Abstract:Phenotypic traits at the organ scale hold significant importance in the realm of plant breeding, notably in evaluating genetic diversity, selecting innovative cultivars, and forecasting potential yield. Ground platforms, despite delivering accurate results in organ-scale phenotypic studies, frequently exhibit limitations concerning efficiency and adaptability, and can adversely affect the soil. The Unmanned Aerial Vehicle (UAV) platform emerges as an innovative medium, distinguished by its high-throughput and adaptability, thus offering immense prospects in phenotypic research. Nonetheless, its precision within the organ scale warrants further investigation. As a notable economic plant species, cotton (Gossypium spp.) provides the textile industry with its essential raw material through the fruit it bears, the cotton boll. This study, employing the characterization of cotton boll breeding materials as an example, scrutinizes the capability of UAV platforms in executing precise enumeration and yield estimation of three-dimensional (3D) cotton bolls in agricultural fields. An innovative route, the Cross-circling oblique (CCO) route, was employed in conjunction with lightweight UAVs for reconstructing a 3D point cloud of cotton bolls in the field, and juxtaposed against the traditional nadir route. A comprehensive process was also developed for characterizing the quantity, volume, and spatial distribution of cotton bolls, further facilitating yield estimation. The CCO-produced point cloud demonstrated a superior completeness compared to the nadir-produced point cloud, proving capable of effectively mitigating occlusion in the middle and lower sections, thereby procuring a more comprehensive 3D cotton boll. In terms of cotton boll counting (R 2 = 0.92 vs R 2 = 0.73) and yield estimation (seed cotton: R 2 = 0.7 vs R 2 = 0.62, lint cotton: R 2 = 0.75 vs R 2 = 0.55), the precision of the CCO-derived point cloud surpassed that of the nadir-derived point cloud considerably. Notably, yield estimations were more precise when based on cotton boll count for seed cotton, and on boll volume for lint cotton. Utilizing the high-precision CCO-produced cotton boll point cloud, an analysis was conducted on the spatial distribution of cotton bolls for all breeding materials, and the yield was subsequently projected. The contribution of this study lies in the high-throughput capture of organ-scale traits of field cotton bolls using the UAV-based CCO route, enabling accurate yield estimation. This approach may also be applicable to extracting organ-scale traits of other field crops. It demonstrates the immense potential of combining UAV platforms with the CCO route for high-throughput acquisition of crop organ-scale traits, which could become a powerful tool in plant breeding.

Organ Segmentation and Phenotypic Trait Extraction of Cotton Seedling Point Clouds Based on a 3D Lightweight Network

Cotton Phenotypic Trait Extraction Using Multi-Temporal Laser Point Clouds

Stem–Leaf Segmentation and Phenotypic Trait Extraction of Individual Plant Using a Precise and Efficient Point Cloud Segmentation Network

Three-dimensional Photogrammetric Mapping of Cotton Bolls in Situ Based on Point Cloud Segmentation and Clustering

In-field High Throughput Phenotyping and Cotton Plant Growth Analysis Using LiDAR

3D Deep Learning-based Segmentation to Reveal the Spatial Distribution of Cotton Bolls

Automatic acquisition, analysis and wilting measurement of cotton 3D phenotype based on point cloud

Segmentation of Crop Organs Through Region Growing in 3d Space

A Point-Cloud Segmentation Network Based on SqueezeNet and Time Series for Plants

Eff-3DPSeg: 3D organ-level plant shoot segmentation using annotation-efficient point clouds

Improved 3D point cloud segmentation for accurate phenotypic analysis of cabbage plants using deep learning and clustering algorithms

Cucumber Seedling Segmentation Network Based on a Multiview Geometric Graph Encoder from 3D Point Clouds

Cotton plant part 3D segmentation and architectural trait extraction using point voxel convolutional neural networks

Three-dimensional Reconstruction of Cotton Plant with Internal Canopy Occluded Structure Recovery

Automated segmentation of soybean plants from 3D point cloud using machine learning

Delving into the Potential of Deep Learning Algorithms for Point Cloud Segmentation at Organ Level in Plant Phenotyping

Research on the evolutionary history of the morphological structure of cotton seeds: a new perspective based on high-resolution micro-CT technology

PlantNet: A dual-function point cloud segmentation network for multiple plant species

3D reconstruction and characterization of cotton bolls in situ based on UVA technology

3D Reconstruction and Characterization of Cotton Bolls in Situ Based on UAV Technology

Panicle-3D: Efficient Phenotyping Tool for Precise Semantic Segmentation of Rice Panicle Point Cloud