CropQuant-Air: an AI-powered system to enable phenotypic analysis of yield- and performance-related traits using wheat canopy imagery collected by low-cost drones

Abstract

As one of the most consumed stable foods around the world, wheat plays a crucial role in ensuring global food security. The ability to quantify key yield components under complex field conditions can help breeders and researchers assess wheat’s yield performance effectively. Nevertheless, it is still challenging to conduct large-scale phenotyping to analyse canopy-level wheat spikes and relevant performance traits, in the field and in an automated manner. Here, we present CropQuant-Air, an AI-powered software system that combines state-of-the-art deep learning (DL) models and image processing algorithms to enable the detection of wheat spikes and phenotypic analysis using wheat canopy images acquired by low-cost drones. The system includes the YOLACT-Plot model for plot segmentation, an optimised YOLOv7 model for quantifying the spike number per m2 (SNpM2) trait, and performance-related trait analysis using spectral and texture features at the canopy level. Besides using our labelled dataset for model training, we also employed the Global Wheat Head Detection dataset to incorporate varietal features into the DL models, facilitating us to perform reliable yield-based analysis from hundreds of varieties selected from main wheat production regions in China. Finally, we employed the SNpM2 and performance traits to develop a yield classification model using the Extreme Gradient Boosting (XGBoost) ensemble and obtained significant positive correlations between the computational analysis results and manual scoring, indicating the reliability of CropQuant-Air. To ensure that our work could reach wider researchers, we created a graphical user interface for CropQuant-Air, so that non-expert users could readily use our work. We believe that our work represents valuable advances in yield-based field phenotyping and phenotypic analysis, providing useful and reliable toolkits to enable breeders, researchers, growers, and farmers to assess crop-yield performance in a cost-effective approach.

Panicle-Cloud: An Open and AI-Powered Cloud Computing Platform for Quantifying Rice Panicles from Drone-Collected Imagery to Enable the Classification of Yield Production in Rice

“…To collect large-scale rice imagery under field conditions, current methods range from satellite remote sensing, light airplane, to unmanned aerial vehicles (UAVs; such as drones) and ground-based phenotyping devices. These phenotyping devices had diverse advantages and disadvantages: (a) Satellite sensing system was long distance and ultrascale, but its imagery was relatively low resolution, making it difficult to capture small organ-level objects such as rice panicles [ 18 ]; (b) light airplanes were normally equipped with hyper- and multispectral image sensors for large-scale field surveillance; still, it was difficult to acquire clear panicle-level objects due to high flight altitudes and speed [ 19 ]; (c) low-altitude UAVs fitted with high-resolution red–green–blue (RGB) cameras have been used in field phenotyping popularly, which were reported for their capabilities to acquire imagery with organ-level resolutions when flying at low altitudes [ 20 ]; (d) ground-based devices were usually used to collect very high-resolution 2-dimensional (2D) or 3D plant images at fixed locations and angles; however, their applications were restricted in scalability because of their mobility in rice paddy fields [ 21 ]. Among the above phenotyping approaches, drone-based phenotyping clearly possesses advantages in flexibility, scalability, and cost effectiveness, which is likely to be useful when collecting rice panicle signals at a large scale.…”

The Dissection of Nitrogen Response Traits Using Drone Phenotyping and Dynamic Phenotypic Analysis to Explore N Responsiveness and Associated Genetic Loci in Wheat

“…In wheat, several case studies have been reported in using N response-related patterns to examine N responsiveness and NUE features, including (a) site-specific N management was enabled by aerial phenotyping to assess plant canopy’s N status to estimate available N content in the soil and thus high NUE genotypes [ 31 ]; (b) 3D point clouds were used to measure canopy structural differences to classify varietal NUE performance [ 32 ]; (c) vegetative indices measured under different N applications were used to determine optimized N requirement [ 33 ]; a range of growth traits such as plant height [ 34 ], leaf area index [ 35 ], leaf greenness [ 36 ], phenology [ 37 ], and spike number [ 38 ] were employed to derive N utilization-related indices such as N-utilization efficiency, NupE, and NHI, facilitating studies on N uptake, N utilization, N responsiveness, and N management. Nevertheless, much research focused on specific traits or proxies measured at limited time points during the season, which overly simplified the dynamic feature of N response and related phenotypic changes [ 22 , 39 ].…”