کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Design of a Plant Row Detection Algorithm for Agricultural Images Using Dynamic Stripping and Adaptive Parameters

نویسندگان :

Ali Pahlavan¹ Saeed Khankalantary²

1- K.N.Toosi University of Technology 2- K.N.Toosi University of Technology

کلمات کلیدی :

Plant Row Detection،Dynamic Stripping،Hough Transform،Adaptive Parameter Selection،Kernel Size Selection

چکیده :

Accurate detection and representation of plant rows in agricultural imagery play a pivotal role in applications such as crop monitoring, yield estimation, and automated harvesting. This paper introduces a novel approach that integrates adaptive image processing techniques with the DBSCAN clustering algo- rithm to improve line detection in large-scale agricultural images. The methodology encompasses key steps, including morphologi- cal operations, skeletonization via the Zhang-Suen algorithm, and the dynamic selection of Hough Transform parameters. These techniques are further optimized through adaptive kernel size adjustment, which is guided by pixel density, ensuring enhanced precision across diverse imaging conditions. A distinctive feature of this approach is the implementation of Dynamic Stripping, where the image is segmented into horizontal strips to focus line detection efforts, thus boosting accuracy. The DBSCAN algorithm is then employed to cluster and select representative lines for each plant row, while addressing intersection issues and retaining only the most relevant lines. The proposed method excels in managing local variations, reducing noise, and resolving intersecting lines, making it an effective solution for precise plant row identification. The results demonstrate notable improvements in line detection accuracy, alongside reduced computational complexity, presenting a robust and efficient tool for agricultural imagery analysis