Image Processing Computational Algorithm for Classification of Katokkon Toraja Chili Peppers Syndromes in Greenhouse Smart Farming System

The real-time and automatic detection and classification of various plant syndromes through the adaptation of artificial intelligence (AI) computational algorithms and image processing have significantly influenced the improvement of the quality and productivity of smart greenhouse and open field farming systems during the harvest period. Therefore, this study examines a general computational algorithm to automate the detection and classification of Toraja Katokkon chili diseases. Katokkon chili is an economically important crop in Indonesia, but its cultivation process can decrease yield and harvest quality due to pest and disease attacks. Practically, image samples of the health condition of the processed plant growth will be categorized into healthy and unhealthy ones. The dataset used in this study consists of 200 images, divided equally between healthy and unhealthy samples, which are manually labeled based on the symptoms seen. The images were first processed to a resolution of 640 x 640 pixels to meet the requirements of the developed YOLOv8 computational model. The model achieved a classification accuracy of around 81.48%, with precision and recall values showing moderate performance in distinguishing between healthy and unhealthy chilies. The results show that although the designed computational algorithm shows promise for real-time detection of agricultural diseases, increasing the size of the dataset, data augmentation, and integration of attention mechanisms are required to improve detection accuracy. These remarkable results highlight the potential of utilizing AI-based systems in smart agriculture to optimize crop management and reduce losses.