成熟度
成熟
RGB颜色模型
线性判别分析
数学
园艺
人工智能
模式识别(心理学)
色调
质量(理念)
质量评定
索引(排版)
图像处理
表征(材料科学)
植物
集合(抽象数据类型)
图像(数学)
计算机视觉
作者
Coral Ortiz,Nikita Dapurkar,Vicente Alegre,Francisco Rovira-Más
出处
期刊:Sensors
[Multidisciplinary Digital Publishing Institute]
日期:2026-02-09
卷期号:26 (4): 1113-1113
被引量:2
摘要
The increasing demand for high-quality dragon fruit in the European market requires efficient quality assessment methods. This study explores a non-destructive image analysis approach for classifying ripe dragon fruits based on fruit ripeness and weight. A low-cost system equipped with visible and ultraviolet lighting was employed to capture images of two sets of samples of 60 and 92 ripe dragon fruits, extracting non-destructive parameters such as visible and ultraviolet perimeter, maximum and minimum diameter and area, and RGB color coordinates. Fruit destructive characterization parameters were also measured. The first set of samples was used to develop a discriminant classification model. In a first step, the main characterization magnitudes were confirmed. A ripening index was calculated based on soluble solid content and acidity. Then, a cluster analysis was used to segregate the fruits into three quality characteristics based on the ripening index and weight. In a second step, a step-by-step discriminant analysis was conducted to classify the fruits into the three quality categories (based on the laboratory-measured weight, soluble solid content and total acidity) using the non-destructive magnitudes extracted from the image analysis. The proposed classification system achieved an accuracy of nearly 85% of well-classified dragon fruits, effectively segregating dragon fruits into the three established categories. Furthermore, the established model could select the very high-quality dragon fruit (riper and larger fruits) with 93% of correctly identified products. A comparable procedure was subsequently applied to the additional set of samples (set 2), obtaining consistent results and confirming that image analysis magnitudes related to size and color enable fruit classification into the predefined weight- and ripeness-based categories. Compared to conventional destructive methods, this non-destructive approach offers a promising, cost-effective, and reliable solution for quality assessment. The findings highlight the potential for integrating smart technologies into fruit classification processes, during automatic harvest and postharvest operations, ultimately improving efficiency, reducing labor costs, and enhancing product consistency in the dragon fruit industry.
科研通智能强力驱动
Strongly Powered by AbleSci AI