Information Extraction Method of Dendrolimus punctatus Based on Multi-scale Texture and Spectral Features

Taking Shaxian County of Fujian Province as the research area, the SPOT-5 multi-spectral image and panchromatic image were merged, the texture quantity was extracted based on the gray level co-occurrence matrix method, and the spectral band was combined. The support vector machine classification method was used to extract the pest information. Exploring the influence of texture features on the accuracy of pest monitoring information extraction. The results show that the support vector machine classification method combining multi-scale texture and spectral features has the highest total accuracy of pest information extraction, which is 80.48%. The support vector machine classifier method combined with single-scale texture and spectral features has the second highest accuracy of pest information extraction, which is 78.81%. Based on the maximum likelihood method of spectral features, the total accuracy of pest information extraction is the lowest, which is 70.48%. The support vector machine classifier method combining multi-scale texture and spectral features has better surface performance and reduces the fine spots on the surface. Therefore, extracting multi-scale textures combined with spectral features enriches the amount of image information and helps to improve the extraction accuracy of pest information.