Orange Fruit Disease Detection and Classification Using AI-Based Techniques

Orange fruit cultivation is a significant economic activity in Ethiopia. However, many diseases pose a major threat to its productivity and production in many areas of our country. Currently, the detection process is done manually, which is inaccurate, inefficient, expensive, time-consuming, and requires special people. Therefore, this research study is AI-Based Orange Fruit Disease Detection and Classification with Python 3.12.3. This has great potential to enhance early disease identification, reduce crop losses, and increased food security in the country. The paper started by collecting 1252 orange fruits from different environmental locations in the Tigray Region using digital camera. These images were collected as disease categories such as orange canker (201 images), orange scabs (206 images), orange melanosis (88 images), orange greening (369 images), as well as healthy (388 images).After that, the dataset is divided into 20% for testing, for validation 10% and 70% for training. Then, image preprocessing technique is applied on the image to removing the noise and then color threshold technique is used to segment. In feature extraction, texture, shape and color features are considered to facilitate disease detection and classification. Subsequently, a deep learning approach using Convolution Neural Networks (CNN) is then implemented, along with three machine learning approaches - Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Machine (SVM), to recognize orange fruit diseases in the early stages of growth. The proposed CNN model achieves an overall accuracy rate of 99%, recall of 99% and precision of 99%, outperforming SVM, RF, and ANN in disease identification.