The Support Vector Machine (SVM) Based Near-Infrared Spectrum Recognition of Leaves Infected by the Leafminers

Using the image processing and spectral analysis technology, some parameters related with leafminer infection were analyzed and confirmed, such as the damaged degrees (DD) of leaves infected by the leafminer (Liriomyza sativae Blanchard), the spectral reflectance and the sensitive wavelengths (771, 821, 891, 945, 1026, 1121, 1256, 1674, 1687 and 1933 nm) related with the DD. Using the support vector machine (SVM) the classifying models were set up to recognize the leaf spectral reflectance with different DD. The numerical experimental results showed that the recognition precisions were very high (using all 10 input vectors, the precisions were beyond 90% with the polynomial-based kernel function and beyond 97.4% with the RBF kernel function.). When the input vectors were fewer, the classification precisions were lower. Therefore, the result provides a spectral method which can be used to measure the damaged degrees of leaves infected by the leafminers with SVM. And the SVM method may help to inspect insect pests quickly