Hyperspectral Imagery Vegetation Index and Temporal Analysis for Corn Yield Estimation

摘要

Aerial hyperspectral imagery has been used to find the temporal relationship between image and corn yield. A total of five hyperspectral images were taken during the growing season. For each image, the optimal vegetation index was selected among many candidate vegetation indices. At the same time, the optimal band subset was selected to calculate the vegetation index. The optimal band subset has the minimum number of bands and represents the most significant image bands (or wavelength) for yield prediction. The optimization process used the EAVI (Evolutionary Algorithm based Vegetation Index generation) algorithm. Results showed that the EAVI algorithm generated the best vegetation index among many comparison indices for yield estimation. For image taken at different date, the algorithm selected a different optimal vegetation index and image bands. The most common sensitive wavelength identified was in the red edge at 700 nm and in the NIR region at 826 nm. This study showed that images taken from the beginning of full canopy coverage to the corn ear formation period provided the best and stable result for corn yield estimation. It is suggested that this period of time during the growing season would have great potential for remote sensing based corn yield prediction.