Classification of weed using machine learning techniques: a review-challenges, current and future potential techniques

Weed detection and classification are considered one of the most vital tools in identifying and recognizing plants in agricultural fields. Recently, machine learning techniques have been rapidly growing in the precision farming area related to plants, as well as weed detection and classification techniques. In digital agricultural analysis, these techniques have played and will continue to play a vital role in mitigating health, agricultural, and environmental impacts, improving sustainability, and reducing herbicides. Deep learning-based models are employed to solve the more sophisticated agricultural issues using individual CNN networks and hybrid models. Such models showed promising results. This paper highlights the major trends from the particular review of detection and classification approaches for weed plants. This review elaborates on the aspects of using traditional methods and deep learning-based methods to solve weed detection problems. It provides an overview of various methods for weed detection in recent years, analyzes the benefits and limitations of existing machine learning techniques, including deep learning techniques, and introduces several related plant leaves, weed datasets, and weeding machinery. Evaluation of the existing techniques has been compared, taking into account the real-world dataset used, images’ capacity, and shortcomings. Furthermore, this study helps to introduce the promising results and identify critically the remaining challenges in achieving robust weed detection, which could support noteworthy agricultural problems and assist researchers in the future. The significance of this study is to provide the potential techniques for solving illumination, overlapping, and occlusion issues of leafy plants, as well as other plant issues.