EasyIDP: A python package for intermediate data processing in 3D based plant phenotyping

Abstract

Abstract Background:The use of 3D based high-throughput phenotyping improves theefficiency of crop management and monitoring practices. Thestructure-from-motion and multi-view stereo photogrammetry (SfM-MVS)technique, applicable to common RGB digital cameras, has been widely used forthis and can be implemented by many commercial and open-source tools. Byusing such tools, several outputs such as digital orthophoto map (DOM), digitalsurface model (DSM), and point cloud data (PCD) can be generated. However,there is a gap between these outputs and the final 3D plant phenotyping. Forexample, calculating plant height and canopy ground cover requires thesegmentation of each plot from the whole DOM, DSM, or original image. Theseintermediate processes are time-consuming, and to the best of our knowledge,there are no easy-to-use alternatives currently available. Results: In this study, a software package called EasyIDP (easy intermediatedata processor) was developed to link the products of SfM-MVS techniques with3D based plant phenotyping. A lotus (Nelumbo nucifera) breeding field was usedto demonstrate the following points: 1) clipping (segmenting) SfM-MVS productsaccording to a given plot boundary or region of interest (ROI); 2) transformingthe ROI of the SfM-MVS products into high-quality raw images to assist inobject detection; and 3) evaluating the accuracy of the previous transformationusing manual annotation. Conclusions: The proposed intermediate data processing tool showed anacceptable accuracy and potential to process the products from SfM-MVStechniques. By using the EasyIDP, a bridge between SfM-MVS products andplant phenotyping was conveniently achieved.

Plant phenotyping relevance

EasyIDPはSfM-MVS生成物を植物表現型抽出へ接続する中間処理ソフトウェアとして開発・評価されており、表現型取得ワークフローが中心である。

Code and data availability

The paper is a software article for EasyIDP, whose source code is publicly released on GitHub, and the authors explicitly state that the example data (UAV/SfM-MVS phenotyping case-study data) and Jupyter notebook analysis codes are available in a companion public repository (EasyIDP.paper). Both are paper-specific,公开,和