A novel NIR-image segmentation method for the precise estimation of above-ground biomass in rice crops

Abstract

Traditional methods to measure spatio-temporal variations in biomass rely on a labor-intensive destructive sampling of the crop. In this paper, we present a high-throughput phenotyping approach for the estimation of Above-Ground Biomass Dynamics (AGBD) using an unmanned aerial system. Multispectral imagery was acquired and processed by using the proposed segmentation method called GFKuts, that optimally labels the plot canopy based on a Gaussian mixture model, a Montecarlo based K-means, and a guided image filtering. Accurate plot segmentation results enabled the extraction of several canopy features associated with biomass yield. Machine learning algorithms were trained to estimate the AGBD according to the growth stages of the crop and the physiological response of two rice genotypes under lowland and upland production systems. Results report AGBD estimation correlations with an average of r = 0.95 and R 2 = 0.91 according to the experimental data. We compared our segmentation method against a traditional technique based on clustering. A comprehensive improvement of 13% in the biomass correlation was obtained thanks to the segmentation method proposed herein.

Novel Feature-Extraction Methods for the Estimation of Above-Ground Biomass in Rice Crops

“…Several approaches are used to address the above-ground biomass estimation (AGBE) problem [ 3 , 4 , 5 , 6 , 15 ]. For instance, authors in [ 15 ] conducted a comprehensive survey to identify which VIs were suitable for estimating rice biomass as a function of the growth stages of the crop.…”

“…Rice is an essential grain to ensure global food security [ 1 ], contributing to of food energy requirements worldwide. Major efforts have emerged to develop novel high-throughput methods for enabling precision biomass characterization aimed at improving crop yield, grain quality and crop management [ 2 , 3 , 4 ].…”

“…Monitoring rice biomass at larger crop scales requires remote sensing approaches for precision sampling, mostly based on unmanned aerial vehicle (UAV-driven) multispectral imagery [ 3 , 4 , 5 , 6 ]. For this purpose, Above-Ground Biomass Estimation (AGBE) methods have recently gained significant traction [ 7 , 8 , 9 , 10 ], since machine learning models can be trained by features extracted according to different plant reflectance wavelengths, namely vegetation indices (VIs).…”

“…Most of the AGBE methods reported in the literature require image segmentation algorithms to extract the VI formulas from the corresponding canopy [ 3 , 4 , 5 , 6 , 15 ]. Therefore, the accuracy of the extracted VI-features strictly depends on the quality of the segmentation, where several factors such as image resolution, saturation and radiation noise, play an important role in achieving an accurate correlation between the segmented pixels (VI-features) with the biomass.…”

“…In this paper, we compare two approaches for the extraction of relevant features from multispectral aerial imagery, allowing the estimation of above-ground biomass based on training classical machine learning models. In previous work reported in [ 3 , 4 ], we introduced the aforementioned methods. The former method is called GFKuts, which was primarily used as an image segmentation method to optimally segment the crop canopy.…”