Enhanced Plant Disease Detection Model Using Yolov10 with De-Hazing and Histogram Equalization

Disease threatens farmers by reducing crop yield and global food availability. Plant disease detection models had been deployed to mitigate these challenges, but poor image quality due to environmental factors and low quality devices affects the accuracy of the approaches. This study enhances disease detection using YOLOV-10 with de-hazing and histogram equalization techniques. Maize and cassava plants were selected for primary data collection, yielding 2,141 images (1,024 maize, 1,117 cassava). The PlantVillage dataset provided 54,303 images across nine plant classes, totalling 56,444 images. De-hazing and histogram equalization improved the YOLOV-10 model for better detection accuracy. Performance evaluation over 200 epochs used precision, recall, bounding box loss, and distribution focus loss metrics. The model achieved a bounding box loss of 0.3, a precision of 0.91 (indicating minimal false detections), and a recall approaching 1, confirming its ability to detect diseased plants accurately. With 94% accuracy, it outperformed many reviewed models.