High throughput canopy characterization of a commercial apple orchard using aerial RGB imagery

Rapid assessment of fruit crop canopy attributes is essential to drive orchard management decisions. Therefore, this study was aimed to estimate canopy volume from high throughput aerial imagery. A small unmanned aerial system with a consumer-grade visible range (RGB) camera was flown in grid missions at 45 m above ground level (1.3 cm/pixel) and camera inclinations of 90° and 65° from vertical, respectively, over a commercial apple orchard trained in modern v-trellis architecture. Manual ground reference measurements of tree height, width, and volume were collected for 12 trees sampled randomly within the orchard. These trees were identified in imagery data by strategically placed reference markers on the ground. Dense geospatial 3D point clouds of the orchard trees were derived from the collected imagery on a photogrammetry platform. Three different 3D surface reconstruction methods were used to estimate canopy volume for each of the sample trees. Convex hull (r = 0.64) and alpha-shape (r = 0.62) methods correlated well with the manual volume calculations based on trapezoidal geometry whereas the voxel grid method (r = 0.25) showed a weaker correlation. The methods will be further evaluated for more accurate estimates. The canopy volume estimated by aerial images may be used to guide commercial tree fruit growers towards site-specific and resource-efficient orchard management.