Establishing a model to predict the single boll weight of cotton in northern Xinjiang by using high resolution UAV remote sensing data
Introduction
China has the largest cotton production area worldwide, and cotton planting area accounts for 30% of the total cultivated land area. The total cotton production area in China was 6,841,593 ha in 2019 (National bureau of statistics, 2019). Xinjiang is the main production area in China, with the corresponding cotton production area and output accounting for more than 50% and approximately 70% of the national cotton production area and total cotton production, respectively (Yu et al., 2019). The single boll weight is the weight of the seed cotton in a single cotton boll, and it exhibits complex characteristics. Moreover, this weight is positively correlated with the burr weight, number of chambers per boll, seed weight, number of seeds, and fiber weight, and negatively correlated with the number of bolls per plant and lint percentage (Karademir et al. 2009). The clear correlation between the single boll weight and yield is has been verified experimentally (Batool et al. 2010). Nevertheless, if the cotton boll is extremely large, the shell is expected to be thick, leading to an extended growth stage and decreased rates of cotton boll filling and boll opening. Generally, in the process of selection and breeding of cotton in China, the suitable weight of a single boll is 5.0–6.5 g (Yu et al., 2016). The efficient prediction of the single boll weight in a large area can help improve the efficiency of cotton variety breeding, thereby improving the cotton yield and quality.
With the rapid development of remote sensing technology, it is being widely used in agricultural management and agricultural monitoring (He and Mostovoy, 2019, Liu et al., 2019, Wei et al., 2019, Weissteiner et al., 2019). As a type of low altitude remote sensing system, UAVs are being widely used owing to their convenience, flexibility, low cost and capacity for high-resolution imaging (Xu et al., 2019, Campos et al., 2019, Yan et al., 2019). Several researchers have attempted to study cotton by using remote sensing systems based on a UAV platform. For example, a UAV installed with a multispectral camera was used to collect data and calculate the NDVI to predict the final yield of cotton. However, single-parameter modeling is not sufficiently accurate (R2 = 0.47) (Huang et al., 2013), and the introduction of additional dependent variables to the prediction model can help improve the prediction accuracy. By installing a high-resolution digital camera in the UAV to collect RGB images and introducing parameters such as color characteristics and vegetation coverage into the model, a higher prediction accuracy (R2 = 0.97) can be obtained (Zhang et al., 2019a, Zhang et al., 2019b). Moreover, a LiDAR sensor can be installed on the UAV to collect quantitative information regarding the crop canopy and generate a digital surface model. These aspects can facilitate the quantitative modeling of biomass (Liu et al., 2019, Simpson et al., 2016, Wang et al., 2019). Moreover, the canopy height and surface model of the crop canopy can be obtained by using an aerial triangulation algorithm in combination with a high-resolution digital camera. Moreover, because of the positive correlation between plant height and biomass, the canopy digital surface model can also be used to realize the inversion of the plant height (Lati et al., 2013). The information provided by various information sources can be extended by using multisource information fusion technology to fuse the remote sensing data pertaining to different time phases. This aspect can considerably improve the effectiveness of feature extraction and classification (Zhang et al., 2019a, Zhang et al., 2019b), and it plays a very important role in the study of farmland classification, vegetation coverage, land desertification and crop growth, which change over time (Huang et al., 2017, Yi et al., 2019, Yu et al., 2019).
The cotton boll in the remote sensing images is extracted through machine vision, which is of great help in studying the rate of cotton boll opening, defoliant effects and to direct intelligent agricultural machinery (Li et al.,2015). In the past decade, researchers have used machine vision technology to extract cotton boll development data. For example, the cotton boll and its background have been segmented by using random forest classification (Zhang et al.,2011) and an object-oriented method based on near-infrared imaging (Li et al.,2016). The rapid improvement in the operation ability of computers has facilitated the advancement of deep learning technology. At present, convolutional neural networks represent the branch of deep learning with the most research and the best development, and they have been applied to yield excellent results (Rawat and Wang 2017). The fully convolutional network (FCN) used in this study was improved using convolutional neural networks (CNN). The convolutional layer was used to replace the last fully connected layer of the CNN. Subsequently, the deconvolutional layer was used to upsample the featured heat map to restore its size to that of the input image to classify the image at the pixel level and retain the spatial information in the input image (Long et al., 2015, Shelhamer et al., 2017). Some researchers attempted to realize cotton extraction using fully convolutional networks (FCN), which is the foundation of the present study (Li et al., 2017).
In this paper, multisource information fusion technology was used to fuse UAV remote sensing data from different time phases at the pixel level. The fully convolutional neural network was used to achieve pixel-level semantic segmentation of the remote sensing images, and the cotton segments in the images were extracted to eliminate the influence of the soil pixels on the model accuracy. The boll opening area in the images was extracted, and analyzed the correlation among the single boll weight of the upper, middle and lower layers of the cotton plant with VDVI and other parameters. VDVI (Du and Noguchi, 2017, Wang et al., 2015a) is a vegetation index used in the case of visible light remote sensing images that can distinguish vegetation from nonvegetation objects without using the near infrared (NIR) domain. Moreover, the correlation coefficients between different parameters and the single boll weight were determined. The objective was to establish a method that could enable the large scale prediction of single boll weight, in order to provide a novel opportunity for cotton yield estimation and variety selection.
Section snippets
Study region
To ensure that the model had a high applicability and could comprehensively consider the distribution of major cotton producing counties and certain production and construction crops in north Xinjiang (Li et al., 2011), 29 cotton fields with a total area of 2,166,666 m2 were selected in 3 different areas located in Changji, Shihezi and Shawan, Xinjiang Uygur Autonomous Region, the People’s Republic of China. Fig. 1 shows the distribution of the field observation areas.
The cotton area in
Correlation analysis
Linear regression was performed for the 4 parameters described in Section 2.6 (Boll opening pixel percentage, VDVI during the blooming period, VDVI during the boll opening stage and RGB average) considering the single boll weight of the upper, middle and lower layers. A total of 12 regression equations and correlation coefficients were used, as shown in Fig. 11.
Only the “boll opening pixel percentage” and “VDVI during the blooming period” exhibited a strong correlation with the single boll
Discussion
Many studies have reported that for cotton, the single boll weight and planting density are negatively correlated. Gao et al. (2009) found that burr weight and single boll weight gradually decreased with increased planting density. Zhang et al., 2019a, Zhang et al., 2019b demonstrated that a close correlation (R2 = 0.97) existed between the NDVI at the flowering and boll setting stages. The single boll weight is one of the main factors affecting cotton yield (Li et al., 2009). Therefore,
Conclusions
A method based on high-resolution UAV visible light remote sensing images was proposed to estimate the upper single boll weight in north Xinjiang. First, the cotton field data pertaining to the flowering and boll setting stages and the boll opening stage by using the UAV. Second, the errors between remote sensing data collected in 2 periods was eliminated by polynomial correction after splicing. Third, the modeling parameters was obtained by performing a correlation analysis with the single
CRediT authorship contribution statement
Weicheng Xu: Conceptualization, Validation, Formal analysis, Investigation, Software, Writing - original draft. Weiguang Yang: Investigation, Resources, Data curation. Shengde Chen: Investigation, Resources, Data curation. Changsheng Wu: Writing - review & editing, Funding acquisition. Pengchao Chen: Investigation, Data curation. Yubin Lan: Conceptualization, Investigation, Resources, Writing - review & editing, Supervision, Project administration, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This research was supported by the Leading Talents of Guangdong Province Program (2016LJ06G689), Science and Technology Planning Project of Guangdong Province (2019B020208007), China Agriculture Research System (CARS-15-22), The 111 Project (D18019) and Science and Technology Planning Project of Guangzhou (201807010039).
References (45)
- et al.
LAI and chlorophyll estimation for a heterogeneous grassland using hyperspectral measurements
ISPRS Journal Of Photogrammetry And Remote Sensing
(2008) - et al.
Estimating plant growth parameters using an energy minimization-based stereovision model
Computers And Electronics In Agriculture
(2013) - et al.
In-field cotton detection via region-based semantic image segmentation
Computers And Electronics In Agriculture
(2016) - et al.
Crop Yield Estimation Using Time-Series MODIS Data and the Effects of Cropland Masks in Ontario
(2019) - et al.
Enhancement of radar rainfall estimates for urban hydrology through optical flow temporal interpolation and Bayesian gauge-based adjustment
J. Hydrol.
(2015) - et al.
Improving the estimation of fractional vegetation cover from UAV RGB imagery by colour unmixing
ISPRS Journal 0f Photogrammetry and Remote Sensing
(2019) - et al.
A fast segmentation method for high-resolution color images of foreign fibers in cotton
Computers And Electronics In Agriculture
(2011) - et al.
Image registration methods: a survey
Image and Vision Computing
(2003) - et al.
Low-weight and UAV-based Hyperspectral Full-frame Cameras for Monitoring Crops: Spectral Comparison with Portable Spectroradiometer Measurements
Photogrammetrie Fernerkundung Geoinformation
(2015) - Batool, S., Khan, N.U., Makhdoom, K., Bibi, Z., Hassan, G., Marwat, K.B., Farhatullah, Mohammad, F., Raziuddin, & Khan,...
No unbiased estimator of the variance of K-fold cross-validation
JOURNAL OF MACHINE LEARNING RESEARCH
Development of canopy vigour maps using UAV for site-specific management during vineyard spraying process
Precis. Agric.
Multi-temporal monitoring of wheat growth by using images from satellite and unmanned aerial vehicle
Int. J. Agric. Biol. Eng.
Monitoring of Wheat Growth Status and Mapping of Wheat Yield's within-Field Spatial Variations Using Color Images Acquired from UAV-camera System
Remote Sensing
Impacts of Boll Weight and Bell Shell Weight and Lint Percentage Cotton During Different Density
Jiangxi Cotton.
Cotton Yield Estimate Using Sentinel-2 Data and an Ecosystem Model over the Southern US
Remote Sensing
Quantificational effect of reforestation to soil erosion in subtropical monsoon regions with acid red soil by sediment fingerprinting
Environmental Earth Sciences
Estimation of cotton yield with varied irrigation and nitrogen treatments using aerial multispectral imagery
Int. J. Agric. Biol. Eng.
Direct Georeferencing of a Pushbroom
Correlations and Path Coefficient Analysis between Leaf Chlorophyll Content, Yield and Yield Components in Cotton (Gossypium hirsutum L.) under Drought Stress Conditions
Notulae Botanicae Horti Agrobotanici Cluj-Napoca
Cited by (15)
Prediction of cotton FPAR and construction of defoliation spraying prescription map based on multi-source UAV images
2024, Computers and Electronics in AgricultureEstimation of direct-seeded guayule cover, crop coefficient, and yield using UAS-based multispectral and RGB data
2022, Agricultural Water ManagementCitation Excerpt :Aerial data through helicopters and airplanes are limited by the cost and unexpected changes in weather conditions. Some of the MS-based platform limitations may be overcome through the use of unmanned aircraft systems (UAS), which have gained the interest of researchers and farmers due to their flexibility and high spatial resolution (Yan et al., 2019; Campos et al., 2019; Xu et al., 2020). UAS carrying multispectral and RGB (red, green, blue) cameras have been indicated to be appropriate for crop growth and yield monitoring (Lu, 2005; Torres-Sanchez et al., 2014; Huang et al., 2016; Kanning et al., 2018; Li et al., 2018a; Hlatshwayo et al., 2019; Yang et al., 2020; Zhang et al., 2021).
Cotton yield estimation model based on machine learning using time series UAV remote sensing data
2021, International Journal of Applied Earth Observation and GeoinformationCitation Excerpt :It uses TensorFlowtm to train the deep learning model and recognize the features in an image according to the spatial and spectral characteristics of the image. U-Net is more refined and efficient than the traditional semantic segmentation model of the FCN, and it has a wide range of applications in remote sensing image classification (Wiratama et al. 2020; Ren et al. 2020; Xu et al. 2020). The main purpose of this research is to establish a cotton production prediction model based on machine learning.
Monitoring of Cotton Boll Opening Rate Based on UAV Multispectral Data
2024, Remote Sensing