Wheat spike morphology plays a critical role in determining grain yield and has garnered significant interest in genetics and breeding research. However, traditional measurement methods are limited to simple traits and fail to capture complex spike phenotypes with high precision, thus limiting progress in yield-related trait analysis. In this study, a deep learning pipeline, called Speakerphone, for acquiring precise wheat spike phenotypes was developed. Our pipeline achieved a mean intersection over union (mIoU) of 0.948 in spike segmentation. Additionally, the spike traits measured by our method strongly agreed with the manually measured values, with Pearson correlation coefficients of 0.9865 for spike length, 0.9753 for the number of spikelets per spike, and 0.9635 for fertile spikelets. Using experimental data of 221 wheat cultivars from various regions of Zhao County, Hebei Province, China, our pipeline extracted 45 phenotypes and analyzed their correlations with thousand-grain weight (TGW) and spike yield. Our findings indicate that precise measurements of spike area, spikelet area, and other phenotypic traits clarify the correlation between spike morphology and wheat yield. Through hierarchical clustering on the basis of spike morphology, we categorized wheat spikes into six classes and identified the phenotypic differences among these classes and their effects on TGW and yield. Furthermore, phenotypic differences among wheat cultivars from different geographical regions and over decades were revealed in this study, with an increase in the number of large-spike cultivars over time, especially in southern China. This research may help breeders understand the relationship between wheat spike morphology and yield, thus providing an important basis for future wheat breeding efforts.

Global climate change and urbanization have posed challenges to sustainable food production and resource management in agriculture. Vertical farming, in particular, allows for high-density cultivation on limited land but requires precise control of crop height to suit vertical farming systems. Tomato, a globally significant vegetable crop, urgently requires mutant varieties that suppress indeterminate growth for effective cultivation in vertical farming systems. In this study, we utilized the CRISPR-Cas9 system to develop a new tomato cultivar optimized for vertical farming by editing the Gibberellin 20-oxidase (SlGA20ox) genes, which are well known for their roles in the “Green Revolution”. Additionally, we proposed a volumetric model to effectively identify mutants through non-destructive analysis of chlorophyll fluorescence. The proposed model achieved over 84 ​% classification accuracy in distinguishing triple-determinate and slga20ox gene-edited plants, outperforming traditional machine learning methods and 1D-CNN approaches. Unlike previous studies that primarily relied on manual feature extraction from chlorophyll fluorescence data, this research introduced a deep learning framework capable of automating feature extraction in three dimensions while learning the temporal characteristics of chlorophyll fluorescence imaging data. The study demonstrated the potential to classify tomato plants customized for vertical farming, leveraging advanced phenotypic analysis methods. Our approach explores new analytical methods for chlorophyll fluorescence imaging data within AI-based phenotyping and can be extended to other crops and traits, accelerating breeding programs and enhancing the efficiency of genetic resource management.

Reliable and automated three-dimensional segmentation of plant organs is essential for extracting phenotypic traits at the organ level. However, existing methods for plant organ segmentation predominantly rely on fully supervised learning, which still necessitates extensive point-by-point annotated datasets and fails to overcome the challenges associated with annotating plant point cloud data. In recent years, self-supervised learning-based point cloud segmentation methods have garnered widespread attention in both industry and academia because of their potential to alleviate the difficulties of point cloud data annotation to some extent. In this study, the paradigm of self-supervised learning is innovatively applied to the field of plant phenotyping through the development of the Plant-MAE, a self-supervised learning-based point cloud segmentation framework. The innovations of the Plant-MAE include a kernel-based point convolution embedding module and a multiangle feature extraction block (MAFEB) based on attention mechanisms. To validate the effectiveness of the model, extensive experiments were conducted on multiple point cloud datasets, which achieved competitive performance, with average precision, recall, F1 score, and IoU values of 92.08 ​%, 88.50 ​%, 89.80 ​%, and 84.03 ​%, respectively. The Plant-MAE outperforms advanced deep learning networks, including PointNet++, point transformer, and Point-M2AE, achieving average improvements of at least 0.53 ​%, 1.36 ​%, 0.88 ​%, and 2.38 ​% in precision, recall, F1 score, and IoU, respectively. Additionally, on the Pheno4D dataset, only half of the training data were necessary for fine-tuning to achieve performance comparable to that of the point transformer and PointNet++. This study provides technical support for the estimation of crop phenotypic parameters, thereby advancing the development of modern smart agriculture.

Canopy photosynthetic productivity is crucial for the formation of crop yields. Identifying limiting factors and adjustment targets for canopy photosynthesis in specific climates is important for yield increase. However, conducting relevant quantitative research remains challenging. In this study, two typical regions with distinct climatic characteristics were selected for a two-year trial of greenhouse tomatoes grown in different seasons. A three-dimensional canopy photosynthesis model was developed to quantify the factor contributions to the regional differences in accumulated canopy photosynthesis throughout the entire growing season (ACP), and to predict gains in ACP through three scenarios: leaf photosynthetic modifications (S1), plant layout adjustments (S2), and greenhouse film haze increase (S3). The results indicated that differences in ACP were mainly influenced by light environment (LE), leaf photosynthetic physiology (PP), and LE-PP interaction in spring, and canopy structure (CS), PP, LE, and LE-PP interaction in autumn. The predicted ACP enhancement showed as S1 ​> ​S2 ​> ​S3, with S3 showing a more limited effect. The light quantum efficiency under limiting light () and maximum electron transport rate () were identified as key biochemical phenotypes for tomato high photosynthetic efficiency breeding in different environments. Additionally, adjusting row spacing under current planting density could further improve ACP. Our conclusions could assist researchers in deepening their understanding of canopy photosynthesis limitations under real production conditions, and provide a theoretical foundation for optimizing greenhouse tomato yield in the context of climate change.

Canopy nitrogen content (CNC) and canopy phosphorus content (CPC) of vegetation in wetlands are key physiological traits, which can be associated with the process of wetland ecosystems. Because of the spectral signals obscured by pigments and water content, it is challenging to accurately estimate CNC and CPC of vegetation species in wetlands using multispectral images. Therefore, we developed the constrained PROSAIL-PRO spectra matching (CPSM) approach to extend multispectral reflectance of unmanned aerial vehicle measurements to 400 ∼ 2500 ​nm. We verified the matched accuracy and spectral reliability of CPSM's spectra from two aspects of reflectance and vegetation spectral characteristic based on field-measured spectral data. We proposed a novel hybrid retrieval strategy to achieve the high-precision estimation of CNC and CPC for seven karst wetland vegetation species. Finally, we evaluated the applicability of combining CPSM with our strategy to estimate CNC and CPC for two typical species in mangrove wetlands. Our results proved that CPSM-based spectra had good consistency with original reflectance of UAV images (R2 ​= ​0.82 ∼ 0.86), and they could maintain similar spectral characteristics to measured spectra. Besides, this study found that the optimal spectral features of CNC and CPC were distributed near the red edge position and water-absorption valley of vegetation spectra. We obtained high-precision estimation of CNC and CPC in karst wetland using CPSM and our hybrid retrieval strategy (R2 ​= ​0.60 ∼ 0.98, MRE ​= ​5.91 ​% ​∼ ​26.25 ​%). The approach also showed a better transferring performance in estimating CNC and CPC of mangrove species (R2 ​= ​0.77 ∼ 0.89, MRE ​= ​9.65 ​% ​∼ ​16.87 ​%). The CPSM approach is effective to achieve high-precision estimation of vegetation CNC and CPC.

Freeze injury during the seedling stage significantly impacts wheat growth and yield, making the development of freeze-tolerant varieties crucial for ensuring stable yields. To identify key genetic factors for wheat freeze tolerance, an accurate assessment of freeze tolerance is necessary. However, traditional methods, such as visual inspection, are subjective and can vary significantly among observers. In this study, we developed FreezeNet, a lightweight deep learning model designed to accurately quantify freeze injury using an image-based phenotyping method. Freeze tolerance traits, including vegetation area (VA), green vegetation area (GVA), yellow vegetation fraction (YVF), and mean hue value (mHue), were extracted for freeze tolerance assessment. We captured standardized images with a smartphone and used FreezeNet to extract the freeze tolerance traits for 220 wheat accessions. These traits were strongly correlated with traditional injury scores estimated through visual inspection. Moreover, they presented relatively high heritability. Using these traits, we conducted genome-wide association studies (GWASs) to identify genetic loci associated with freeze tolerance. Eleven significant QTLs associated with freeze tolerance were identified, including 8 novel loci. By integrating four of these loci into a wheat germplasm that lacked any of the 11 QTLs, we significantly enhanced its freeze resistance, demonstrating the practical application of these genetic loci in breeding for improved freeze tolerance. Our results highlight FreezeNet as an advanced tool for assessing wheat freeze injury and identifying the genetic factors responsible for freeze tolerance, with the potential to guide breeding efforts toward the development of more resilient wheat varieties.

Unmanned aerial vehicle (UAV)-based multispectral imaging is one of the most widely used technologies for rapid crop monitoring, essential for crop-growth management. However, the technology's complex optical structure and difficulty in interpreting real-time crop-growth information seriously restrict its application. This paper presents a newly designed UAV-based snapshot multispectral imaging crop-growth sensor (SMICGS) aimed at simplifying the optical structure and realizing the online interpretation of crop spectral information. Mosaic filters based on the special spectral characteristics of crops were designed to achieve multiband co-optical imaging. A spectral crosstalk correction method based on the pixel response characteristics of SMICGS was proposed, and a processing system based on the coupling of sensor information and crop-growth monitoring models was developed to realize real-time online processing of crop spectral information. Field experiments showed that the vegetation indices obtained by SMICGS combined with the machine learning algorithm random forest (RF) achieved better results in predicting leaf area index (LAI) and above-ground biomass (AGB) for wheat and rice. For wheat, the R2 and root mean square error (RMSE) values for the LAI and AGB prediction models were 0.81 and 0.85, and 0.682 and 1.127 ​t/ha, respectively. For rice, the R2 and RMSE values for the LAI and AGB prediction models were 0.89 and 0.93, and 0.818 and 0.866 ​t/ha, respectively. Overall, SMICGS provides a reliable foundational tool for real-time, non-destructive monitoring of field crop growth information, offering significant potential for the precise management of agricultural production.

Capturing crop physiological information by phenotyping is a key trend in smart agriculture. However, current studies underutilize spatial structural information in phenotypic imaging. To evaluate the feasibility of crop cold stress monitoring based on phenotypic spatial variability, we conducted controlled experiments on ‘Toyonoka’ strawberry plants under four dynamic cooling gradients and three stress durations and analyzed the dependence of their photosynthetic physiology and phenotypic traits on temperature-time interactions. The results revealed that NPQ/1D-Parallel/TENT, Y(NO)/2D-Region/INEM, and qP/1D-Parallel/TENT presented the highest mutual information, with the maximum net photosynthetic rate (Pmax), relative electrolyte conductivity (REC), and total chlorophyll content (Chla ​+ ​b), respectively. The difference between the Photosynthetic Physiological Potential Index (PPPI) and relative negative accumulated temperature (RNAT)/650 effectively was used to calculate the cold damage risk (CDRI). An XGBoost-based model integrating the PPPI and RNAT outperformed AdaBoost and RandomForest, achieving an R2 of 0.98, an RMSE of 0.337, a classification accuracy of 92.13 ​%, and a Kappa coefficient of 0.904. qP/1D-Parallel/TENT contributed the most to the model. This study provides a scientific basis for phenotypic information mining and agro-meteorological disaster monitoring.

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.