EasyIDP V2.0: An Intermediate Data Processing Package for Photogrammetry-Based Plant Phenotyping

Photogrammetry techniques (3D reconstruction) have been commonly used in applications of high-throughput plant phenotyping. However, some intermediate processing still requires manual operation, which is often time-consuming and labor-intensive. For example, only some Regions of Interest (ROIs) rather than the entire plot or field are worth analyzing, and manually cropping them from images, geotiffs, and point clouds necessitates alternating use of different software programs. In addition, since the orthomosaic is stitched from different image views, it is easily affected by wind, causing ghost effects and seamline distortion, the raw images are clearer and more competitive for image analysis, but matching the corresponding regions from raw images to actual field positions is challenging. Thus, to decrease the workload required for the intermediate data processing, we improved our previous EasyIDP package for several commonly used solutions, including (1) cropping and saving ROIs from the entire field for raw images, geotiffs (orthomosaic), and point clouds; (2) finding the corresponding ROI from a 3D model to raw 2D images (backward projection), and returning image analysis results back to the 3D model (forward projection). By using this package, researchers are able to crop ROIs from digital orthomosaic (DOM), improve ROI image quality, and map deep learning results on raw images to real-world positions, with just a few lines of code.