Deep Learning Techniques for In-Crop Weed Identification: A Review

Kun Hu,Zhiyong Wang,Guy Coleman,Asher Bender,Tingting Yao,Shan Zeng,Dezhen Song,Arnold Schumann,Michael Walsh

DOI: https://doi.org/10.1007/s11119-023-10073-1

2024-02-05

Abstract:Weeds are a significant threat to the agricultural productivity and the environment. The increasing demand for sustainable agriculture has driven innovations in accurate weed control technologies aimed at reducing the reliance on herbicides. With the great success of deep learning in various vision tasks, many promising image-based weed detection algorithms have been developed. This paper reviews recent developments of deep learning techniques in the field of image-based weed detection. The review begins with an introduction to the fundamentals of deep learning related to weed detection. Next, recent progresses on deep weed detection are reviewed with the discussion of the research materials including public weed datasets. Finally, the challenges of developing practically deployable weed detection methods are summarized, together with the discussions of the opportunities for future research.We hope that this review will provide a timely survey of the field and attract more researchers to address this inter-disciplinary research problem.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper attempts to address the problem of achieving effective weed identification within large-scale grain production systems. Specifically, weeds pose a significant threat to agricultural productivity and the environment, while traditional weed control methods (such as the use of herbicides) have many limitations. Therefore, researchers are committed to developing alternative, sustainable weed control technologies that require precise identification of weeds without harming crops. However, existing weed identification algorithms perform poorly in complex and variable farmland environments, especially when the morphological differences between weeds and crops are small under different growth stages and environmental conditions.

To address these issues, the paper reviews recent advances in the application of deep learning-based image recognition technology in the field of weed detection. The main content includes:

1. **Fundamentals of Deep Learning**: Introduces the basic concepts and technologies of deep learning related to weed detection.
2. **Recent Advances**: Discusses the latest deep learning weed detection algorithms and their research results, including publicly available weed datasets.
3. **Challenges and Opportunities**: Summarizes the challenges faced in developing practically usable weed detection methods and explores future research opportunities.

The paper aims to provide researchers with a timely overview of the field by reviewing these contents and to attract more attention to this interdisciplinary research issue.