GUI based Detection of Unhealthy Leaves using Image Processing Techniques

Abstract

Increasing the agricultural productivity improves the Indian economy. Keeping this as objective, in order to achieve an efficient and smart farming system, identification of unhealthy leaf using image processing techniques is contributed in this paper. For this, ladies finger plant leaves are chosen and examined to find an early stage of various diseases such as yellow mosaic vein, leaf spot, powdery mildew etc. Leaf images are captured, processed, segmented, features extracted, and classified to know if they are healthy or unhealthy. Due to practical limitations in climatic conditions and other terrain regions, noisy image data sets are also created and taken into consideration. K-means clustering is used for segmentation and for classification, SVM and ANN are used. This work uses PCA to reduce the feature set. Results show that, the average accuracy of detection in SVM and ANN are 85% and 97% respectively. Without noise they are observed to be 92% and 98% respectively. This work paves the way to reach complete automation in agricultural industries.