חיפוש מתקדם

Elnashef, B., Mapping and Geo-Information Engineering, Technion–Israel Institute of Technology, Haifa, Israel; Filin, S., Mapping and Geo-Information Engineering, Technion–Israel Institute of Technology, Haifa, Israel;

Plant breeding is the key to genetic improvement and increased crop yield. Breeding projects require the determination of plant phenotypes (i.e., phenotyping), a task that becomes tedious and biased when performed manually. Thus, to speed-up and improve breeding projects, novel autonomous phenotyping methods must be developed. Among the currently available methods, three-dimensional (3-D) plant modeling offers a high-resolution description of plant morphology and hence a detailed analysis of plant growth. However, advanced organ-level scale analysis of 3-D data is a complex process that requires preliminary segmentation of the 3-D point cloud into leaf- and stem-related points. Despite their obvious importance, only a few segmentation models have been reported, and many studies still perform this task manually. This paper presents a novel data-driven segmentation model as a basis for organ-level plant analysis. It shows that a point-wise computation of first- and second-order tensors is sufficient to characterize the data, without the need for pre-defined shape assumptions and constraints. Application of the model to a variety of species at different growth stages gave accurate segmentation results regardless of the species morphology and the plant growth stage. The derived information provides an organ-level phenotyping ability, and the paper shows how this ability remains accurate for various geometries and growth stages, even when the model is applied to noisy data or when the reconstruction is incomplete. An accurate, nondestructive, and autonomous organ-level phenotyping pipeline, such as the one described here, will find application for trait selection processes in breeding programs. In addition, the short processing time facilitate the use of our pipeline for precision agriculture field applications. © 2018

פותח על ידי קלירמאש פתרונות בע"מ -
הספר "אוצר וולקני"
אודות
תנאי שימוש
Tensor-based classification and segmentation of three-dimensional point clouds for organ-level plant phenotyping and growth analysis
156

Elnashef, B., Mapping and Geo-Information Engineering, Technion–Israel Institute of Technology, Haifa, Israel; Filin, S., Mapping and Geo-Information Engineering, Technion–Israel Institute of Technology, Haifa, Israel;

Tensor-based classification and segmentation of three-dimensional point clouds for organ-level plant phenotyping and growth analysis

Plant breeding is the key to genetic improvement and increased crop yield. Breeding projects require the determination of plant phenotypes (i.e., phenotyping), a task that becomes tedious and biased when performed manually. Thus, to speed-up and improve breeding projects, novel autonomous phenotyping methods must be developed. Among the currently available methods, three-dimensional (3-D) plant modeling offers a high-resolution description of plant morphology and hence a detailed analysis of plant growth. However, advanced organ-level scale analysis of 3-D data is a complex process that requires preliminary segmentation of the 3-D point cloud into leaf- and stem-related points. Despite their obvious importance, only a few segmentation models have been reported, and many studies still perform this task manually. This paper presents a novel data-driven segmentation model as a basis for organ-level plant analysis. It shows that a point-wise computation of first- and second-order tensors is sufficient to characterize the data, without the need for pre-defined shape assumptions and constraints. Application of the model to a variety of species at different growth stages gave accurate segmentation results regardless of the species morphology and the plant growth stage. The derived information provides an organ-level phenotyping ability, and the paper shows how this ability remains accurate for various geometries and growth stages, even when the model is applied to noisy data or when the reconstruction is incomplete. An accurate, nondestructive, and autonomous organ-level phenotyping pipeline, such as the one described here, will find application for trait selection processes in breeding programs. In addition, the short processing time facilitate the use of our pipeline for precision agriculture field applications. © 2018

Scientific Publication
You may also be interested in