Surface defect detection in lime by using computer vision

Limes are one of the key ingredients used in Thailand, both in food and beverages. Due to such high market demand for Thai limes, quality inspection is then very important. While manual sorting by humans can lead to inaccuracy, leveraging on technology could alternatively make the process more accurate. Limes are graded by quality. The factors that affect the lime quality or market price are texture, shape, color, and maturity. This research focuses on detecting defects or diseases on lime surface by using computer vision through data collection from image capturing. The researcher used k-means clustering as a segmentation technique to separate the surface of limes into clusters. Each cluster was classified into a color indication of diseases or health. To evaluate the cluster, we used artificial neural network (ANN) and support vector machine (SVM), in which results showed that ANN has led to better performance than SVM. Defect percentage on a lime surface can be measured by computation of pixels from clustering. Consequently, this system will reduce both cost and time.青柠是泰国餐饮与饮品领域的核心原料之一。鉴于泰国青柠市场需求量极高，其品质检测环节尤为关键。传统人工分拣易产生误差，而借助技术手段则可有效提升检测精度。青柠按照品质等级进行分级，影响其品质与市场售价的因素包括质地、外形、色泽及成熟度。本研究依托计算机视觉技术，通过图像采集获取数据，致力于检测青柠表面的缺陷与病害。研究人员采用K均值聚类（K-means clustering）作为分割技术，将青柠表面划分为多个聚类区域，并将每个区域按照病害或健康状态的色泽特征进行分类。为评估聚类效果，本研究使用了人工神经网络（Artificial Neural Network, ANN）与支持向量机（Support Vector Machine, SVM）两种模型，实验结果显示人工神经网络的性能优于支持向量机。通过计算聚类后的像素占比，可量化青柠表面的缺陷占比。最终，该系统能够有效降低检测成本与时间投入。

搜集汇总

背景与挑战

背景概述

该数据集是一份2022年由泰国法政大学发布的学术论文，专注于利用计算机视觉技术检测柠檬表面缺陷。研究通过图像采集、K-means聚类分割和人工神经网络（ANN）与支持向量机（SVM）分类方法，评估了缺陷检测效果，其中ANN表现更优，旨在提高柠檬质量检测的准确性和效率。