Multi-Channel Feature Fusion Extraction Method Based on Wavelet Packet in Plant Diseases' Diagnosis

To improve the accuracy of plant disease identification, on the basis of wavelet packet image and YUV color space, multi-channel feature fusion extraction algorithm has been proposed. Firstly, the input image is converted into the Y, U, V channels sub-graph, respectively. Then the fusion features based on wavelet packet energy, the Y, U and V-component images are calculated and as the inputs of disease image recognition model. Finally, we compare several wavelet basis, wavelet decomposition levels, color models and feature extraction methods to verify the robustness of our algorithm. The experimental results show that when the particle swarm optimization support vector machine (PSO-SVM) and the ‘haar’-based 2-level wavelet packet decomposition are used, the highest average F1 score is 0.963. Compared with the feature extraction method based on wavelet transforming, the average Fl score of the proposed approach is improved by 0.277.