Contents lists available at ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag Establishing a model to predict the single boll weight of cotton in northern Xinjiang by using high resolution UAV remote sensing data Weicheng Xu a,b , Weiguang Yang a,b , Shengde Chen a,b, ⁎ , Changsheng Wu a,b , Pengchao Chen a,b , Yubin Lan a,b, ⁎ a College of Electronic Engineering, South China Agriculture University, Guangzhou 510642, China b National Center for International Collaboration on Precision Agricultural Aviation Pesticide Spraying Technology, Guangzhou 510642, China ARTICLE INFO Keywords: UAV remote sensing Neural networks Multiple linear regression Single boll weight prediction of cotton Cotton boll recognition ABSTRACT Single boll weight is the main factor of cotton yield and a key index used to evaluate the quality of cotton. Predicting the single boll weight in a large area is important for variety selection and yield improvement. A model was established to predict the single boll weight by using the multitemporal high-resolution visible light remote sensing data obtained from UAV. Specifically, remote sensing data were collected for 29 fields in the Changji, Shihezi and Shawan areas in northern Xinjiang during the blooming period and boll opening stage. Five circular areas with a radius of 1 m were selected from each field as the ground investigation area for collection of the cotton boll samples. Fully convolutional networks (FCN) was used to recognize and extract bolls at the boll opening stage in the remote sensing images as a dependent variable of the model. Correlation analysis was carried out by combining VDVI (Visible-band difference vegetation index) at the flowering and boll setting stages, VDVI at boll opening stages, VDVI at boll opening areas (Extracted by FCN) and RGB mean values, then use the least squares linear regression and BP neural networks to model the upper, middle, lower cotton layers and average single boll weight in investigation area. Subsequently, K-fold cross-validation was performed to evaluate the results. The results showed that the results of the least squares linear regression (R 2 = 0.8162) and BP neural networks (R 2 = 0.8170) were nearly equivalent. The percentage of boll opening in the area and VDVI at the flowering and boll setting stages were highly correlated with upper single boll weight. This study proposes a method to realize the large scale prediction of single boll weight, which provided a new idea for cotton yield prediction and breeding screening. 1. 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, respec- tively (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 be- tween 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 mon- itoring (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, flex- ibility, 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 https://doi.org/10.1016/j.compag.2020.105762 Received 22 February 2020; Received in revised form 17 August 2020; Accepted 30 August 2020 ⁎ Corresponding authors at: College of Electronic Engineering, South China Agriculture University, Guangzhou 510642, China. E-mail addresses: shengde-chen@scau.edu.cn (S. Chen), ylan@scau.edu.cn (Y. Lan). Computers and Electronics in Agriculture 179 (2020) 105762 0168-1699/ © 2020 Published by Elsevier B.V. T