Plant Phenomics最新文献

Accurate acquisition of forest spatial competition and tree 3D structural phenotype parameters is crucial for exploring tree-environment interactions. However, due to the occlusion between tree crowns, current UAV-based and ground-based LiDAR struggles to capture complete crown information in dense stands, making parameter extraction challenging such as maximum crown width height (HMCW). This study proposes a canopy spatial relationship-based method for constructing forest spatial structure units and employs five ensemble learning techniques to train 11 machine learning model combinations. By coupling spatial competition with phenotype parameters, the study identifies the optimal fitting model for HMCW of Chinese fir. The results demonstrate that the constructed spatial structure units align closely with existing research while addressing issues of incorrectly selected or omitted neighboring trees. Among the 10,191 trained HMCW models, the Bagging model integrating XGBoost, Random Forest (RF), Support Vector Regression (SVR), Gradient Boosting (GB), and Ridge exhibited the best performance. Compared to the best single model (RF), the Bagging model achieved improved accuracy (R² ​= ​0.8346, representing a 1.6 ​% improvement; RMSE ​= ​1.4042, reduced by 6.66 ​%; EVS ​= ​0.8389; MAE ​= ​0.9129; MAPE ​= ​0.0508; and MedAE ​= ​0.5076, with corresponding improvements of 1.63 ​%, 1.49 ​%, 0.1 ​%, and 7.06 ​%, respectively). This study provides a viable solution for modeling HMCW in all species with similar structural characteristics and offers a method for extracting other hard-to-measure parameters. The refined spatial structure units better link 3D structural phenotypes with environmental factors. This approach aids in canopy morphology simulation and forest management research.

Location-based methods for counting rice panicles have often been underestimated, primarily due to their perceived inferior performance when compared to detection-based techniques. However, we argue that the potential of these location-based methods has not been fully realized, largely owing to the limitations of existing model architectures. In response to this challenge, we introduce LKNet, an innovative model developed on the foundation of the location-based framework P2Pnet. To enhance the performance of panicle counting across diverse types and growth stages, we implemented several key strategies. Firstly, we reconstructed the localization loss function as a predictive probability distribution to reduce the influence of manual labeling. Additionally, we dynamically adapted the receptive field to better accommodate different panicle types through the use of large kernel convolutional blocks. We evaluated LKNet on several publicly available counting task datasets and achieved state-of-the-art performance on the Diverse Rice Panicle Detection dataset. Furthermore, we employed a rice panicle dataset collected at an altitude of 7 ​m, which includes various panicle types and growth stages for model training and evaluation. The results showed that LKNet effectively accommodates variations in panicle morphology, with R² values ranging from 0.903 to 0.989. These findings highlight LKNet's potential to enhance precision in panicle counting in rice breeding programs.

Accurate and real-time monitoring true leaf area index (LAI) is an essential for assessing crop growth status and predicting yields. Conventional LAI inversion approaches have been constrained by insufficient data representativeness and environmental variability, particularly when applied across interannual variations and different phenological stages. This study presented a novel methodology integrating three-dimensional radiative transfer modeling (3D RTM) with knowledge-guided deep learning to address these limitations. We developed a knowledge-guided convolutional neural network (KGCNN) architecture incorporating 3D canopy structural physics, enhanced through transfer learning (TL) techniques for cross-temporal adaptation. The KGCNN model was initially pre-trained on synthetic datasets generated by the large-scale remote sensing scattering model (LESS), followed by domain-specific fine-tuning using 2021 field measurements, and culminating in cross-year validation with 2022-2023 datasets. Our results demonstrated significant improvements over conventional approaches, with the 3D RTM-based KGCNN achieving superior performance compared to 1D RTM implementations (PROSAIL + CNN + TL). Specially, for the 2022 dataset, the overall R² increased by 0.27 and RMSE decreased by 2.46; for the 2023 dataset, the overall RMSE decreased by 1.62, compared to the PROSAIL ​+ ​TL method. Our method (3D RTM ​+ ​KGCNN ​+ ​TL) delivered superior LAI retrieval accuracy on the two-year datasets compared to LSTM ​+ ​TL, RNN ​+ ​TL, and 3D RTM ​+ ​RF models. This study also introduced an effective 3D scene modeling strategy that integrates scenarios representing the measured data range with additional synthetic scenes generated through random combinations of structural parameters. By incorporating detailed 3D crop structural information into the KGCNN network and fine-tuning the model with measured data, the approach significantly enhanced the model's adaptability to varying data distributions across different years and growth stages. This approach thus improved both the accuracy and stability of true LAI retrieval.

Maize is pivotal in supporting global agriculture and addressing food security challenges. Crop root systems are critical for water uptake and nutrient acquisition, which impacts yield. Quantitative trait phenotyping is essential to understand better the genetic factors underpinning maize root growth and development. Root systems are challenging to phenotype given their below-ground, soil-bound nature. In addition, manual trait annotations of root images are tedious and can lead to inaccuracies and inconsistencies between individuals, resulting in data discrepancies. In this study, we explored juvenile root phenotyping in the presence and absence of auxin treatment, a key phytohormone in root development, using manual curation and gene expression analyses. In addition, we developed an automated phenotyping pipeline for field-grown maize crown roots by leveraging open-source software. By examining a test set of 11 diverse maize genotypes for juvenile-adult root trait correlations and gene expression patterns, an inconsistent correlation was observed, underscoring the developmental plasticity prevalent during maize root morphogenesis. Transcripts involved in hormone signaling and stress responses were among differentially expressed genes in roots from 20 diverse maize genotypes, suggesting many molecular processes may underlie the observed phenotypic variance. In particular, co-expressed gene expression networks associated with module-trait relationships included 1,3-β-glucan, which plays a crucial role in cell wall dynamics. This study furthers our understanding of genotype-phenotype relationships, which is relevant for informing agricultural strategies to improve maize root physiology.

In quantitative genomic analysis of wheat plant height (PH), the average height of a few representative plants is typically used to represent the PH of the entire plot, which overlooks the variation in height among other plants. Extracting different height quantiles from canopy point clouds can address this limitation. For this purpose, low-cost UAV cross-circling oblique (CCO) imaging, combined with structure-from-motion (SfM) and multi-view stereopsis (MVS), was employed to generate precise canopy point clouds for 262 F5 recombinant inbred lines (Zhongmai 578 ​× ​Jimai 22) across seven environments. Multi-level 3D-PH measurements were extracted from six height quantiles, revealing a strong correlation (mean r ​= ​0.95) between 3D-PH and field-measured PH (FM-PH) across environments. The 90 ​% and 92 ​% height quantiles showed the closest agreement with FM-PH compared to other quantiles. Eleven stable quantitative trait loci (QTLs) associated with multi-level 3D-PH were identified using a 50K single nucleotide polymorphism array. Among these, QPhzj.caas-3A.2 (detected by 3D-PH) and QPhzj.caas-7A.1 (detected by both FM-PH and 3D-PH) represented potential novel loci. KASP markers for these QTLs were developed and validated. Furthermore, within the intervals of QPhzj.caas-5A and QPhzj.caas-3B (both were detected by 3D-PH), two candidate genes associated with PH regulation were identified: TaGL3-5A and Rht5, respectively. Corresponding KASP markers for these genes were also developed and validated. This study highlighted the advantages of 3D model and multi-level 3D-PH in elucidating the genetic basis of crop height, and provided a precise and objective basis for advancing wheat breeding programs.

We present Depth-Informed Crop Segmentation (DepthCropSeg), an almost unsupervised crop segmentation approach without manual pixel-level annotations. Crop segmentation is a fundamental vision task in agriculture, which benefits a number of downstream applications such as crop growth monitoring and yield estimation. Over the past decade, image-based crop segmentation approaches have shifted from classic color-based paradigms to recent deep learning-based ones. The latter, however, rely heavily on large amounts of data with high-quality manual annotation such that considerable human labor and time are spent. In this work, we leverage Depth Anything V2, a vision foundation model, to produce high-quality pseudo crop masks for training segmentation models. We compile a dataset of 17,199 images from six public plant segmentation sources, generating pseudo masks from depth maps after normalization and thresholding. After a coarse-to-fine manual screening, 1378 images with reliable masks are selected. We compare four semantic segmentation models and enhance the top-performing one with depth-informed two-stage self-training and depth-informed post-processing. To evaluate the feasibility and robustness of DepthCropSeg, we benchmark the segmentation performance on 10 public crop segmentation testing sets and a self-collect dataset covering in-field, laboratory, and unmanned aerial vehicle (UAV) scenarios. Experimental results show that our DepthCropSeg approach can achieve crop segmentation performance comparable to the fully supervised model trained with manually annotated data (86.91 vs. 87.10). For the first time, we demonstrate almost unsupervised, close-to-full-supervision crop segmentation successfully.

Pear fruit typically contains abundant highly lignified cells, known as stone cells, which have a negative impact on the fruit's edibility and processing quality. Despite extensive physiological and molecular research, there remains a limited understanding of the precise spatiotemporal aspects of lignification in flesh cells during pear development, particularly regarding the initiation of lignification and expansion of stone cell clusters. Here, an emerging bioorthogonal chemistry-based imaging technique was employed to in vivo visualize cell lignification dynamics in developing pear fruit. Specific identification of active sites undergoing lignification revealed that initial lignification of flesh cells occurred at 10 days after full bloom (DAFB), resulting in the formation of primordial stone cells (PSCs). These PSCs exhibited a random distribution and showed significantly larger diameter and area compared to normal parenchyma cells. Subsequently, PSCs developed pit canals and initiated lignification process in their neighboring cells at 15 DAFB. A cascading effect in the formation of stone cell aggregations was visualized by tracing of the lignification trajectory. This expansion process exhibited a domino effect, whereby lignification progressively spread from one cell to the next, creating a cascading pattern of stone cell formation. Finally, a cellular developmental model was proposed for stone cell formation. This study presented a procedure for applying the cutting-edge technology, click chemistry imaging, to get insights into practical scientific questions. The findings elucidated the spatiotemporal dynamics of active lignification sites in pear fruit at the cellular level, thereby enhancing our understanding of the initiation and aggregation processes in stone cell formation.

Genotype-environment interaction (G ​× ​E) models have potential in digital breeding and crop phenotype prediction. Using genotype-specific parameters (GSPs) as a bridge, crop growth models can capture G ​× ​E and simulate plant growth and development processes. In this study, a dataset containing multi-environmental planting and flowering data for 169 genotypes, each with 700K single nucleotide polymorphism (SNP) markers was used. Three rice growth models (ORYZA, CERES-Rice, and RiceGrow), SNPs, and climatic indices were integrated for flowering time prediction. Significant associations between GSPs and quantitative trait nucleotides (QTNs) were investigated using genome-wide association study (GWAS) methods. Several GSPs were associated with previously reported rice flowering genes, including DTH2, DTH3 and OsCOL15, demonstrating the genetic interpretability of the models. The rice models driven by SNPs-based GSPs showed a decrease in goodness of fit as reflected by increased root mean square errors (RMSE), compared to the traditional model calibration. The predictions of crop model were further modified using the machine learning (ML) methods and climate indicators. The accuracy of the modified predictions were comparable to what was achieved using the traditional calibration approach. In addition, the Multi-model ensemble (MME) was comparable to that of the best individual model. Implications of our findings can potentially facilitate molecular breeding and phenotypic prediction of rice.

The threat of crop diseases can lead to reduced yields and significantly hinder progress towards achieving the sustainable development goal of "zero hunger." In the process of detecting crop diseases, variations in data collection conditions can lead to significant differences in the spatial distribution features of training and testing data. Models trained on specific datasets often perform poorly when applied to detect crop diseases in new datasets, significantly affecting the performance of cross-domain object detection tasks. To address the challenges of cross-domain crop disease object detection, this paper proposes a Multi-Granularity Alignment (MGA) domain adaptation framework that is compatible with other object detectors and is generalizable. This approach integrates the multi-granularity alignment and omni-scale gated fusion domain adaptation components into an enhanced object detector, aiming to align features between the source and target domains and reduce their disparities. MGA conducts scale-aware convolutional aggregation on the feature maps of the object detector, and it utilizes three different levels of discriminators-category, instance, and pixel-to identify the domain source, aligning features across different domains from a granularity-dependent perspective, thereby achieving cross-domain object detection. The experimental results demonstrate that MGA achieves the highest mAP in various datasets collected from different regions, environments, and styles, with scores of 47.9 ​% (dataset: PVi → CDi), 48.3 ​% (dataset: PDc → PVi), and 49.2 ​% (dataset: Data with style transfer → CDi). This performance significantly surpasses other object detection technologies. When integrated into Faster R-CNN, MGA also achieves a remarkable mAP of 44.7 ​% on the CDi → Data w/style transfer dataset, demonstrating robust generalization capabilities.

Automatic plant growth monitoring is an important task in modern agriculture for maintaining high crop yield and boosting the breeding procedure. The advancement of 3D sensing technology has made 3D point clouds to be a better data form on presenting plant growth than images, as the new organs are easier identified in 3D space and the occluded organs in 2D can also be conveniently separated in 3D. Despite the attractive characteristics, analysis on 3D data can be quite challenging. We present 3D-NOD, a framework to detect new organs from time-series 3D plant data by spatiotemporal point cloud deep semantic segmentation. The design of 3D-NOD framework drew inspiration from how a well-experienced human utilizes spatiotemporal information to identify growing buds from a plant at two different growth stages. In the training phase, by introducing the Backward & Forward Labeling, the Registration & Mix-up, and the Humanoid Data Augmentation step, our backbone network can be trained to recognize growth events with organ correlation from both temporal and spatial domains. In testing, 3D-NOD has shown better sensitivity at segmenting new organs against the conventional way of using a network to conduct direct semantic segmentation. On a time-series dataset containing multiple species, Our method reached a mean F1-measure at 88.13 ​% and a mean IoU at 80.68 ​% on detecting both new and old organs with the DGCNN backbone.