Nondestructive detection and grading of flesh translucency in pineapples with visible and near-infrared spectroscopy

Authors: XU, SAI - Guangdong Academy Of Agricultural Sciences; REN, JINCHANG - Robert Gordon University; LU, HUAZHONG - Guangdong Academy Of Agricultural Sciences; WANG, XU - Guangdong Academy Of Agricultural Sciences; Sun, Xiuxiu; LIANG, XIN - Guangdong Academy Of Agricultural Sciences

Submitted to: Postharvest Biology and Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/5/2022
Publication Date: 7/14/2022
Citation: Xu, S., Ren, J., Lu, H., Wang, X., Sun, X.N., Liang, X. 2022. Nondestructive detection and grading of flesh translucency in pineapples with visible and near-infrared spectroscopy. Postharvest Biology and Technology. 192. https://doi.org/10.1016/j.postharvbio.2022.112029.
DOI: https://doi.org/10.1016/j.postharvbio.2022.112029

Interpretive Summary: Pineapple is one of the most economically important crops in tropical and subtropical areas, however, it suffers from flesh translucency. VIS/NIR spectrum was found to be able to transmit the whole pineapple with enough light intensity and quickly acquire abundant information about the internal quality, which is suitable for pineapple translucency detection. In this research, a VIS/NIR spectroscopy based method is developed for the nondestructive detection of pineapple translucency. Comprehensive experiments have validated the efficiency, cost-effectiveness and efficacy of the proposed methodology.

Technical Abstract: The increasing incidence of pineapple translucency and the lack of effective control or treatment for it have increased the industry need for a rapid and nondestructive detection of translucency degree. However, accurate internal quality detection for large fruit is particularly challenging and needs more comprehensive analysis and modeling, especially for pineapples with their highly rough surface and large size. Besides, existing work fails to address the balance between effectiveness and cost and model adaptation among different batches of samples, leading to an inability to meet the needs of industry. To tackle these issues, a visible and near infrared (VIS/NIR) spectrum based platform is proposed in this paper for optimized detection of pineapple translucency. The internal quality of three batches of samples harvested on different dates were acquired by different spectral settings, i.e. visible (Vis) to very near-infrared (NIR) (400 to 1100nm), NIR (900- 1700 nm) and Vis-NIR (400 to 1700 nm). The pineapple samples were manually cut open and divided into three translucency degrees ( no, slight, and heavy) according to the marketing requirements. The Savitzky Golay (SG) and standard normal variate (SNV) were applied to remove the jitter noise and scattering noise, respectively. The successive projections algorithm (SPA), principal component analysis (PCA) and Euclidean distance (ED) were combined for feature extraction; partial least squares regression (PLSR) and probabilistic neural network (PNN) were applied for modeling; data correction, data supplement and the combination of them were applied for model updating. The experimental results showed that the optimal roadmap for pineapple translucency detection is to use 400-1100 nm full wavelength spectrums plus SG, SNV, PNN and data supplement for model updating. With the first and second batch samples used for building and updating the model, the detection accuracy on the third batch samples can achieve 100%. The proposed methodologies therefore can be used as rapid, nondestructive, and cost-effective tools to detect pineapple translucency to guarantee the marketing of high quality fruit, which can instruct postharvest treatment for the pineapple industry around the world, and improve market competitiveness. The research results can also provide references for the industrial application of nondestructive detection technology on other large sized fruits.