Real-time oil palm FFB ripeness grading system based on ANN, KNN and SVM classifiers

The high accuracy performance, cost and processing time are commonly the key factors for evaluate and validate the systems of agricultural crop quality inspections. External and internal grading system based on image and signal processing and analysis were implemented based on general steps as acquisition, pre-processing, segmentation, feature measurement, and classification for fruit maturity grading system. The classification step using supervised machine learning classifiers was commonly based on the method namely; two-thirds for training and the other third for testing. In the other hand, the purpose of the supervised machine learning is to construct predicted model for the training data of the distribution of the different class labels as known features which can be used for the further testing data as unknown features. In this paper, real-time oil palm FFB ripeness grading system based on ANN, KNN and SVM Classifiers was demonstrated. System training and testing were implemented in order to appropriately classify the ripeness of oil palm FFB based on the external feature of the fruit. System evaluation has been carried out based on the system performance, processing time and the system cost in order to enhance the system methodology for classification proficiency. The results show that the real-time oil palm FFB ripeness grading system has achieved the highest accuracy 93 % and the fastest image processing speed 0.40 (s) by using BGLAM texture feature based on ANN classifier compared to the other feature extraction techniques and machine learning classifiers.