Identification of corn and weeds on the leaf scale using polarization spectroscopy

In order to explore the feasibility of accurate identification between crop and weed species using polarization spectroscopy, Field Imaging Spectral System (FISS) was utilized with a polaroid configuration to collect imagery data of corn and five kinds of weeds in the laboratory. Through comparisons and analysis of spectral response curves, characteristic difference and identification model accuracy between corn and weeds under four polarization angles, it was found that there was a consistency for spectral changing trends between corn and five kinds of weeds, and the spectral intensity of corn and weeds displayed highest in the no polarization status. Moreover, the selected sensitive bands under four polarization conditions to distinguish corn and weed species indicated that there were similar characteristics, as well as some differences. Finally, for overall accuracy of the identification models between corn and weeds, and the corresponding Kappa coefficients were all more than 90%. The accuracy was the highest, close to 100%, when data were measured at 0 polarization angle. Therefore, polarization technology can be used to identify corn and weeds on the leaf scale, providing an important data foundation for further application on a field scale.