Rapeseed Stand Count Estimation at Leaf Development Stages With UAV Imagery and Convolutional Neural Networks

Abstract

Rapeseed is an important oil crop in China. Timely estimation of rapeseed stand count at early growth stages provides useful information for precision fertilization, irrigation, and yield prediction. Based on the nature of rapeseed, the number of tillering leaves is strongly related to its growth stages. However, no field study has been reported on estimating rapeseed stand count by the number of leaves recognized with convolutional neural networks (CNNs) in unmanned aerial vehicle (UAV) imagery. The objectives of this study were to provide a case for rapeseed stand counting with reference to the existing knowledge of the number of leaves per plant and to determine the optimal timing for counting after rapeseed emergence at leaf development stages with one to seven leaves. A CNN model was developed to recognize leaves in UAV-based imagery, and rapeseed stand count was estimated with the number of recognized leaves. The performance of leaf detection was compared using sample sizes of 16, 24, 32, 40, and 48 pixels. Leaf overcounting occurred when a leaf was much bigger than others as this bigger leaf was recognized as several smaller leaves. Results showed CNN-based leaf count achieved the best performance at the four-to six-leaf stage with F-scores greater than 90% after calibration with overcounting rate. On average, 806 out of 812 plants were correctly estimated on 53 days after planting (DAP) at the four-to sixleaf stage, which was considered as the optimal observation timing. For the 32-pixel patch size, root mean square error (RMSE) was 9 plants with relative RMSE (rRMSE) of 2.22% on 53 DAP, while the mean RMSE was 12 with mean rRMSE of 2.89% for all patch sizes. A sample size of 32 pixels was suggested to be optimal accounting for balancing performance and efficiency. The results of this study confirmed that it was

Investigation Of Genetic Diversity Of Different Rapeseed (brassica napus l.) Genotypes And Yield Prediction Using Machine Learning Models

“…Similarly, the validation was carried out by comparing the values predicted by the models with the actual values obtained from the phenotypic and molecular data. To compare the performance of the models in the prediction of hybrid performance, the statistical criteria such as MAE, RMSE and R 2 were used in this research [23]. In order to create and train the neural network in the programming environment, MATLAB 2018b software was used.…”

Automatic counting of rapeseed inflorescences using deep learning method and UAV RGB imagery

“…UAV equipped with RGB cameras (UAV-RGB) has the advantages of higher resolution and less weather affection, which makes it possible to acquire and process large-scale field information conveniently. Recently, combined with the deep learning technology, UAV-RGB system extracted purple leaves ( Zhang et al., 2020a ), recognized frozen ( Li et al., 2022 ), and estimated stand count ( Zhang et al., 2020b ) of rapeseed effectively. Many experts are also attracted to study crop counting.…”

Heading percentage estimation in proso millet (Panicum miliaceum L.) using aerial imagery and deep learning

“…UAVs is one of the most popular HTPPs used to collect and track various traits of crops because of their ease of deployment, low cost, nondestructive, and noninvasive advantages (Sankaran et al., 2015; Shi et al., 2016; Yang et al., 2017). Many phenotyping studies with UAV platforms have been published in rice (J. Wang et al., 2021; J. Wu et al., 2019), wheat (Holman et al., 2016; Li et al., 2019), maize (Buchaillot et al., 2019; Su et al., 2019), soybean (Borra‐Serrano et al., 2020; Trevisan et al., 2020), and other plants (Shafian et al., 2018; J. Zhang et al., 2020). Various phenotypes can be sensed at the canopy level by UAVs including plant height, canopy cover, and spectral reflectance (Galli et al., 2020; Kamilaris & Prenafeta‐Boldú, 2018; Sankaran et al., 2015; Volpato et al., 2021).…”

“…Plant height and leaf rotation angle are derived from high‐resolution red‐green‐blue (RGB) images (Holman et al., 2016; Kawamura et al., 2020; Xu et al., 2021). In addition, high‐resolution RGB cameras can monitor plant growth in seedling, heading, and maturity stages (Das et al., 2019; Madec et al., 2019; J. Zhang et al., 2020). Hence, aerial images captured by UAVs could be a potential data source showing the details of proso millet panicles and tillers for panicle detection and heading percentage estimation in proso millet breeding.…”

“…Supervised algorithms such as Support Vector Machine were used in cereal spike classification (Narisetti et al., 2020). In recent years, deep learning‐based computer vision has been developed and applied to object classification and detection in agriculture, such as weed detection (Dos Santos Ferreira et al., 2017), seedling detection (Pang et al., 2020; J. Zhang et al., 2020), disease recognition (Ashqar & Abu‐Naser, 2018; Kumar et al., 2020), as well as head detection for grains (Alzadjali et al., 2021; Bai et al., 2018; Madec et al., 2019; Narisetti et al., 2020). Hence, this state‐of‐the‐art algorithm is promising to facilitate automated proso millet high‐throughput phenotyping in panicle detection and heading percentage estimation in UAV aerial images.…”