Research Article | Open Access

Volume 2023 |Article ID 0065 | https://doi.org/10.34133/plantphenomics.0065

CountShoots: Automatic Detection and Counting of Slash Pine New Shoots Using UAV Imagery

Xia Hao,¹ Yue Cao,¹ Zhaoxu Zhang,¹ Federico Tomasetto,² Weiqi Yan,³ Cong Xu,⁴ Qifu Luan,⁵ and Yanjie Li ⁵

¹College of Information Science and Engineering, Shandong Agricultural University, No. 61, Daizong Road, Taian 271018, Shandong Province, China
²AgResearch Ltd., Christchurch 8140, New Zealand
³Department of Computer Science, Auckland University of Technology, Auckland 1010, New Zealand
⁴School of Forestry, University of Canterbury, Private Bag 4800, 8041 Christchurch, New Zealand
⁵Research Institute of Subtropical Forestry, Chinese Academy of Forestry, No. 73, Daqiao Road, Fuyang, Hangzhou 311400, Zhejiang Province, China

Abstract

The density of new shoots on pine trees is an important indicator of their growth and photosynthetic capacity. However, traditional methods to monitor new shoot density rely on manual and destructive measurements, which are labor-intensive and have led to fewer studies on new shoot density. Therefore, in this study, we present user-friendly software called CountShoots, which extracts new shoot density in an easy and convenient way using unmanned aerial vehicles based on the YOLOX and Slash Pine Shoot Counting Network (SPSC-net) models. This software mainly consists of 2 steps. Firstly, we deployed a modified YOLOX model to identify the tree species and location from complex RGB background images, which yielded a high recognition accuracy of 99.15% and 95.47%. These results showed that our model produced higher detection accuracy compared to YOLOv5, Efficientnet, and Faster-RCNN models. Secondly, we constructed an SPSC-net. This methodology is based on the CCTrans network, which outperformed DM-Count, CSR-net, and MCNN models, with the lowest mean squared error and mean absolute error results among other models (i.e., 2.18 and 1.47, respectively). To our best knowledge, our work is the first research contribution to identify tree crowns and count new shoots automatically in slash pine. Our research outcome provides a highly efficient and rapid user-interactive pine tree new shoot detection and counting system for tree breeding and genetic use purposes.