Skip to main content
ARS Home » Southeast Area » Byron, Georgia » Fruit and Tree Nut Research » Research » Publications at this Location » Publication #412979
Research Project: Genetic Improvement of Stone Fruit Crops/Cultivars for the Southeastern United States

Location: Fruit and Tree Nut Research

Title: Estimation of Physical Characteristics of Unhealthy Peach Leaves Using a Two-Step Algorithm

Author
item WANG, HAIXIN - Fort Valley State University
item Chen, Chunxian

Submitted to: IEEE International Conference on Computer Vision
Publication Type: Proceedings
Publication Acceptance Date: 3/18/2024
Publication Date: 8/14/2024
Citation: Wang, H., Chen, C. 2024. Estimation of Physical Characteristics of Unhealthy Peach Leaves Using a Two-Step Algorithm. IEEE International Conference on Computer Vision . 1: 14-20. https://doi.org/10.1109/CIPCV61763.2024.00013.
DOI: https://doi.org/10.1109/CIPCV61763.2024.00013

Interpretive Summary: Characteristics of diseased leaves reflect the health status of a plant. In this study, we compared six image processing approaches to identify the best approach to estimate the physical characteristics of diseased peach leaves. The results showed the active contour method yielded the highest performance with a metric score of 0.9535., suggesting it was the most effective and viable method to process diseased peach leaves. The proposed two-step algorithm offers an effective framework for quantifying a leaf’s diseased status by estimating color ratios, allowing for disease progression tracking. The results validate the effectiveness and viability of this approach.

Technical Abstract: Evaluation of the health status of peach leaves is a common practice in both botany and agriculture. The study aims to identify physical characteristics including width, length, area, perimeter, and color ratios to assess the health status of peach trees. The study compares six image segmentation algorithms: K-means clustering, active contour model, region growing algorithm, OTSU’s method, K-nearest neighbors (KNN), and support vector machine (SVM), using a custom metric function incorporating pixel accuracy, sensitivity, precision, Dice coefficient, Jaccard index, and specificity. The active contour method yields the highest performance with a metric score of 0.9535, demonstrating its effectiveness in segmenting the leaf from its background. The proposed two-step algorithm offers an effective framework for quantifying a leaf’s diseased status by estimating color ratios, allowing for disease progression tracking. The results validate the effectiveness and viability of this approach.