REAL-TIME CLASSIFICATION ALGORITHM FOR RIPENESS EVALUATION OF CAYENNE PEPPER BASED ON ENHANCED IMAGE PROCESSING

Developing video processing models in classification systems requires appropriate adaptation to maximize accuracy. The object of this research is a Cayenne Pepper fruit image taken using a conveyor with recording parameters and technical aspects tailored to the needs of the processing industry. How to develop an image processing model on real-time data to classify the quality of Cayenne Pepper adaptively with conveyor devices is a problem for the industry. The concept of identification using the image processing system developed in this research uses a segmentation system that converts RGB (Red, Green, Blue) to HSV (Hue, Saturation, and Value), image masking by producing a binary image combined with the results of edge detection using the Canny operator, and then morphological operations for object edge scraping. The segmentation system model makes a strong contribution to improving the accuracy of the identification and classification system of the quality of Cayenne Pepper divided into four categories: immature, half-ripe, ripe, and rotten, using a specially designed conveyor device. In addition, the RGB average approach of the object is a new development from previous research to minimize detection errors from adjacent objects. The classification model uses the SVM method with the best parameter values on the Radial Basis Function (RBF) kernel, Cost (C) = 100 and gamma (γ) = 0.2 obtained from Grid Search, and multi-class OAA (One Against All). The approach with the segmentation system developed by the Support Vector Machine (SVM) gave the best classification result of 95.92%.