Optimizing Intra-Row Weed Management in Apple Orchards: Enhancing Detection and Localization Through Advanced Machine Vision Techniques
Research Poster Engineering 2025 Graduate ExhibitionPresentation by Lawrence Arthur
Exhibition Number 176
Abstract
Intra-row weed management in apple orchards is essential for establishing a pest-free environment and maintaining high fruit quality, particularly from the blossoming period through mid-July. Historically, chemical applications have been the predominant method for controlling weeds; however, excessive reliance on these methods and blanket herbicide applications can lead to significant challenges. These challenges include the development of herbicide-resistant weed populations and adverse effects on the surrounding environment, including soil and water contamination. To address these issues, advanced spot spraying systems equipped with machine vision technology have emerged, revolutionizing chemical weed management through what is known as the "see and spray" concept. Despite these advancements, the effectiveness of machine vision can be compromised by variable light conditions and differing crop arrangements, which may hinder the system's ability to detect and localize weeds accurately. This study aims to optimize intra-row weed management in apple orchards by enhancing machine vision capabilities for more precise detection and localization of weeds. The specific goals of this research include improving the accuracy of weed detection and refining the tracking of identified weeds. The methodology leverages the YOLOv7 segmentation model, integrating dynamic line positioning to enhance localization and implementing an improved tracking system for accurate object recognition. Importantly, depth information is utilized to tackle overlapping detection issues. Initial results indicate that the algorithm significantly boosts weed detection accuracy in apple orchards to 91.3%. The effective combination of line positioning and tracking mechanisms has enabled continuous recognition of weeds, substantially reducing missed detections by 96%.
Importance
This study is significant for apple orchard management as it addresses the critical challenge of weed control while minimizing reliance on chemicals. Weeds can compete with apple trees for nutrients and resources, affecting both fruit quality and yield. By enhancing machine vision technology, the research offers a more precise way to identify and locate weeds, which means that farmers can target their efforts more effectively, reducing the overall amount of herbicide used. This not only helps in preventing herbicide resistance and environmental damage but also promotes healthier crops. With improved detection and tracking methods, orchardists can achieve better weed management, leading to higher quality apples and more sustainable farming practices.