WO2023077859A1 - 一种基于空间偏移拉曼光谱的带包装食品信号分离方法 - Google Patents

一种基于空间偏移拉曼光谱的带包装食品信号分离方法 Download PDF

Info

Publication number
WO2023077859A1
WO2023077859A1 PCT/CN2022/105463 CN2022105463W WO2023077859A1 WO 2023077859 A1 WO2023077859 A1 WO 2023077859A1 CN 2022105463 W CN2022105463 W CN 2022105463W WO 2023077859 A1 WO2023077859 A1 WO 2023077859A1
Authority
WO
WIPO (PCT)
Prior art keywords
tested
food
sample
laser source
raman
Prior art date
Application number
PCT/CN2022/105463
Other languages
English (en)
French (fr)
Inventor
黄敏
刘振方
朱启兵
赵鑫
Original Assignee
江南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江南大学 filed Critical 江南大学
Publication of WO2023077859A1 publication Critical patent/WO2023077859A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
    • G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
    • G01N21/65—Raman scattering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques

Definitions

  • the invention relates to the technical field of space offset Raman spectroscopy, in particular to a signal separation method for packaged food based on space offset Raman spectroscopy.
  • Food packaging is widely used to ensure the physical characteristics and quality of food during transportation, storage and sales.
  • the traditional food inspection method is to sample and destroy the food packaging to detect the internal quality of the food. This method not only causes a lot of waste of resources, but also has great randomness in detection, so there is an urgent need for a non-destructive quality analysis method for packaged food.
  • Raman spectroscopy is an optical technology that reflects the composition and content of substances through photon-molecular interactions. It is insensitive to water, rich in information, and simple in sample preparation. It is widely used in food, medicine, biology, materials and other fields, but traditional Raman spectroscopy and other optical detection methods are limited to surface detection or samples with clear surfaces (such as glass), which are difficult to apply to packaged food Mass non-destructive analysis.
  • Space-shifted Raman technology is a special Raman spectroscopy technology, which collects Raman scattering signals at different distances from the incident point of excitation light according to the difference in photon transmission between multi-layer materials, and identifies the internal signal of the material.
  • the light intensity spectrum of space-shifted Raman technology can penetrate the turbid surface layer and penetrate deep into the interior of diffuse scattering objects, such as subcutaneous tissue, packaged food, etc., so it can overcome complex layered tissue
  • the measurement defect can extend the penetration depth to two orders of magnitude.
  • the present inventor proposes a kind of band-packaged food signal separation method based on space offset Raman spectrum for above-mentioned problem and technical demand, technical scheme of the present invention is as follows:
  • a signal separation method for packaged food based on space offset Raman spectroscopy comprising:
  • the laser light is emitted by the laser source to irradiate the surface of the sample to be tested, and the Raman spectrum of the sample to be tested is obtained at several positions with different offset distances relative to the laser source to obtain the initial spectral data.
  • the sample to be tested includes the food to be tested and its exterior food packaging;
  • the Raman spectra of some regions are intercepted from the initial spectral data as observation data;
  • spectral peak identification technology uses the spectral peak identification technology to identify the spectral peaks of the initial spectral data, and normalize the attenuation curve of the identified spectral peaks with the increase of the offset distance relative to the laser source;
  • Cluster the normalized attenuation curve, and the number of categories of the cluster is equal to the number of layers of the sample to be tested;
  • the category corresponding to the food to be tested is determined, and the spectral peaks belonging to the category are taken as typical spectral peaks corresponding to the food to be tested.
  • the method also includes:
  • the independent signal component of the food to be tested adopts adaptive iterative weighted penalty least squares to fit the corrected baseline of the independent signal component of the food to be tested;
  • the Raman spectrum whose offset distance is in the range of 0 to i is intercepted from the initial spectral data as the observation data.
  • H( xi ) is the information entropy of the Raman spectrum x i at the offset distance i
  • H is the average value of all information entropy
  • g is the size of the sliding window
  • d represents the initial spectral data relative to the laser The maximum offset distance of the source.
  • the outliers were removed from the sample spectra extracted by each independent repeated acquisition operation, and the average value was taken for smoothing, and the data of the region of interest was intercepted according to the predetermined principle as the initial spectral data.
  • the detection device includes a laser source, several optical fiber probes, optical fibers, focusing lenses, high-pass filters and spectrometers; the laser source and several optical fibers
  • the probes are arranged at intervals to form a row in the horizontal direction and are located above the sample to be tested.
  • the vertical distance between the laser source and each fiber optic probe and the sample to be tested is the same, and the distance between each fiber optic probe and the laser source is Relative to the offset distance of the laser source, several Raman spectra at different offset distances relative to the laser source are collected through each fiber optic probe, and each fiber optic probe is connected to the spectrometer through a focusing lens and a high-pass filter in turn.
  • This application discloses a signal separation method for packaged food based on spatial offset Raman spectroscopy.
  • the method selects observation data through the information entropy extraction of Raman spectroscopy at different offset distances, and combines the clustering and identification of characteristic spectrum peaks. Solve the problem of the attribution of independent signal components obtained by independent component analysis, and finally determine the internal Raman signal of the food to be tested, which is less affected by human factors and has a good separation effect.
  • This method can effectively solve the problem of traditional methods that are difficult to separate overlapping signals and weak signals.
  • the signal problem plays a certain role in promoting the identification and quantification of trace components in packaged foods.
  • the spectral baseline correction strategy is used to improve the reconstruction effect of the spectrum. It can adapt to different packaging, food materials and thickness, and can effectively separate the internal food signal under the premise that the laser can penetrate the food packaging. It can be used as a pre-data processing method for food quality inspection, and can also assist human observation Composition identification of layer samples.
  • Fig. 1 is a schematic flow chart of a method for signal separation of packaged food based on spatial offset Raman spectroscopy disclosed in the present application.
  • Fig. 2 is a schematic structural diagram of a detection device for emitting laser light and obtaining a Raman spectrum of a sample to be tested in the present application.
  • Fig. 3 is a schematic diagram of the effect comparison of the separation of four different groups of samples to be tested using the method of the present application.
  • This application discloses a signal separation method for packaged food based on space-shifted Raman spectroscopy. Please refer to the flow chart shown in Figure 1. The method includes the following steps:
  • the sample to be tested includes the food to be tested 1 and its external food packaging2.
  • the detection device emits laser light and obtains the Raman spectrum of the sample to be tested.
  • the laser source 3 and several fiber optic probes 4 are arranged at intervals to form a row in the horizontal direction and are located above the sample to be tested.
  • the surface of the sample to be tested is basically flat.
  • the distance between the laser source 3 and each fiber optic probe 4 and the sample to be tested is The vertical distance is the same.
  • the distance between each fiber probe 4 and the laser source 3 is the offset distance of the fiber probe relative to the laser source 3, as shown in Figure 2, the distances between each fiber probe 4 and the laser source 3 are respectively D1, D2... Dn.
  • each fiber optic probe 4 is sequentially connected to a focusing lens 5 and a high-pass filter 6 through an optical fiber 8 , and is finally connected to a spectrometer 7 .
  • the laser source 3 When working, the laser source 3 emits laser light to irradiate the surface of the sample to be tested, and the laser can penetrate the food package 2 to reach the food 1 to be tested inside without causing damage to the food package 2 and the food 1 to be tested.
  • the Raman spectrum is collected by each fiber optic probe 4 , and what is collected is the Raman spectrum at the offset distance of the fiber optic probe 4 relative to the laser source 3 .
  • a line scan acquisition device or the like may also be used to acquire Raman spectra at different offset distances.
  • this application performs several independent repeated acquisition operations on the sample to be tested under the same acquisition conditions.
  • the laser is emitted according to the above process and the laser beams at different offset distances are obtained.
  • Mann spectrum the Raman spectrum of the sample to be measured obtained at several locations with different offset distances relative to the laser source is spliced to obtain the sample spectrum.
  • the abscissa of the spliced sample spectrum represents the total offset distance, and the ordinate Raman signals representing different bands.
  • the sample spectra extracted by each independent repeated acquisition operation are averaged and smoothed, and the data of the region of interest is intercepted as the initial spectral data according to a predetermined principle.
  • the predetermined principle is based on the sample It is determined by volume, effective data interval and device characteristics, aiming to reduce the amount of data.
  • the Raman spectra of some regions are intercepted from the initial spectral data as observation data. Specifically, the following steps are included:
  • H( xi ) is the information entropy of the Raman spectrum x i at the offset distance i
  • H is the average value of all information entropy
  • g is the size of the sliding window
  • d represents the initial spectral data relative to the laser
  • the maximum offset distance of the source, eg in Figure 2 is Dn.
  • g is typically set to 100.
  • This application uses the FastICA algorithm to perform independent component analysis on the observed data.
  • the process of using the FastICA algorithm to perform independent component analysis is to use fixed-point iteration to obtain the largest non-Gaussian function value to determine the independent variable, which is expressed as constructing the following function so that the matrix W is in the iterative process Converge to maximum non-Gaussianity in the same direction:
  • Matrix W' is the value obtained by iterative update of matrix W
  • X w is a matrix composed of observation data
  • Decorrelation and scaling are performed by singular value decomposition, resulting in several independent signal components.
  • the same number of singular vectors and corresponding singular values as the number of independent signal components to be separated is selected to reduce the number of parameters to be estimated.
  • spectral peak identification technology uses the spectral peak identification technology to identify the spectral peaks of the initial spectral data, and normalize the attenuation curve of the identified spectral peaks as the offset distance relative to the laser source increases.
  • Cluster the normalized attenuation curves.
  • the number of clustering categories is equal to the number of layers of samples to be tested. Typically, a hierarchical clustering method can be used. If the food to be tested 1 has only one layer of food packaging outside 2, the number of clustering categories is 2.
  • the category corresponding to the food to be tested is determined. Since the internal food to be tested 1 has a spectral characteristic that attenuates slowly, the attenuation value of the spectral peaks that belong to the same category formed after clustering is calculated. The average value, the category with the largest average value of the corresponding attenuation value is taken as the category corresponding to the food to be tested. Then, match the peak positions of the spectral peaks belonging to the category as typical spectral peaks corresponding to the food to be tested with the peak positions of the spectral peaks of each independent signal component, and determine that the category belongs to the independent signal component of the food to be tested.
  • this application does not directly use them for non-destructive quality analysis. Instead, adaptive iterative weighted penalty least squares is used to fit the corrected baseline of the independent signal component of the food to be tested, and then the corrected baseline is subtracted from the independent signal component of the food to be tested to obtain the food to be tested The baseline-free Raman spectrum is then used for non-destructive quality analysis of the food to be tested.
  • the sliding variance of window 5 is used to set the variance threshold, which is considered as a noise signal if it is less than the threshold, thereby determining the fluorescent background and Raman peak area containing random noise in the independent signal component;
  • the selected background baseline was then fitted using air partial least squares with a parameter of 10th order polynomial to achieve the fitting of the corrected baseline.
  • the simulation test results of the simulated packaging sugar composed of white sugar and its external food-grade PE sheet are shown in Figure 3 (a), wherein the mixed The spectrum is the Raman spectrum of the simulated packaging sugar, the separation spectrum is the Raman spectrum of the separated white sugar, the packaging PE is the Raman spectrum of the separated PE sheet, and the reference spectrum is the Raman spectrum of the white sugar without food packaging.
  • the simulation test results for packaged sugar are shown in Figure 3(b)
  • the simulation test results for bagged rice are shown in Figure 3(c)
  • the simulation test results for boxed butter are shown in Figure 3(d) shown.
  • the separation spectrum has a high similarity with the reference spectrum when the food to be tested is not packaged, especially in the test of 3(a) because the packaging material is pure and has a standard thickness, so the similarity is the highest and the signal separation effect is the best. It can be shown that the method of the present application is effective, and the separation effect of the Raman signal of the packaged food is good, and the separation spectrum can be used for quality non-destructive testing.
  • SAM shape-based spectral angle mapping
  • JS amplitude-based Jensen-Shannon divergence
  • PCC linear trend-based Pearson correlation coefficient

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

一种基于空间偏移拉曼光谱的带包装食品信号分离方法，涉及空间偏移拉曼光谱技术领域，该方法基于不同偏移距离处的拉曼光谱的信息熵从初始光谱数据中截取部分区域范围的拉曼光谱作为观测数据，对观测数据进行独立成分分析，分离得到若干个独立信号分量，并结合特征谱峰聚类识别解决独立成分分析分离得到的独立信号分量的归属问题，最终确定内部的待测食品的拉曼信号，该方法受人为因素影响较小，分离效果好，该方法可以有效解决传统方法难以分离重叠信号和微弱信号的问题，在促进带包装食品微量成分识别和定量上起到一定的促进作用。

Description

一种基于空间偏移拉曼光谱的带包装食品信号分离方法 技术领域
本发明涉及空间偏移拉曼光谱技术领域，尤其是一种基于空间偏移拉曼光谱的带包装食品信号分离方法。
背景技术
食品包装广泛用于保证食品在运输、储存和销售过程中的物理特性和质量。传统的食品检测方法是对食品包装进行取样和破坏，以检测食品内部质量。这种方法不仅造成了大量的资源浪费，而且检测的随机性也很大，因此亟待出现一种针对带包装食品的质量无损分析方法。
目前常见的物品质量无损检测方法主要利用拉曼光谱，拉曼光谱是一种通过光子-分子相互作用来反映物质组成和含量的光学技术，因其具有对水不敏感、信息丰富、样品制备简单等特点，被广泛应用于食品、医药、生物、材料等领域，但传统的拉曼光谱和其他光学检测方法仅限于表面检测或表面清晰的样品(如玻璃)，很难应用于带包装食品的质量无损分析。
空间偏移拉曼技术是一种特殊的拉曼光谱技术，其是根据多层材料之间光子传输的差异，采集激发光入射点不同距离的拉曼散射信号，识别材料内部信号的一种技术，与传统的拉曼光谱相比，空间偏移拉曼技术的光强光谱可以穿透浑浊的表面层，深入到扩散散射物体内部，如皮下组织、包装食品等，因此可以克服复杂层状组织的测量缺陷，能将穿透深度扩展至两个数量级。但是在将空间偏移拉曼技术应用于带包装食品的质量无损分析时，如何准确的从获取到的光谱信号中将包装信号和食品信号进行分离，从而准确提取食品的拉曼光谱进行质量检测，成为了该技术应用的难点。以往对层状样品下深层探测的研究主要集中在选择最佳偏移距离以减少地面信号干扰、增强井下信号，也即采用适当尺度系数下的非零偏移处光谱减去零偏移处光谱来表示地下信号。然而，这种减少表面干扰的方法要求操作人员具有较高的领域专业知识，偏移距离和相减比的参数难以确定，受人为因素影响较大，准确度和处理效果都难以保证。
发明内容
本发明人针对上述问题及技术需求，提出了一种基于空间偏移拉曼光谱的 带包装食品信号分离方法，本发明的技术方案如下：
一种基于空间偏移拉曼光谱的带包装食品信号分离方法，该方法包括：
通过激光源发射激光照射在待测样本的表面，并在若干个相对于激光源具有不同偏移距离处获取待测样本的拉曼光谱得到初始光谱数据，待测样本包括待测食品及其外部的食品包装；
基于不同偏移距离处的拉曼光谱的信息熵从初始光谱数据中截取部分区域范围的拉曼光谱作为观测数据；
对观测数据进行独立成分分析，分离得到若干个独立信号分量；
对初始光谱数据基于谱峰识别技术确定对应于待测食品的典型光谱峰，将典型光谱峰的峰位与各个独立信号分量的谱峰的峰位进行匹配，确定对应于待测食品的独立信号分量。
其进一步的技术方案为，对初始光谱数据基于谱峰识别技术确定对应于待测食品的典型光谱峰，包括：
利用谱峰识别技术对初始光谱数据进行谱峰识别，并将识别得到的谱峰随着相对于激光源的偏移距离的增大的衰减曲线进行归一化；
对归一化后的衰减曲线进行聚类，聚类的类别数等于待测样本的层数；
根据每一个类别的数值特征确定对应于待测食品的类别，并将属于类别的谱峰作为对应于待测食品的典型光谱峰。
其进一步的技术方案为，根据每一个类别的数值特征确定对应于待测食品的类别，包括：
将聚类后形成的属于同一个类别的谱峰的衰减值求取平均值，将对应的衰减值的平均值最大的一个类别作为对应于待测食品的类别。
其进一步的技术方案为，该方法还包括：
对待测食品的独立信号分量采用自适应迭代加权惩罚最小二乘拟合待测食品的独立信号分量的校正基线；
从待测食品的独立信号分量中减去校正基线，得到待测食品的无基线的拉曼光谱。
其进一步的技术方案为，基于不同偏移距离处的拉曼光谱的信息熵从初始光谱数据中截取部分区域范围的拉曼光谱作为观测数据，包括：
计算不同偏移距离处的拉曼光谱的信息熵，并确定信息熵的拐点对应的偏移距离i；
从初始光谱数据中截取偏移距离在0到i范围内的拉曼光谱作为观测数据。
其进一步的技术方案为，计算不同偏移距离处的拉曼光谱的信息熵，包括：
通过公式
Figure PCTCN2022105463-appb-000001
计算偏移距离i处的拉曼光谱x i的信息熵，其中
Figure PCTCN2022105463-appb-000002
为偏移距离i处的拉曼光谱x i中的j波段的向下舍入值，参数j∈[1,2,...,J]表示拉曼光谱x i的所有波段中光强数值不同的元素，
Figure PCTCN2022105463-appb-000003
表示这些元素对应的波段在所有波段中出现的概率。
其进一步的技术方案为，按照如下公式采用滑动方差法确定信息熵的拐点对应的偏移距离i：
Figure PCTCN2022105463-appb-000004
其中，H(x i)为偏移距离i处的拉曼光谱x i的信息熵，H为所有信息熵的平均值，g为滑动窗口的大小，d表示初始光谱数据中相对于所述激光源的最大偏移距离。
其进一步的技术方案为，利用FastICA算法对观测数据进行独立成分分析，并选择与待分离得到的独立信号分量的数量相同数量的奇异向量及对应奇异值通过奇异值分解进行去相关和缩放，得到若干个独立信号分量。
其进一步的技术方案为，在若干个相对于激光源具有不同偏移距离处获取待测样本的拉曼光谱得到初始光谱数据，包括：
在相同的采集条件下对待测样本进行若干次独立重复采集操作，在进行每一次独立重复采集操作时，将在若干个相对于激光源具有不同偏移距离处获取到的待测样本的拉曼光谱进行拼接得到样本光谱；
对各次独立重复采集操作提取到的样本光谱剔除异常值后取均值进行平滑，并按照预定原则截取感兴趣区域的数据作为初始光谱数据。
其进一步的技术方案为，通过检测装置发射激光并获取待测样本的拉曼光谱，检测装置包括激光源、若干个光纤探头、光纤、聚焦镜头、高通滤波片和光谱仪；激光源和若干个光纤探头间隔排布在水平方向上形成一排并位于待测样本的上方，激光源和各个光纤探头与待测样本之间的垂直距离相同，每个光纤探头与激光源之间的距离为光纤探头相对于激光源的偏移距离，通过各个光纤探头采集若干个相对于激光源具有不同偏移距离处的拉曼光谱，各个光纤探头依次通过聚焦镜头和高通滤波片后连接至光谱仪。
本发明的有益技术效果是：
本申请公开了一种基于空间偏移拉曼光谱的带包装食品信号分离方法，该方法通过不同偏移距离处的拉曼光谱的信息熵提取选择出观测数据，并结合特征谱峰聚类识别解决独立成分分析分离得到的独立信号分量的归属问题，最终确定内部的待测食品的拉曼信号，受人为因素影响较小，分离效果好，该方法可以有效解决传统方法难以分离重叠信号和微弱信号的问题，在促进带包装食品微量成分识别和定量上起到一定的促进作用。
进一步的，针对FastICA分离的信号存在异常贡献问题，通过光谱基线的校正策略提高光谱的重构效果。能自适应不同的包装、食品材料和厚度，在激光能穿透食品包装的前提下都可有效分离内部食品信号，即可以作为一种食品品质检测的前数据处理手段，也可辅助人为观测分层样品的成分鉴定。
附图说明
图1是本申请公开的基于空间偏移拉曼光谱的带包装食品信号分离方法的流程示意图。
图2是本申请中发射激光并获取待测样本的拉曼光谱的检测装置的结构示意图。
图3是利用本申请的方法对四组不同的待测样本进行分离的效果对比示意图。
具体实施方式
下面结合附图对本发明的具体实施方式做进一步说明。
本申请公开了一种基于空间偏移拉曼光谱的带包装食品信号分离方法，请参考图1所示的流程图，该方法包括如下步骤：
1、通过激光源发射激光照射在待测样本的表面，并在若干个相对于激光源具有不同偏移距离处获取待测样本的拉曼光谱得到初始光谱数据，待测样本包括待测食品1及其外部的食品包装2。
本申请通过检测装置发射激光并获取待测样本的拉曼光谱，请参考图2，检测装置包括激光源3、若干个光纤探头4、聚焦镜头5、高通滤波片6和光谱仪7。激光源3和若干个光纤探头4间隔排布在水平方向上形成一排并位于待测样本的上方，待测样本的表面基本平整，激光源3和各个光纤探头4与待测样本之间的垂直距离相同。每个光纤探头4与激光源3之间的距离即为该光纤探头相对于激光源3的偏移距离，如图2所示，各个光纤探头4与激光源3之 间的距离分别为D1、D2……Dn。实际应用时，激光源3与各个光纤探头4的每两个相邻组件之间的距离相等从而形成均匀排布结构，或者也可以不相等。各个光纤探头4依次通过光纤8连接聚焦镜头5和高通滤波片6，最终连接至光谱仪7。
在工作时，通过激光源3发射激光照射待测样本的表面，激光能穿透食品包装2到达内部的待测食品1且不对食品包装2和待测食品1产生损伤。通过各个光纤探头4采集拉曼光谱，采集到的即为光纤探头4相对于激光源3的偏移距离处的拉曼光谱。实际也可以采用线扫描采集设备等来获取不同偏移距离处的拉曼光谱。
为了提高数据的有效性，本申请在相同的采集条件下对待测样本进行若干次独立重复采集操作，在进行每一次独立重复采集操作时，按照如上过程发射激光并获取不同偏移距离处的拉曼光谱，将在若干个相对于激光源具有不同偏移距离处获取到的待测样本的拉曼光谱进行拼接得到样本光谱，拼接得到的样本光谱的横坐标表示总的偏移距离，纵坐标表示不同波段的拉曼信号。对各次独立重复采集操作提取到的样本光谱剔除异常值后(包括宇宙噪声引起的峰谷)取均值进行平滑，并按照预定原则截取感兴趣区域的数据作为初始光谱数据，该预定原则根据样本体积、有效数据区间及设备特性来确定，旨在缩减数据量。
2、基于不同偏移距离处的拉曼光谱的信息熵从初始光谱数据中截取部分区域范围的拉曼光谱作为观测数据。具体的，包括如下步骤：
(1)计算不同偏移距离处的拉曼光谱的信息熵。
通过公式
Figure PCTCN2022105463-appb-000005
计算偏移距离i处的拉曼光谱x i的信息熵，其中
Figure PCTCN2022105463-appb-000006
为偏移距离i处的拉曼光谱x i中的j波段的向下舍入值，这是拉曼光谱信号是一种高分辨率的近似连续信号，必须通过数据量化将其分割成区间整数值进行统计计算，下舍入是为了降低拉曼光谱的随机噪声。参数j∈[1,2,...,J]描述无重复元素，表示拉曼光谱x i的所有波段中光强数值不同的元素，
Figure PCTCN2022105463-appb-000007
表示这些元素对应的波段在所有波段中出现的概率。
(2)确定信息熵的拐点对应的偏移距离i。随着偏移距离的增加，信息熵值逐渐减小，波动较大，因此本申请采用滑动方差法确定信息熵的拐点，表示为：
Figure PCTCN2022105463-appb-000008
其中，H(x i)为偏移距离i处的拉曼光谱x i的信息熵，H为所有信息熵的平均值，g为滑动窗口的大小，d表示初始光谱数据中相对于所述激光源的最大偏移距离，比如在图2中即为Dn。考虑带包装食品的空间分辨率的波动和信息熵的衰减，比较典型的将g设为100。
(3)从初始光谱数据中截取偏移距离在0到i范围内的拉曼光谱作为观测数据，由此对初始光谱数据做了进一步筛选，保留其高阶统计特性。
3、对观测数据进行独立成分分析，分离得到若干个独立信号分量。
本申请利用FastICA算法对观测数据进行独立成分分析，利用FastICA算法进行独立成分分析的过程为使用定点迭代获得最大的非高斯函数值来确定独立变量，表示为构造如下函数使得矩阵W在迭代过程中同方向收敛到最大非高斯性：
Figure PCTCN2022105463-appb-000009
矩阵W′是矩阵W迭代更新得到的值，X w为观测数据构成的矩阵，G(x)＝logcosh(x)为靠近源信号的非二次累积分布函数。
通过奇异值分解进行去相关和缩放，最终得到若干个独立信号分量。选择与待分离得到的独立信号分量的数量相同数量的奇异向量及对应奇异值以减少待估计参数数量。
4、对初始光谱数据基于谱峰识别技术确定对应于待测食品的典型光谱峰，将典型光谱峰的峰位与各个独立信号分量的谱峰的峰位进行匹配，确定对应于待测食品的独立信号分量。
具体的：利用谱峰识别技术对初始光谱数据进行谱峰识别，并将识别得到的谱峰随着相对于激光源的偏移距离的增大的衰减曲线进行归一化。对归一化后的衰减曲线进行聚类，聚类的类别数等于待测样本的层数，比较典型的可以采用层次聚类的方法，比较常规的若待测食品1外部只有一层食品包装2，则聚类的类别数为2。
根据每一个类别的数值特征确定对应于待测食品的类别，由于内部的待测食品1具有衰减较慢的光谱特性，因此将聚类后形成的属于同一个类别的谱峰的衰减值求取平均值，将对应的衰减值的平均值最大的一个类别作为对应于待测食品的类别。然后将属于该类别的谱峰作为对应于待测食品的典型光谱峰与各个独立信号分量的谱峰的峰位进行匹配，确定类别归属于待测食品的独立信号分量。
进一步的，在得到待测食品的独立信号分量后，考虑到独立信号分量中存在负贡献及基线漂移的现象，因此本申请并不直接将其用于质量无损分析。而是对待测食品的独立信号分量采用自适应迭代加权惩罚最小二乘拟合待测食品的独立信号分量的校正基线，然后从待测食品的独立信号分量中减去校正基线，得到待测食品的无基线的拉曼光谱，然后用于待测食品的质量无损分析。具体的，首先由于随机噪声方差较小，利用窗口5的滑动方差，设定方差阈值，小于阈值被认为是噪声信号，由此确定独立信号分量中含有随机噪声的荧光背景和拉曼峰区域；然后利用空气偏最小二乘拟合选定的背景基线，参数为10阶多项式，实现校正基线的拟合。
为了说明本申请方法的有效性，本申请进行了如下几个实例试验：对白糖及其外部的食品级PE板材组成的模拟包装糖的模拟试验结果如图3(a)所示，其中，混合光谱为模拟包装糖的拉曼光谱，分离光谱为分离得到的白糖的拉曼光谱，包装PE为分离得到的PE板材的拉曼光谱，参考光谱为不带食品包装的白糖的拉曼光谱。类似的，对包装白糖的模拟试验结果如图3(b)所示，对袋装大米的模拟试验结果如图3(c)所示，对盒装黄油的模拟试验结果如图3(d)所示。通过基于形状的光谱角映射(SAM)、基于振幅的Jensen-Shannon散度(JS)和基于线性趋势的Pearson相关系数(PCC)等方法将分离光谱与参考光谱进行相似度比较可知，提取得到的分离光谱与待测食品不带食品包装时的参考光谱相似度较高，尤其是3(a)的试验中由于包装材料纯净，拥有标准厚度，所以相似度最高、信号分离效果最佳，由此可以说明本申请方法的有效性，对带包装食品的拉曼信号的分离效果较好，分离光谱可以用于质量无损检测。
以上所述的仅是本申请的优选实施方式，本发明不限于以上实施例。可以理解，本领域技术人员在不脱离本发明的精神和构思的前提下直接导出或联想到的其他改进和变化，均应认为包含在本发明的保护范围之内。

Claims (10)

  1. 一种基于空间偏移拉曼光谱的带包装食品信号分离方法，其特征在于，所述方法包括：
    通过激光源发射激光照射在待测样本的表面，并在若干个相对于所述激光源具有不同偏移距离处获取所述待测样本的拉曼光谱得到初始光谱数据，所述待测样本包括待测食品及其外部的食品包装；
    基于不同偏移距离处的拉曼光谱的信息熵从所述初始光谱数据中截取部分区域范围的拉曼光谱作为观测数据；
    对所述观测数据进行独立成分分析，分离得到若干个独立信号分量；
    对所述初始光谱数据基于谱峰识别技术确定对应于待测食品的典型光谱峰，将所述典型光谱峰的峰位与各个独立信号分量的谱峰的峰位进行匹配，确定对应于待测食品的独立信号分量。
  2. 根据权利要求1所述的方法，其特征在于，所述对所述初始光谱数据基于谱峰识别技术确定对应于待测食品的典型光谱峰，包括：
    利用谱峰识别技术对所述初始光谱数据进行谱峰识别，并将识别得到的谱峰随着相对于所述激光源的偏移距离的增大的衰减曲线进行归一化；
    对归一化后的衰减曲线进行聚类，聚类的类别数等于所述待测样本的层数；
    根据每一个类别的数值特征确定对应于待测食品的类别，并将属于所述类别的谱峰作为对应于所述待测食品的典型光谱峰。
  3. 根据权利要求2所述的方法，其特征在于，所述根据每一个类别的数值特征确定对应于待测食品的类别，包括：
    将聚类后形成的属于同一个类别的谱峰的衰减值求取平均值，将对应的衰减值的平均值最大的一个类别作为对应于待测食品的类别。
  4. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    对所述待测食品的独立信号分量采用自适应迭代加权惩罚最小二乘拟合所述待测食品的独立信号分量的校正基线；
    从所述待测食品的独立信号分量中减去所述校正基线，得到所述待测食品的无基线的拉曼光谱。
  5. 根据权利要求1所述的方法，其特征在于，所述基于不同偏移距离处的拉曼光谱的信息熵从所述初始光谱数据中截取部分区域范围的拉曼光谱作为观 测数据，包括：
    计算不同偏移距离处的拉曼光谱的信息熵，并确定信息熵的拐点对应的偏移距离i；
    从所述初始光谱数据中截取偏移距离在0到i范围内的拉曼光谱作为观测数据。
  6. 根据权利要求5所述的方法，其特征在于，所述计算不同偏移距离处的拉曼光谱的信息熵，包括：
    通过公式
    Figure PCTCN2022105463-appb-100001
    计算偏移距离i处的拉曼光谱x i的信息熵，其中
    Figure PCTCN2022105463-appb-100002
    为偏移距离i处的拉曼光谱x i中的j波段的向下舍入值，参数j∈[1,2,...,J]表示拉曼光谱x i的所有波段中光强数值不同的元素，
    Figure PCTCN2022105463-appb-100003
    表示这些元素对应的波段在所有波段中出现的概率。
  7. 根据权利要求5所述的方法，其特征在于，按照如下公式采用滑动方差法确定信息熵的拐点对应的偏移距离i：
    Figure PCTCN2022105463-appb-100004
    其中，H(x i)为偏移距离i处的拉曼光谱x i的信息熵，
    Figure PCTCN2022105463-appb-100005
    为所有信息熵的平均值，g为滑动窗口的大小，d表示所述初始光谱数据中相对于所述激光源的最大偏移距离。
  8. 根据权利要求1所述的方法，其特征在于，利用FastICA算法对所述观测数据进行独立成分分析，并选择与待分离得到的独立信号分量的数量相同数量的奇异向量及对应奇异值通过奇异值分解进行去相关和缩放，得到若干个独立信号分量。
  9. 根据权利要求1所述的方法，其特征在于，所述在若干个相对于所述激光源具有不同偏移距离处获取所述待测样本的拉曼光谱得到初始光谱数据，包括：
    在相同的采集条件下对所述待测样本进行若干次独立重复采集操作，在进行每一次独立重复采集操作时，将在若干个相对于所述激光源具有不同偏移距离处获取到的所述待测样本的拉曼光谱进行拼接得到样本光谱；
    对各次独立重复采集操作提取到的样本光谱剔除异常值后取均值进行平滑，并按照预定原则截取感兴趣区域的数据作为所述初始光谱数据。
  10. 根据权利要求1所述的方法，其特征在于，通过检测装置发射激光并获取所述待测样本的拉曼光谱，所述检测装置包括激光源、若干个光纤探头、光纤、聚焦镜头、高通滤波片和光谱仪，所述激光源和若干个光纤探头间隔排布在水平方向上形成一排并位于所述待测样本的上方，所述激光源和各个光纤探头与所述待测样本之间的垂直距离相同，每个光纤探头与所述激光源之间的距离为所述光纤探头相对于所述激光源的偏移距离，通过各个光纤探头采集若干个相对于所述激光源具有不同偏移距离处的拉曼光谱，各个光纤探头依次通过所述聚焦镜头和高通滤波片后连接至所述光谱仪。
PCT/CN2022/105463 2021-11-03 2022-07-13 一种基于空间偏移拉曼光谱的带包装食品信号分离方法 WO2023077859A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111293805.6A CN113984736B (zh) 2021-11-03 2021-11-03 一种基于空间偏移拉曼光谱的带包装食品信号分离方法
CN202111293805.6 2021-11-03

Publications (1)

Publication Number Publication Date
WO2023077859A1 true WO2023077859A1 (zh) 2023-05-11

Family

ID=79746117

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/105463 WO2023077859A1 (zh) 2021-11-03 2022-07-13 一种基于空间偏移拉曼光谱的带包装食品信号分离方法

Country Status (2)

Country Link
CN (1) CN113984736B (zh)
WO (1) WO2023077859A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118010649A (zh) * 2024-04-09 2024-05-10 西安大业食品有限公司 用于食品的污染检测方法
CN118094210A (zh) * 2024-04-17 2024-05-28 国网上海市电力公司 一种基于欠定盲源分离的储能系统充放电行为识别方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113984736B (zh) * 2021-11-03 2022-07-29 江南大学 一种基于空间偏移拉曼光谱的带包装食品信号分离方法
CN117007577B (zh) * 2023-10-07 2023-12-08 生态环境部华南环境科学研究所(生态环境部生态环境应急研究所) 一种污染物毒性智能检测系统
CN117434048B (zh) * 2023-11-14 2024-06-18 江南大学 基于蒙特卡洛实现拉曼光源转换的冷鲜食品无损检测方法

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101559513A (zh) * 2009-05-21 2009-10-21 山东大学 基于激光测距的集装箱波纹板焊接轨迹检测与控制方法
CN105300954A (zh) * 2015-09-02 2016-02-03 中南大学 一种锑矿浮选非均质泡沫层矿物的拉曼光谱表征方法
CN105334204A (zh) * 2015-10-16 2016-02-17 东北大学 一种基于傅里叶变换拉曼光谱分析方法
CN106153658A (zh) * 2016-09-21 2016-11-23 中国科学院合肥物质科学研究院 一种能量色散x射线荧光光谱中多元素特征谱峰识别方法
CN108169201A (zh) * 2016-12-08 2018-06-15 同方威视技术股份有限公司 用于扣除包装干扰的拉曼光谱检测方法
CN108827934A (zh) * 2018-05-21 2018-11-16 江南大学 一种基于盲源分离拉曼散射图像的带包装食品品质无损检测方法
CN108844939A (zh) * 2018-03-14 2018-11-20 西安电子科技大学 基于非对称加权最小二乘的拉曼光谱检测基线校正方法
CN109765212A (zh) * 2019-03-11 2019-05-17 广西科技大学 拉曼光谱中不同步褪色荧光的消除方法
CN110243806A (zh) * 2019-07-30 2019-09-17 江南大学 拉曼光谱下基于相似度的混合物组分识别方法
CN112001256A (zh) * 2020-07-24 2020-11-27 东风汽车集团有限公司 混合信号去工频干扰方法及系统
CN112014376A (zh) * 2020-09-09 2020-12-01 江南大学 一种基于空间偏移拉曼光谱的虾肉新鲜度检测方法
CN112834481A (zh) * 2020-12-31 2021-05-25 宁波海关技术中心 一种拉曼光谱增强测量系统及测量方法
WO2021119946A1 (en) * 2019-12-16 2021-06-24 Mekitec Oy Food inspection solution
CN113109318A (zh) * 2021-03-26 2021-07-13 中国科学院西安光学精密机械研究所 基于光谱峰高直接提取的拉曼光谱定量分析方法及系统
CN113984736A (zh) * 2021-11-03 2022-01-28 江南大学 一种基于空间偏移拉曼光谱的带包装食品信号分离方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104749155B (zh) * 2013-12-27 2018-01-16 同方威视技术股份有限公司 用于检测容装体中的样品的拉曼光谱检测方法
CN110580492A (zh) * 2019-04-30 2019-12-17 上海铁大电信科技股份有限公司 一种基于小幅波动检测的轨道电路故障前兆发现方法

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101559513A (zh) * 2009-05-21 2009-10-21 山东大学 基于激光测距的集装箱波纹板焊接轨迹检测与控制方法
CN105300954A (zh) * 2015-09-02 2016-02-03 中南大学 一种锑矿浮选非均质泡沫层矿物的拉曼光谱表征方法
CN105334204A (zh) * 2015-10-16 2016-02-17 东北大学 一种基于傅里叶变换拉曼光谱分析方法
CN106153658A (zh) * 2016-09-21 2016-11-23 中国科学院合肥物质科学研究院 一种能量色散x射线荧光光谱中多元素特征谱峰识别方法
CN108169201A (zh) * 2016-12-08 2018-06-15 同方威视技术股份有限公司 用于扣除包装干扰的拉曼光谱检测方法
CN108844939A (zh) * 2018-03-14 2018-11-20 西安电子科技大学 基于非对称加权最小二乘的拉曼光谱检测基线校正方法
CN108827934A (zh) * 2018-05-21 2018-11-16 江南大学 一种基于盲源分离拉曼散射图像的带包装食品品质无损检测方法
CN109765212A (zh) * 2019-03-11 2019-05-17 广西科技大学 拉曼光谱中不同步褪色荧光的消除方法
CN110243806A (zh) * 2019-07-30 2019-09-17 江南大学 拉曼光谱下基于相似度的混合物组分识别方法
WO2021119946A1 (en) * 2019-12-16 2021-06-24 Mekitec Oy Food inspection solution
CN112001256A (zh) * 2020-07-24 2020-11-27 东风汽车集团有限公司 混合信号去工频干扰方法及系统
CN112014376A (zh) * 2020-09-09 2020-12-01 江南大学 一种基于空间偏移拉曼光谱的虾肉新鲜度检测方法
CN112834481A (zh) * 2020-12-31 2021-05-25 宁波海关技术中心 一种拉曼光谱增强测量系统及测量方法
CN113109318A (zh) * 2021-03-26 2021-07-13 中国科学院西安光学精密机械研究所 基于光谱峰高直接提取的拉曼光谱定量分析方法及系统
CN113984736A (zh) * 2021-11-03 2022-01-28 江南大学 一种基于空间偏移拉曼光谱的带包装食品信号分离方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118010649A (zh) * 2024-04-09 2024-05-10 西安大业食品有限公司 用于食品的污染检测方法
CN118010649B (zh) * 2024-04-09 2024-06-21 西安大业食品有限公司 用于食品的污染检测方法
CN118094210A (zh) * 2024-04-17 2024-05-28 国网上海市电力公司 一种基于欠定盲源分离的储能系统充放电行为识别方法

Also Published As

Publication number Publication date
CN113984736A (zh) 2022-01-28
CN113984736B (zh) 2022-07-29

Similar Documents

Publication Publication Date Title
WO2023077859A1 (zh) 一种基于空间偏移拉曼光谱的带包装食品信号分离方法
Kimuli et al. Application of SWIR hyperspectral imaging and chemometrics for identification of aflatoxin B1 contaminated maize kernels
CN104198432B (zh) 应用近红外光谱技术无损鉴别钢结构防火涂料品牌的方法
Wu et al. Design and development of an on-line fluorescence spectroscopy system for detection of aflatoxin in pistachio nuts
Pan et al. Eggshell crack detection based on computer vision and acoustic response by means of back-propagation artificial neural network
CN102033043A (zh) 基于高光谱图像技术的粮粒含水率检测方法
CN110632021A (zh) 基于便携式近红外光谱仪的光谱检测方法及系统
CN110473194A (zh) 基于多图像块阈值分割算法的水果表面缺陷检测方法
Cui et al. Identification of maize seed varieties based on near infrared reflectance spectroscopy and chemometrics
Jiang et al. Discrimination of moldy wheat using terahertz imaging combined with multivariate classification
Liu et al. A packaged food internal Raman signal separation method based on spatially offset Raman spectroscopy combined with FastICA
Turgut et al. Potential of image analysis based systems in food quality assessments and classifications
Liu et al. Digital image analysis method for rapid measurement of rice degree of milling
WO2020130942A1 (en) A non-destructive system and method for determining the quality of chinese herb using terahertz time-domain spectroscopy
Ovide et al. The analysis of glass from portable electronic devices and glass accessories using µ-XRF for forensic investigations
Liu et al. Estimation of chlorophyll content in maize canopy using wavelet denoising and SVR method
CN110186870B (zh) 一种极限学习机光谱模型判别恩施玉露茶鲜叶产地的方法
CN115718081B (zh) 基于光谱指纹的琥珀产地溯源模型的构建方法及应用
Zhao et al. Method for extracting Raman spectra characteristic variables of biological sample based on Hilbert–Huang transform
Sun et al. Optical methods for firmness assessment of fresh produce: a review
JP6612423B2 (ja) 試料の透過度を測定するためのアセンブリ
Tao et al. Discriminant analysis of red wines from different aging ways by information fusion of NIR and MIR spectra
CN115146219A (zh) 基于导数光谱和三次平滑的拉曼光谱尖峰去除方法和系统
CN113418889A (zh) 一种基于深度学习的干菜水分含量和菌落总数的实时检测方法
Güneş et al. Detection of aflatoxin contaminated figs using Near-Infrared (NIR) reflectance spectroscopy

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22888894

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE