Classification of peanut variety based on hyperspectral imaging and improved extreme learning machine

Peanut as an important crop, plays an important role in agricultural production, which is rich in edible vegetable oil and protein. The variety of peanut affects the content of vegetable oil and protein. Therefore, the classification of peanut variety can better promote the sustainable development of agriculture. In this study, hyperspectral imaging technology is used to achieve peanut variety classification. In addition, the spatial-spectral extreme learning machine (SS-ELM) is proposed to process the hyperspectral data to get the final classification label. To fully explore the spatial structure information of hyperspectral data, propagation filtering is integrated into the framework of extreme learning machine (ELM). The average accuracy of the improved ELM model on five varieties of peanuts dataset (Luhua 11, Dabaisha, Xiaobaisha, Fenghua, and Luohanguo 308) is 98.32%, which is higher than other classic models. The experimental results show that the improved ELM can classify peanut of different varieties by hyperspectral imaging.