Aerial imagery from unmanned aircraft systems (UASs) has been increasingly used for field phenotyping and precision agriculture. Plant height is one important crop growth parameter that has been estimated from 3D point clouds and digital surface models (DSMs) derived from UAS-based aerial imagery. However, many factors can affect the accuracy of aerial plant height estimation. This study examined the effects of image overlap, pixel resolution, and data extraction methods on estimation accuracy. An experimental field containing 16 plots with four crops (cotton, corn, grain sorghum and soybean) and four replications was set up for this study. An imaging system consisting of two consumer-grade Nikon D7100 cameras with 6000 x 4000 pixels was mounted on a rotary hexacopter for image acquisition. One camera was used to capture red-green-blue (RGB) color images, while the other was equipped with an 830-nm long-pass filter to obtain near-infrared (NIR) images. Aerial images were captured from the field at four altitudes (30 m, 60 m, 90 m and 120 m) above ground level three times during a growing season. Plant height was also measured manually from selected sampling points across the 16 plots. The RGB and NIR images taken at the four altitudes with varying overlaps were processed to create orthomosaics, 3D point clouds, DSMs and digital terrain models (DTMs). Two methods were used to extract plant height data. The first method was based on the differences between the DSMs and DTMs and the second was to estimate plant height from the 80th to 99th percentiles and maximum height derived from the 3D point clouds. The data values falling within a circular area centered at each sampling point across the 16 plots were extracted. Statistical analyses were performed to compare the differences between estimates derived from the DSMs and the 3D point clouds and ground measurements for each of the four crops. The effects of image overlap, pixel resolution and data extraction methods were analyzed. The results from this study will be useful for selecting appropriate flight parameters and data extraction methods for accurate plant height estimation using UAS imagery.
