature extraction were used to classify apple images based on whether or notthey included a stem or calyx end. In one experiment, the system successfully classified 98.4% of a test set consisting of254 near-infrared images captured at 740 nm. The network was also tested on gray-level images captured with light in thevisible range. In this test of 242 images, 5% were not classified correctly. In another classification test that included appleimages with prominent defects, 28% were misclassified. However, the majority of the errors occurred because major defectswere mistaken for the stem or calyx. In a practical implementation, errors that could lead to loss in a sorting system wouldamount to only 0.05%.