# Detecting Volunteer Cotton Plants in a Corn Field with Deep Learning on UAV Remote-Sensing Imagery

**Authors:** Pappu Kumar Yadav, J. Alex Thomasson, Robert Hardin, Stephen W. Searcy, Ulisses Braga-Neto, Sorin C. Popescu, Daniel E. Martin, Roberto Rodriguez, Karem Meza, Juan Enciso, Jorge Solorzano Diaz, Tianyi Wang  
**arXiv:** [2207.06673](https://arxiv.org/abs/2207.06673)  
**Date:** 2022-07-14  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2207.06673v1/

## Key points

- The cotton boll weevil has caused over 16 billion USD in damages to the U.S. cotton industry since its introduction from Mexico in the late 1800s.
- Efforts to eradicate the pest have been made, but infestations still occur in southern Texas due to its sub-tropical climate.
- Volunteer cotton (VC) plants that grow alongside inter-seasonal crops like corn can serve as hosts for the pests.
- A recent study used YOLOv3 on aerial images collected by unmanned aircraft systems (UAS) to detect VC plants in a corn field.
- The study found no significant differences in mean average precision (mAP) among three different scales of images.
- However, there were significant differences in average precision (AP) and F1-score between certain image scales.
- The trained YOLOv3 model can be used for near real-time detection and spray application of VC plants using a computer vision-based remotely piloted aerial application system (RPAAS).
- This study provides valuable insights into improving pest management strategies in the U.S. cotton industry.

## Summary

The cotton boll weevil, Anthonomus grandis Boheman, is a destructive pest that has caused over 16 billion USD in damages to the U.S. cotton industry since its introduction from Mexico in the late 1800s. While efforts have been made to eradicate this pest, the southern part of Texas still faces infestations due to its sub-tropical climate which allows cotton plants to grow year-round. Volunteer cotton (VC) plants that grow alongside inter-seasonal crops like corn can serve as hosts for these pests once they reach the pin-head square stage. Therefore, it is crucial to detect and eliminate or spray these VC plants. In a recent study, researchers presented a method for detecting VC plants in a corn field using YOLOv3 on aerial images collected by unmanned aircraft systems (UAS). The study aimed to determine if YOLOv3 could effectively detect VC plants using RGB aerial images and investigate its behavior at different scales. The results showed no significant differences in mean average precision (mAP) among three different scales (320 x 320 pixels; 416 x 416 pixels; and 512 x 512 pixels). However, there were significant differences in average precision (AP) between 320 x 320 pixels and 512 x 512 pixels as well as between 416 x 416 pixels and 512 x 512 pixels. Additionally, a significant difference was found for the F1-score between 416 x 416 pixels and 512 x 512 pixels. These findings suggest that the trained YOLOv3 model can be used on a computer vision-based remotely piloted aerial application system (RPAAS) for near real-time detection and spray application of VC plants. The study provides valuable insights into improving pest management strategies in the U.S. cotton industry by effectively identifying and eliminating potential hosts for the cotton boll weevil.