High-throughput volumetric reconstruction for 3D wheat plant architecture studies

Abstract

For many tiller crops, the plant architecture (PA), including the plant fresh weight, plant height, number of tillers, tiller angle and stem diameter, signi¯cantly a®ects the grain yield. In this study, we propose a method based on volumetric reconstruction for high-throughput three-dimensional (3D) wheat PA studies. The proposed methodology involves plant volumetric reconstruction from multiple images, plant model processing and phenotypic parameter estimation and analysis. This study was performed on 80 Triticum aestivum plants, and the results were analyzed. Comparing the automated measurements with manual measurements, the mean absolute percentage error (MAPE) in the plant height and the plant fresh weight was 2.71% (1.08 cm with an average plant height of 40.07 cm) and 10.06% (1.41 g with an average plant fresh weight of 14.06 g), respectively. The root mean square error (RMSE) was 1.37 cm and 1.79 g for the plant height and plant fresh weight, respectively. The correlation coe±cients were 0. plant height and plant fresh weight, respectively. Additionally, the proposed methodology, including plant reconstruction, model processing and trait extraction, required only approximately 20 s on average per plant using parallel computing on a graphics processing unit (GPU), demonstrating that the methodology would be valuable for a high-throughput phenotyping platform.

Crop Phenomics and High-Throughput Phenotyping: Past Decades, Current Challenges, and Future Perspectives

“…Together with GWASs and 51 i-traits, the HRPF can dissect the complex DR traits into heritable and simple i-traits and discover new genes for DR (Guo et al, 2018a). In addition, with minor adjustments of the image analysis pipeline, the HRPF can be extended to phenotype other species, including 3D wheat plant architecture (Fang et al, 2016), the leaf traits of rape seedlings (Xiong et al, 2017), and the genetic architecture of variation in maize growth (Zhang et al, 2017). However, the establishment of high-throughput conveyor-based phenotyping systems is costly and time consuming and requires in-depth knowledge of engineering and computational sciences to maintain function and flexibility.…”

“…However, the hyperspectral sensor is costly, and the processing of hyperspectral data (gigabyte per sample) can be complicated (Mahlein et al, 2018). 5Compared with 2D imaging, 3D imaging techniques, which mainly include image-based (Fang et al, 2016) and laser scanning-based (Paulus et al, 2014) techniques, can generate 3D models and obtain more spatial and volumetric traits. According to the merits of the different imaging technologies, the current trend is to combine multiple imaging techniques.…”

Performance Evalution of 3D Keypoint Detectors and Descriptors for Plants Health Classification

“…Different traits like plant volume, leaf area, and stem length can be estimated by using 3D plant model. Wei et al obtained some volumetric features after making 3D reconstruction of plant by silhouette method [12]. Extracting features from surfaces meshes is developed recently.…”

Three-Dimensional Point Cloud Reconstruction and Morphology Measurement Method for Greenhouse Plants Based on the Kinect Sensor Self-Calibration

“…As a result, true 3D images of plants cannot be captured, and the 3D reconstruction of plants cannot be achieved at a single AOV. Instead, multiview stereo reconstruction is required for plants [35][36][37][38][39]. Therefore, rapid multiview registration is key to achieving the high-throughput 3D phenotyping of greenhouse plants.…”