基于可见/近红外光谱技术的橘小实蝇侵染柑橘检测研究

Detection of Citrus infestation by Bactrocera dorsalis based on visible/near infrared spectroscopy

中文关键词:  橘小实蝇  柑橘  可见/近红外光谱  快速无损检测  偏最小二乘判别分析

英文关键词:Bactrocera dorsalis  Citrus  visible/near infrared spectroscopy  rapid non-destructive detection  partial least squares discriminant analysis

中文摘要:

      目的 使用可见/近红外光谱技术实施橘小实蝇侵染柑橘不同时期的无损检测。方法 研究选取人工制备的不同侵染时期的柑橘样本作为研究对象, 利用搭建的可见/近红外光谱系统测量的光谱信息结合人工标定的侵染时期, 对原始光谱进行了5种预处理, 采用竞争性自适应重加权算法(competitive adaptive reweighted sampling, CARS)、连续投影算法(successive projections algorithm, SPA)两类方法提取反映侵染柑橘时期变化的光谱特征波长, 应用偏最小二乘判别分析(partial least squares discriminant analysis, PLS-DA)建立基于特征波长光谱的柑橘侵染时期分类模型, 对比分析不同光谱预处理方法的模型分类效果。结果 原始光谱经多元散射校正(multiplicative scatter correction, MSC)预处理的模型分类效果最佳, 分别经CARS方法和SPA方法优选出了34和16个光谱特征波长。采用MSC-CARS-PLS-DA方法构建的模型分类效果最好, 总准确率、假阳率分别为96.8%和0.0%, 模型对健康柑橘和侵染柑橘有较强的分类能力。结论 通过可见/近红外光谱结合PLS-DA判别方法, 可以实现橘小实蝇侵染柑橘的无损检测, 为今后继续开展橘小实蝇侵染柑橘光谱检测研究提供参考依据。

英文摘要:

      Objective To implement nondestructive detection of different infestation periods of Citrus infested by Bactrocera dorsalis using visible/near infrared spectroscopy. Methods Citrus with different infestation periods prepared artificially were selected as the research object, and the spectral information obtained from the constructed visible/near infrared spectroscopy system was combined with the manually defined infestation periods, and five kinds of pre-processing were performed on the original spectra, and competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were used to extract the spectral characteristic wavelengths reflecting the changes of infested Citrus periods, and apply partial least squares discriminant analysis (PLS-DA) to establish the classification model of Citrus infestation periods based on the characteristic wavelength spectra, and compare and analyze the model classification effects of different spectral preprocessing methods. Results The model with raw spectra pretreated by multiplicative scatter correction (MSC) had the best classification results, and 34 and 16 spectral characteristic wavelengths were selected by the CARS and SPA, respectively. The model constructed by MSC-CARS-PLS-DA method had the best classification results, with the total accuracy rate and that false positive rate reach 96.8% and 0.0%, respectively, and the model had good classification ability for healthy and infested Citrus. Conclusion The visible/near infrared spectroscopy combined with PLS-DA can achieve nondestructive detection of Citrus infested by Bactrocera dorsalis, provides a reference for continuing the research on spectral detection of Bactrocera dorsalis infection in the future.