Application of Co-occurrence Matrix on Wavelet Coefficients for Crop-weed Discrimination

The objective of this study was to discriminate the main crop from the common weeds in the field. Textural image analysis was applied to differentiate between the crop and the weeds. In the textural analysis, images were divided in square tiles, and were subjected to a wavelet transform. Wavelet transformation of digital images produces several spatial orientations which are highly effective in analyzing the information content of the images. In this study for each sub-band of wavelet coefficient, co-occurrence matrix was constructed to extract appropriate features for classification. Energy, entropy, contrast, homogeneity and inertia features were extracted from each orientation of the co-occurrence matrix. Finally, these features were fed into a multi-layer perceptron neural network to classify corn and four species of common weeds in the corn field. One hundred images were captured in normal condition of the plants in the field and were used to verify the ability of the proposed method in cropweed discrimination. Results showed that this technique was able to distinguish the corn plants with an accuracy of 99.9%, and 96% for the weeds.