Tzuan Gabriel Tan Hong, Hashim Fazida Hanim, Raj Thinal, Baseri Huddin Aqilah, Sajab Mohd Shaiful
Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia.
Research Centre for Sustainable Process Technology (CESPRO), Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia.
Plants (Basel). 2022 Jul 26;11(15):1936. doi: 10.3390/plants11151936.
The capacity of palm oil production is directly affected by the ripeness of the fresh fruit bunches (FFB) upon harvesting. Conventional harvesting standards rely on rigid harvesting scheduling as well as the number of fruitlets that have loosened from the bunch. Harvesting is usually done every 10 to 14 days, and an FFB is deemed ready to be harvested if there are around 5 to 10 empty sockets on the fruit bunch. Technology aided by imaging techniques relies heavily on the color of the fruit bunch, which is highly dependent on the surrounding light intensities. In this study, Raman spectroscopy is used for ripeness classification of oil palm fruits, based on the molecular assignments extracted from the Raman bands between 1240 cm and 1360 cm. The Raman spectra of 52 oil palm fruit samples which contain the fingerprints of different organic compounds were collected. Signal processing was applied to perform baseline correction and to reduce background noises. Characteristic data of the organic compounds were extracted through deconvolution and curve fitting processes. Subsequently, a correlation study between organic compounds was developed and eight hidden Raman peaks including protein, beta carotene, carotene, lipid, guanine/cytosine, chlorophyll-a, and tryptophan were successfully located. Through ANOVA statistical analysis, a total of six peak intensities from proteins through Amide III (β-sheet), beta-carotene, carotene, lipid, guanine/cytosine, and carotene and one peak location from lipid were found to be significant. An automated oil palm fruit ripeness classification system deployed with artificial neural network (ANN) using the seven signification features showed an overall performance of 97.9% accuracy. An efficient and accurate ripeness classification model which uses seven significant Raman peak features from the correlation analysis between organic compounds was successfully developed.
棕榈油的生产能力直接受到收获时新鲜果串（FFB）成熟度的影响。传统的收获标准依赖于严格的收获计划以及从果串上脱落的小果数量。收获通常每10至14天进行一次，如果果串上有大约5至10个空果柄，则认为FFB已准备好收获。借助成像技术的技术严重依赖于果串的颜色，而果串颜色高度依赖于周围的光照强度。在本研究中，基于从1240厘米至1360厘米之间的拉曼光谱带提取的分子归属，拉曼光谱用于油棕果实成熟度分类。收集了52个油棕果实样品的拉曼光谱，这些光谱包含不同有机化合物的指纹信息。应用信号处理进行基线校正并降低背景噪声。通过去卷积和曲线拟合过程提取有机化合物的特征数据。随后，开展了有机化合物之间的相关性研究，成功定位了包括蛋白质、β-胡萝卜素、胡萝卜素、脂质、鸟嘌呤/胞嘧啶、叶绿素-a和色氨酸在内的八个隐藏拉曼峰。通过方差分析统计分析，发现从蛋白质通过酰胺III（β-折叠）、β-胡萝卜素、胡萝卜素、脂质、鸟嘌呤/胞嘧啶和胡萝卜素得到的总共六个峰强度以及脂质的一个峰位置具有显著性。使用这七个显著特征通过人工神经网络（ANN）部署的自动油棕果实成熟度分类系统显示出97.9%的总体准确率。成功开发了一种高效准确的成熟度分类模型，该模型使用了有机化合物相关性分析中的七个显著拉曼峰特征。
