A Polarized-Spectrum-Based Vegetation Index to Improve Vegetation Health Detection Accuracy by Removing Shadows

Vegetation plays an extremely important role in maintaining the quality and stability of the ecological environment and can be monitored using remote sensing, which is interfered by shadows, resulting in reduced accuracy of chlorophyll content retrieval, and eliminating this shadow interference is challenging. In this paper, a ground-based polarized-multispectral imaging system is built and a polarized-spectrum-based shadow removal method is proposed to eliminate the interference of vegetation shadows and improve the accuracy of vegetation health status detection. Two shadow removal vegetation indices, Polarization-spectrum-based Shadow Removal-Simple Ratio Index (PSR-SR) and Polarization-spectrum-based Shadow Removal-Normalized Difference Vegetation Index (PSR-NDVI), were developed based on SR and NDVI, to evaluate the impact of shadows on vegetation health detection. Compared with the typical un-polarized vegetation indices, PSR-SR and PSR-NDVI can effectively eliminate the vegetation shadow effects with the retrieval accuracy of chlorophyll content (PSR-SR: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>=0.89, RMSE=0.033 and PSR-NDVI: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>=0.95, RMSE=0.053). Different illumination analysis shows that this method is suitable for various illumination environments, and has excellent shadow elimination performance in a low illumination environment of 1.13 lx (PSR-SR: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>= 0.77, RMSE=0.028) (PSR-NDVI: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>=0.91, RMSE=0.053). Moreover, it shows that this method still has excellent shadow elimination performance for real outdoor applications (PSR-SR: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>=0.85, RMSE=0.044) (PSR-NDVI: R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>=0.92, RMSE=0.05). In summary, our method shows the potential of utilizing polarization and spectrum to remove vegetation shadows and provides a strategy for eliminating vegetation shadows with different health states to achieve accurate and efficient vegetation health detection.