Development of a Damage Level Classification Model for Banana Base on Image Processing and Machine Learning

This study focuses on the design and development of a model to classify the levels of banana damage using image processing combined with machine learning. The dataset consists of 568 digital images of bananas, each with a resolution of 615x440 pixels, captured using a Nikon D750 camera with settings of ISO 800, 1/200 shutter speed, f/4 aperture, and a 50mm lens at a distance of approximately 25 cm from the object. The dataset categorizes banana damage into three levels: good condition 203 images, medium damage 218 images, and poor or unacceptable quality 147 images. Image processing techniques were employed to prepare the dataset, starting with converting color images to grayscale, followed by feature extraction to transform the image data into numerical data. The resulting dataset was used to create classification models, which were then evaluated using five algorithms: Multiclass Classifier Updateable, SMO, Multilayer Perceptron, Iterative Classifier Optimizer, and Random Forest. The experimental design involved splitting the data into training and testing sets using a simple random sampling approach with a 70:30 ratio. The experimental results showed the highest accuracy rates achieved by each algorithm as follows: 91.11%, 88.89%,84.44%,82.22%, and 80.00%, respectively.