Analysis of the effects of cover crop residue on hyperspectral reflectance discrimination of soybean and weeds via Haar transform

A Haar-based automatic classification system is designed to discriminate between the hyperspectral reflectance curves of soybean and weeds. The study focuses on whether or not the cover crop residue influences the ability to distinguish grass from soybean when using hyperspectral reflectance curves. Four sets of experiments with or without cover crop residue associated with using or not using conventional tillage are designed and conducted. In each experiment, a three-class classification problem is solved using the leave-one-out test method based on the nearest mean classifier. The classification accuracies and the associated confidence intervals are computed to evaluate the classification performance of the system. The experimental results show that both cover crop residue and conventional tillage have little influence on overall classification accuracies of the automated classification system. In general, the system provides a classification accuracy of over 90%, which indicates a promising discrimination capability of the automated Haar-based classification system