Seed cotton mulching film identification and classification method based on hyperspectral image pixel block machine learning

The invention discloses a seed cotton mulching film identification and classification method based on hyperspectral image pixel block machine learning, and the method comprises the steps: obtaining a material hyperspectral image after impurity mixing through a hyperspectral imaging device, cutting the hyperspectral image into pixel blocks corresponding to a removal requirement, extracting spatial texture features, combining the spatial texture features with spectral features, and enabling the spatial texture features to serve as the features of the pixel blocks; according to the method, a machine learning model is constructed, an impurity docking matrix is constructed, the recognition precision of continuous pixel blocks is improved, and finally online classification and recognition of materials in the pixel blocks are realized. According to the method, the importance of original spectral features is sorted by using a limit gradient lifting algorithm, a weighted spectral feature value concept is introduced, a