Drone‐based polarization imaging system for leaf spot severity determination in peanut plants

Abstract In this study, we introduce a new approach for enhancing peanut phenotyping through a polarization imaging platform. With leaf spot disease posing significant threats to peanut ( Arachis hypogae L.) crops, our research addresses the need for accurate and efficient detection methods. Polarization imaging offers unique advantages over more traditional spectral imaging solutions. Polarization correlates strongly with the geometric properties of an object, such as surface roughness or its orientation relative to the sensor or light source. Leveraging a drone‐based system, we conducted extensive field trials, collecting approximately 30,184 images over two growing seasons and locations. Images were processed for panchromatic (400–800 nm wavelengths) degree of linear polarization (DOLP) and compared to conventional red, green, and blue (RGB) imagery against conventional visual severity scores (modified nine‐point scale). Results indicated that when attempting to determine ground truth infection severity, panchromatic DOLP alone provided 1.34 root mean square error, RGB alone provided 1.09 root mean square error accuracy, and both modalities provided 1.03 root mean square error, indicating that adding a polarization capability can enhance or augment conventional RGB scoring pipelines. We expect that polarization may allow for phenotypic scoring models to mitigate—or leverage—confounding factors related to a leaf's geometric properties without the need for 3D imaging.