Table 1 Comparison of proposed technique with existing methods
Method | Year | Segmentation technique | Feature type | Feature selection | Classifier | Disease | Accuracy |
|---|---|---|---|---|---|---|---|
[2] | 2016 | K-means clustering | Color, texture | – | Random Forest Classifier | Apple scab, apple blotch, and rot | 80%, 80%, 80% |
[35] | 2012 | K-means clustering | GCH, CCV, LBP, CLBP | – | M-SVM | Apple scab | 96.96% |
[5] | 2016 | K-means clustering | Color, texture, shape | – | M-SVM | Apple blotch, rot, and scab | 97.50%, 92.50%, 93.75% |
[29] | 2017 | Region growing algorithm | Color, texture, shape | GA and correlation-based feature selection | SVM | Rust | 94.28% |
[36] | 2016 | Otsu and watershed | SGDM, GLCM | – | NN | Grape rot, powdery mildew | 95.23%, 92.85% |
[37] | 2016 | K-means clustering | Color, texture | – | SVM | Powdery mildew | 83.33% |
Proposed | 2017 | Adaptive and quartile deviation-based segmentation with correlation coefficient | SLBP, texture, color | Entropy-rank correlation-based feature selection | M-SVM | Apple scab, apple rust, grape rot leaves, grape powdery mildew | 97.10%, 94.70%, 96.60%, 96.30 |