Proceedings of the 2013 International Conference on Advanced Computer Science and Electronics Information (ICACSEI 2013)

Image Recognition of Wheat Disease Based on RBF Support Vector Machine

Lian zhong Liu, Wu Zhang, Shuang bao Shu, Xiu Jin
Corresponding author
Plant disease, Computer vision, Image processing, Support Vector Machine.
The paper proposes an image recognition method of wheat disease. Image background is first removed by image segmentation using green feature of wheat leaf to obtain only disease pixels from original leaf image. Then disease features are calculated through 3 schemes: 1) mean values of R, G, B; 2) normalized mean values of R, G, B; 3) green ratios of R/G, B/G. Using disease features as input, image samples are trained and recognized using multi-class RBF SVM. The method has been tested on healthy leaves and leaves infected by leaf powdery mildew, stripe rust, leaf rust and leaf blight. The result shows normalized R, G, B achieved the best recognition rate up to 96%, and the overall recognition rate decreases dramatically while including more disease types in samples.
Download article (PDF)