The detection of durian ripeness using thermal imaging is an essential study geared towards improving the current analytical methods which rely heavily on routine analysis and human labour skills. Thermal imaging was investigated in this study in order to evaluate the ripeness of durian based on the relationship of physicochemical properties and thermal image parameters. Thermal images of durians were acquired at three different ripening stages (unripe, ripe, and overripe) and the physicochemical properties of the soluble solids content, pH, firmness, moisture content, and colour changes were determined. Partial least squares (PLS) regression was used to develop quantitative prediction models with R² values greater than 0.94 for all the physicochemical properties of durians. Principal component analysis (PCA) showed successful clustering ability of three different ripeness levels of durians. Linear discriminant analysis (LDA), k-nearest neighbour (kNN), and support vector machine (SVM) were applied for the establishment of the optimal classification modelling algorithms. The SVM classifier gave the overall best performance for the discrimination of durian ripeness with a classification accuracy of 97 %. The feasibility of thermal imaging coupled with multivariate methods demonstrated huge potential for non-destructive evaluation of durian ripeness levels.

使用热成像技术检测榴莲的成熟度是一项重要的研究，旨在改善当前的分析方法，这些方法在很大程度上依赖于常规分析和人工技能。在本研究中对热成像进行了研究，以基于理化性质和热成像参数之间的关系来评估榴莲的成熟度。在三个不同的成熟阶段（未成熟，成熟和成熟）获取榴莲的热图像，并测定可溶性固形物含量，pH，硬度，水分含量和颜色变化的理化性质。使用偏最小二乘（PLS）回归开发具有R ^2的定量预测模型榴莲的所有理化特性值均大于0.94。主成分分析（PCA）显示了榴莲三个不同成熟度水平的成功聚类能力。应用线性判别分析（LDA），k最近邻（kNN）和支持向量机（SVM）来建立最佳分类建模算法。SVM分类器在判别榴莲成熟度方面总体表现最佳，分类精度为97％。热成像与多元方法相结合的可行性证明了榴莲成熟度无损评估的巨大潜力。