Paddy is one of the major crops in India.The various leaf diseases affect the crop yield and it is difficult to detect these diseases efficiently by farmers with their limited knowledge. These diseases directly or indirectly affect paddy crop yield. To ensure better yield, it is more essential to detect and classify diseases with prior time. In this paper, we are proposing a system that detects and classify paddy leaf diseases by using YUV color space. Initially, the input RGB color space image is transformed to YUV color space, further converted to binary image by eliminating false positive using proposed Binary conversion and segmentation in YUV color space, extracted disease infected region boundary by applying Maximum Gradient Difference (MGD) and sobel operator. Further, we have extracted 13-features from segmented image and fed to Support Vector Machine (SVM) classifier, which classifies Healthy leaves, Rice Blast, Brown Spot, leaf Blight and Hispa paddy leaf diseases. The proposed system trained and tested on the Rice Diseases Image Dataset: collected from Kaggle along with our own dataset. The proposed system gives accuracy of 94.64%, 92.59%, 93.61%, 88.23% and 97.84% for Brown Spot, Leaf blast, Hispa, Leaf blight and Healthy leaves respectively for combined dataset (Kaggle dataset+ our own dataset), it is noticed that the larger the dataset for SVM classifier, better in accuracy.

Binary conversion, Crop yield, Maximum Gradient Difference, Support Vector Machine (SVM), Kaggle Dataset.