Classification of lemon fruit ripe using convolutional network Available to Purchase

Lemon is a type of citrus fruit that grows in the tropics and has high vitamin C. Maturity in lemons affects the quality of lemons and nutrients in the lemon. Although the maturity level of lemon can be seen with vision, it cannot be measured entirely. Therefore needs technologies that can support in determining the maturity level of lemon fruit, such as machine learning of image processing technology. Image processing is used to assist humans in recognizing images or classifying objects quickly, precisely and can process with multiple data simultaneously. This study proposed a convolutional neural network (CNN) to classify the maturity of lemons. The training used a dataset available from the GitHub website with a size of 1056x1056 pixels. Then the image data enters the pre-processing stage, which consists of segmentation, cropping, and resizing. After the pre-processing stage, the image data is extracted, and a feature map is obtained, then it enters ReLU activation and finally Max Pooling to be used in the classification of lemons. There are three classes in this study, including “Unripe”, “Mature”, and “Over Ripe”. By using the CNN method, the accuracy is quite good with the SGD and Adam optimizer models. The accuracy obtained in this study reached 100% on the training data while 93.14% on the new data. SGD optimization model gaave better performance than Adam in this study.