Above-ground biomass (AGB) is a basic agronomic parameter for field investigation and is frequently used to indicate crop growth status, the effects of agricultural management practices, and the ability to sequester carbon above and below ground. The conventional way to obtain AGB is to use destructive sampling methods that require manual harvesting of crops, weighing, and recording, which makes large-area, long-term measurements challenging and time consuming. However, with the diversity of platforms and sensors and the improvements in spatial and spectral resolution, remote sensing is now regarded as the best technical means for monitoring and estimating AGB over large areas. In this study, we used structural and spectral information provided by remote sensing from an unmanned aerial vehicle (UAV) in combination with machine learning to estimate maize biomass. Of the 14 predictor variables, six were selected to create a model by using a recursive feature elimination algorithm. Four machine-learning regression algorithms (multiple linear regression, support vector machine, artificial neural network, and random forest) were evaluated and compared to create a suitable model, following which we tested whether the two sampling methods influence the training model. To estimate the AGB of maize, we propose an improved method for extracting plant height from UAV images and a volumetric indicator (i.e., BIOVP). The results show that (1) the random forest model gave the most balanced results, with low error and a high ratio of the explained variance for both the training set and the test set. (2) BIOVP can retain the largest strength effect on the AGB estimate in four different machine learning models by using importance analysis of predictors. (3) Comparing the plant heights calculated by the three methods with manual ground-based measurements shows that the proposed method increased the ratio of the explained variance and reduced errors. These results lead us to conclude that the combination of machine learning with UAV remote sensing is a promising alternative for estimating AGB. This work suggests that structural and spectral information can be considered simultaneously rather than separately when estimating biophysical crop parameters.

Plant height (PH) is a key factor in maize (Zea maysL.) yield, biomass, and plant architecture. We investigated the PH of diverse maize inbred lines (117 temperate lines, 135 tropical lines) at four growth stages using unmanned aerial vehicle high-throughput phenotypic platforms (UAV-HTPPs). We extracted PH data using an automated pipeline based on crop surface models and orthomosaic model. The correlation between UAV and manually measured PH data reached 0.95. Under temperate field conditions, temperate maize lines grew faster than tropical maize lines at early growth stages, but tropical lines grew faster at later growth stages and ultimately became taller than temperate lines. A genome-wide association study identified 68 unique quantitative trait loci (QTLs) for seven PH-related traits, and 35% of the QTLs coincided with those previously reported to control PH. Generally, different QTLs controlled PH at different growth stages, but eight QTLs simultaneously controlled PH and growth rate at multiple growth stages. Based on gene annotations and expression profiles, we identified candidate genes controlling PH. The PH data collected by the UAV-HTPPs were credible and the genetic mapping power was high. Therefore, UAV-HTPPs have great potential for use in studies on PH.

The number of rice seedlings in the field is one of the main agronomic components for determining rice yield. This counting task, however, is still mainly performed using human vision rather than computer vision and is thus cumbersome and time-consuming. A fast and accurate alternative method of acquiring such data may contribute to monitoring the efficiency of crop management practices, to earlier estimations of rice yield, and as a phenotyping trait in breeding programs. In this paper, we propose an efficient method that uses computer vision to accurately count rice seedlings in a digital image. First, an unmanned aerial vehicle (UAV) equipped with red-green-blue (RGB) cameras was used to acquire field images at the seedling stage. Next, we use a regression network (Basic Network) inspired by a deep fully convolutional neural network to regress the density map and estimate the number of rice seedlings for a given UAV image. Finally, an improved version of the Basic Network, the Combined Network, is also proposed to further improve counting accuracy. To explore the efficacy of the proposed method, a novel rice seedling counting (RSC) dataset was built, which consisted of 40 images (where the number of seedlings varied between 3732 and 16,173) and corresponding manually-dotted annotations. The results demonstrated high average accuracy (higher than 93%) between counts according to the proposed method and manual (UAV image-based) rice seedling counts, and very good performance, with a high coefficient of determination (R²) (around 0.94). In conclusion, the results indicate that the proposed method is an efficient alternative for large-scale counting of rice seedlings, and offers a new opportunity for yield estimation. The RSC dataset and source code are available online.

Phenotyping under field environmental conditions is often considered as a bottleneck in crop breeding. Unmanned aerial vehicle high throughput phenotypic platform (UAV-HTPP) mounted with multi-sensors offers an efficiency, non-invasive, flexible and low-cost solution in large-scale breeding programs compared to ground investigation, especially where measurements are time-sensitive. This study was conducted at the research station of the Xiao Tangshan National Precision Agriculture Research Center of China. Using the UAV-HTPP, RGB and multispectral images were acquired during four critical growth stages of maize. We present a method of extracting plant height (PH) at the plot scale using UAV-HTPP based on the spatial structure of the maize canopy. The core steps of this method are segmentation and spatial Kriging interpolation based on multiple neighboring maximum pixels from multiple plants in a plot. Then, the relationships between the PH extracted from imagery collected using UAV-HTPP and the ground truth were examined. We developed a semi-automated pipeline for extracting, analyzing and evaluating multiple phenotypic traits: canopy cover (CC), normalized vegetation index (NDVI), PH, average growth rate of plant height (AGRPH), and contribution rate of plant height (CRPH). For these traits, we identify genotypic differences and analyze and evaluate dynamics and development trends during different maize growth stages. Furthermore, we introduce a time series data clustering analysis method into breeding programs as a tool to obtain a novel representative trait: typical curve. We classified and named nine types of typical curves of these traits based on curve morphological features. We found that typical curves can detect differences in the genetic background of traits. For the best results, the recognition rate of an NDVI typical curve is 59%, far less than the 82.3% of the CRPH typical curve. Our study provides evidence that the PH trait is among the most heritable and the NDVI trait is among the most easily affected by the external environment in maize.

Above-ground biomass (AGB) is an important indicator for effectively assessing crop growth and yield and, in addition, is an important ecological indicator for assessing the efficiency with which crops use light and store carbon in ecosystems. However, most existing methods using optical remote sensing to estimate AGB cannot observe structures below the maize canopy, which may lead to poor estimation accuracy. This paper proposes to use the stem-leaf separation strategy integrated with unmanned aerial vehicle LiDAR and multispectral image data to estimate the AGB in maize. First, the correlation matrix was used to screen optimal the LiDAR structural parameters (LSPs) and the spectral vegetation indices (SVIs). According to the screened indicators, the SVIs and the LSPs were subjected to multivariable linear regression (MLR) with the above-ground leaf biomass (AGLB) and above-ground stem biomass (AGSB), respectively. At the same time, all SVIs derived from multispectral data and all LSPs derived from LiDAR data were subjected to partial least squares regression (PLSR) with the AGLB and AGSB, respectively. Finally, the AGB was computed by adding the AGLB and the AGSB, and each was estimated by using the MLR and the PLSR methods, respectively. The results indicate a strong correlation between the estimated and field-observed AGB using the MLR method (R² = 0.82, RMSE = 79.80 g/m², NRMSE = 11.12%) and the PLSR method (R² = 0.86, RMSE = 72.28 g/m², NRMSE = 10.07%). The results indicate that PLSR more accurately estimates AGB than MLR, with R² increasing by 0.04, root mean square error (RMSE) decreasing by 7.52 g/m², and normalized root mean square error (NRMSE) decreasing by 1.05%. In addition, the AGB is more accurately estimated by combining LiDAR with multispectral data than LiDAR and multispectral data alone, with R² increasing by 0.13 and 0.30, respectively, RMSE decreasing by 22.89 and 54.92 g/m², respectively, and NRMSE decreasing by 4.46% and 7.65%, respectively. This study improves the prediction accuracy of AGB and provides a new guideline for monitoring based on the fusion of multispectral and LiDAR data.