Maize tassels play a critical role in plant growth and yield. Extensive RGB images obtained using unmanned aerial vehicle (UAV) and the prevalence of deep learning provide a chance to improve the accuracy of detecting maize tassels. We used images from UAV, a mobile phone, and the Maize Tassel Counting dataset (MTC) to test the performance of faster region-based convolutional neural network (Faster R-CNN) with residual neural network (ResNet) and a visual geometry group neural network (VGGNet). The results showed that the ResNet, as the feature extraction network, was better than the VGGNet for detecting maize tassels from UAV images with 600 × 600 resolution. The prediction accuracy ranged from 87.94% to 94.99%. However, the prediction accuracy was less than 87.27% from the UAV images with 5280 × 2970 resolution. We modified the anchor size to [852, 1282, 2562] in the region proposal network according to the width and height of pixel distribution to improve detection accuracy up to 89.96%. The accuracy reached up to 95.95% for mobile phone images. Then, we compared our trained model with TasselNet without training their datasets. The average difference of tassel number was 1.4 between the calculations with 40 images for the two methods. In the future, we could further improve the performance of the models by enlarging datasets and calculating other tassel traits such as the length, width, diameter, perimeter, and the branch number of the maize tassels.

Background and Aims High-throughput phenotyping is a limitation in plant genetics and breeding due to large-scale experiments in the field. Unmanned aerial vehicles (UAVs) can help to extract plant phenotypic traits rapidly and non-destructively with high efficiency. The general aim of this study is to estimate the dynamic plant height and leaf area index (LAI) by nadir and oblique photography with a UAV, and to compare the integrity of the established three-dimensional (3-D) canopy by these two methods. Methods Images were captured by a high-resolution digital RGB camera mounted on a UAV at five stages with nadir and oblique photography, and processed by Agisoft Metashape to generate point clouds, orthomosaic maps and digital surface models. Individual plots were segmented according to their positions in the experimental design layout. The plant height of each inbred line was calculated automatically by a reference ground method. The LAI was calculated by the 3-D voxel method. The reconstructed canopy was sliced into different layers to compare leaf area density obtained from oblique and nadir photography. Key Results Good agreements were found for plant height between nadir photography, oblique photography and manual measurement during the whole growing season. The estimated LAI by oblique photography correlated better with measured LAI (slope = 0.87, R2 = 0.67), compared with that of nadir photography (slope = 0.74, R2 = 0.56). The total number of point clouds obtained by oblique photography was about 2.7–3.1 times than those by nadir photography. Leaf area density calculated by nadir photography was much less than that obtained by oblique photography, especially near the plant base. Conclusions Plant height and LAI can be extracted automatically and efficiently by both photography methods. Oblique photography can provide intensive point clouds and relatively complete canopy information at low cost. The reconstructed 3-D profile of the plant canopy can be easily recognized by oblique photography.

No abstract

No abstract

No abstract