Coupling of machine learning methods to improve estimation of ground coverage from unmanned aerial vehicle (UAV) imagery for high-throughput phenotyping of crops

Abstract

Ground coverage (GC) allows monitoring of crop growth and development and is normally estimated as the ratio of vegetation to the total pixels from nadir images captured by visible-spectrum (RGB) cameras. The accuracy of estimated GC can be significantly impacted by the effect of 'mixed pixels', which is related to the spatial resolution of the imagery as determined by flight altitude, camera resolution and crop characteristics (fine vs coarse textures). In this study, a two-step machine learning method was developed to improve the accuracy of GC of wheat (Triticum aestivum L.) estimated from coarse-resolution RGB images captured by an unmanned aerial vehicle (UAV) at higher altitudes. The classification tree-based per-pixel segmentation (PPS) method was first used to segment fine-resolution reference images into vegetation and background pixels. The reference and their segmented images were degraded to the target coarse spatial resolution. These degraded images were then used to generate a training dataset for a regression tree-based model to establish the sub-pixel classification (SPC) method. The newly proposed method (i.e. PPS-SPC) was evaluated with six synthetic and four real UAV image sets (SISs and RISs, respectively) with different spatial resolutions. Overall, the results demonstrated that the PPS-SPC method obtained higher accuracy of GC in both SISs and RISs comparing to PPS method, with root mean squared errors (RMSE) of less than 6% and relative RMSE (RRMSE) of less than 11% for SISs, and RMSE of less than 5% and RRMSE of less than 35% for RISs. The proposed PPS-SPC method can be potentially applied in plant breeding and precision agriculture to balance accuracy requirement and UAV flight height in the limited battery life and operation time.

Estimation of Maize Foliar Temperature and Stomatal Conductance as Indicators of Water Stress Based on Optical and Thermal Imagery Acquired Using an Unmanned Aerial Vehicle (UAV) Platform

“…In recent years, unmanned aerial vehicles (UAVs) have become a popular field phenotyping platform for precision agricultural applications. UAV-based phenotyping is the measurement and analysis of plant characteristics in a far more advanced and detailed manner [23,24,[28][29][30][31][32][33][34][35]. UAVs mounted with very-high-resolution (VHR) cameras offer advanced crop image throughput analytics and are effective in overcoming the limitations of satellite imagery [36,37].…”

A novel approach for surveying flowers as a proxy for bee pollinators using drone images

“…Our results showed that, the higher the spatial resolution of the RGB UAV images for the assessment of flower cover, the higher the accuracy (through the goodness of fit) with the in situ flower cover and with the bee abundance, diversity and richness. This is probably largely the result of the fact that images with higher spatial resolution are better able to differentiate between flowers, grass or soil, offering higher information details and reducing the mixed pixel issue (Hu et al, 2021). Images with coarse spatial resolution result in mixed signal at pixel scale, integrating the spectral signature of various vegetative organisms (e.g.…”

“…In this paper we aim to assess flower cover as a proxy for bee abundance using images from a UAV. The fundamental step in estimating flower cover is distinguishing flower pixels from grass or soil pixels by means of differences in the spectral signatures (Hu et al, 2021). Different types of images can be used for this purpose.…”

“…The spatial resolution of the images also plays an important role in the assessment of flower cover. Generally, fine spatial resolution offers more detailed information and reduces the mixed pixel issue (Hu et al, 2021). On the other hand, fine spatial resolution requires lower flight heights which limits the geographical area covered by the UAV in a single flight (Jin et al, 2017).…”

Comparison of Modelling Strategies to Estimate Phenotypic Values from an Unmanned Aerial Vehicle with Spectral and Temporal Vegetation Indexes

“…Unmanned aerial vehicle (UAV) based remote sensing is flexible in data acquisition regarding diverse sensors as well as spatial and temporal resolutions [7,9]. Given the growing demand for high-throughput phenotyping to support crop breeding, great interest has been expressed in using UAV-based platforms to rapidly and non-destructively collect phenotypic data (e.g., imagery) under field conditions [9][10][11]. UAV-based high-throughput phenotyping usually estimates crop traits through the integration of UAV-derived vegetation indexes (VIs) with statistical models [12].…”