Early-Stage Brown Spot Disease Recognition in Paddy Using Image Processing and Deep Learning Techniques

India is an agricultural country. Paddy the main crop here on which livelihood of millions people depends. Brown spot disease caused by fungus most predominant infection that appears as oval and round lesions paddy leaves. If not addressed time, it might result in serious loss. Pesticide use for plant treatment should be limited because raises costs pollutes environment. Usage pesticide loss both can minimized if we recognize a timely manner. Our aim to develop simple, fast, effective deep learning structure early-stage brown detection utilizing severity estimation using image processing techniques. The suggested approach consists two phases. In first phase, infected leaf dataset partitioned into sets named developed stage spot. This partition done basis calculated severity. Infection computed ratio pixel count total count. Total counts are determined segmenting region from background Otsu's thresholding technique. Infected regions Triangle segmentation. second fully connected CNN architecture built automatic feature extraction classification. CNN-based classification model trained validated spot, healthy leaves images rice plants. Early-stage used training validation same obtained phase 1. experimental analysis shows proposed recognition approach. accuracy found 99.20%. method compared with those existing methods have diseases. It observed performance our significantly better than methods.