Winter wheat plant area monitoring using GF-1 WFV imagery

This study presents a modified model for feature extraction of multivariate texture from remotely sensed imagery, which measures the spatial correlation of all the bands of a multispectral image with a designed distance metric using Cholesky decomposition of matrix and the transformation for Mahalanobis-Euclidean space. On this basis, a method of crop mapping and area extracting is developed and compared with other two traditional methods. Accuracy assessment was used to evaluate the identification results. Contrastive analysis was applied to measure the consistencies of the extracted area of winter wheat from GF-1 WFV images at three different growth stages with the reference data. Through experiments using data sets from a county in Middle China, it was found that the proposal method could significantly improve the classification accuracy and can also promoted application.