WO2018121122A1 - 用于物品查验的拉曼光谱检测方法和电子设备 - Google Patents

用于物品查验的拉曼光谱检测方法和电子设备 Download PDF

Info

Publication number
WO2018121122A1
WO2018121122A1 PCT/CN2017/111624 CN2017111624W WO2018121122A1 WO 2018121122 A1 WO2018121122 A1 WO 2018121122A1 CN 2017111624 W CN2017111624 W CN 2017111624W WO 2018121122 A1 WO2018121122 A1 WO 2018121122A1
Authority
WO
WIPO (PCT)
Prior art keywords
raman spectrum
item
standard
raman
inspected
Prior art date
Application number
PCT/CN2017/111624
Other languages
English (en)
French (fr)
Inventor
王红球
左佳倩
谭玲玉
苟巍
刘德平
黄健华
刘来福
戴晓理
田睿
王康琳
钱宏伟
王保刚
周小平
刘德臣
孙晓东
Original Assignee
同方威视技术股份有限公司
中华人民共和国北京出入境检验检疫局
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 同方威视技术股份有限公司, 中华人民共和国北京出入境检验检疫局 filed Critical 同方威视技术股份有限公司
Publication of WO2018121122A1 publication Critical patent/WO2018121122A1/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
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2201/00—Features of devices classified in G01N21/00
    • G01N2201/10—Scanning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Definitions

  • the present invention relates generally to the field of Raman spectroscopy, and in particular to a Raman spectroscopy method for inspection of articles, and more particularly to a Raman spectroscopy method for rapid inspection of entry and exit special items.
  • Raman spectroscopy is a molecular vibrational spectroscopy that reflects the fingerprint characteristics of molecules and can be used to detect substances. Raman spectroscopy detects and identifies a substance by detecting a Raman spectrum produced by the Raman scattering effect of the analyte on the excitation light. Raman spectroscopy has been widely used in liquid security, jewelry testing, explosives testing, drug testing, drug testing, pesticide residue testing and other fields.
  • the existing on-site inspection for special items for entry and exit mainly adopts the real-time monitoring and supervision function of the on-site high-definition camera device.
  • the on-site inspection personnel open the box to check whether the cargo information and the declaration information are consistent. If the cargo information is consistent, it is considered The goods can be released.
  • the above methods mainly have the following disadvantages: (1) The on-site inspection personnel have a large workload. Checking whether the approval form and the label information are uncoordinated requires the on-site inspection staff to check the label information one by one, and manually compare the complex information such as product name, specification, traits and quantity (mostly non-Chinese language information such as English). . (2) The inspection vulnerability is obvious. Checking the label information cannot verify that the label information of the goods is consistent with the actual contents, and cannot effectively identify the storage or transportation of high-risk special items in low-risk or non-special items.
  • the present invention has been made in order to overcome or eliminate at least one of the problems and disadvantages of the prior art.
  • At least one object of the present invention is to provide a Raman spectroscopy method and an electronic device for inspection of articles, and more particularly to a Raman spectroscopy method for rapid inspection of entry and exit special articles, which can enhance the accuracy of on-site inspection of articles, and It can realize simultaneous matching and checking of multiple information through one test, and speed up the on-site inspection efficiency of articles.
  • a Raman spectroscopy method for inspection of an item comprising the steps of:
  • Raman spectroscopy acquisition step collecting the Raman spectrum of the item to be inspected.
  • Alignment and determination step comparing the Raman spectrum of the collected item to be inspected with the Raman spectrum of the standard item stored in the standard database to determine whether the item to be inspected matches the standard item,
  • the comparing step comprises: classifying the Raman spectrum of the item to be inspected by using a support vector machine to achieve an alignment of the Raman spectrum of the item to be inspected with the Raman spectrum of the standard item stored in the standard database.
  • the step of classifying the Raman spectrum of the item to be inspected using the support vector machine comprises:
  • the Raman spectrum of the item to be inspected is classified using the established classifier.
  • performing a sparse transform on the Raman spectrum of the training sample comprises the steps of:
  • the target vector c for sparse representation of the Raman spectral information y is calculated using the following objective function:
  • the number of non-zero elements in the vector c is fixed to k, N is the length of the vector c, k ⁇ N, and i is the number of the vector c.
  • the comparing and determining steps further comprise:
  • the Raman spectrum of the item to be inspected is similar to the Raman spectrum of the standard item stored in the standard database;
  • the Raman spectrum of the item to be inspected is classified by the support vector machine to realize the Raman spectrum of the item to be inspected and the Raman of the standard item stored in the standard database. Spectral alignment.
  • the similarity measure of the Raman spectrum of the item to be inspected and the Raman spectrum of the standard item stored in the standard database includes:
  • a correlation coefficient between a feature vector of a Raman spectrum of the article to be inspected and a feature vector of a Raman spectrum of a standard article stored in a standard database is calculated, and the calculated correlation coefficient is used as a result of the similarity measure.
  • the Raman spectroscopy detection method further comprises the following steps:
  • the Raman spectroscopy acquisition step comprises: performing surface and internal scanning of the article to be inspected to acquire a Raman spectrum of a surface component of the article to be inspected and a Raman spectrum of a component of the actual content;
  • the step of measuring the similarity of the Raman spectrum of the article to be inspected with the Raman spectrum of the standard article stored in the standard database includes: separately analyzing the Raman spectrum of the surface component of the article to be inspected and the surface of the standard article stored in the standard database The Raman spectrum of the composition and the Raman spectrum of the composition of the actual content of the item to be inspected are similarly measured with the Raman spectrum of the composition of the actual content of the standard item stored in the standard database;
  • the result that the result of the similarity measure is greater than the preset threshold includes only the case where the results of the two similarity measures are greater than the preset threshold.
  • the step of establishing a standard database further comprises: collecting a product name, a source company, a size, and a picture information of the standard item, and storing the information in a standard database.
  • an electronic device including:
  • a memory for storing executable instructions
  • a processor for executing executable instructions stored in the memory to perform a Raman spectroscopy method for item inspection as described in any aspect or embodiment of the present invention.
  • the detection method according to the embodiment of the present invention eliminates the process of on-site inspection personnel comparing the plurality of information of the articles to be inspected one by one, and speeds up the on-site inspection speed; and the identification method of matching the feature information is adopted, thereby improving the accuracy of the on-site inspection.
  • FIG. 1 is a flow chart showing a method for detecting an entry and exit special item inspection according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of the principle of a support vector machine
  • Figure 3 shows a flow chart for establishing a support vector machine classifier
  • FIG. 4 shows a flow chart of a detection method using an improved support vector machine in accordance with an embodiment of the present invention.
  • FIG. 5 shows an example hard of an electronic device for performing a detecting method according to an embodiment of the present invention.
  • the entry and exit special articles refer to microorganisms, human tissues, human genetic resources, biological products, blood, blood products and the like. Due to the wide variety of special items on the site of inspection and quarantine inspection, the variety of goods and the complexity of packaging, many special items have the characteristics of less sample, difficult to observe by the naked eye and unclear composition. In addition, the circulation of special items is more difficult than the general goods. The potential risks are even greater; for on-site inspections, many special items such as microorganisms, blood and other biological agents can not be tested at the site.
  • the Raman spectroscopy method for inspection of articles according to the embodiment of the present invention can quickly check the entry and exit special items, and at the same time improve the accuracy of on-site inspection.
  • a complete Raman spectrum information library of a relatively complete entry and exit special item is first established, and the information base may be referred to as a standard database. Then, during the on-site inspection process, according to the item name indicated on the manifest, the standard items already existing in the standard database are selected, and then the surface to be inspected is scanned, and then the spectrum of the object to be inspected is scanned and The standard spectrum in the standard database is compared and analyzed, and the "similarity measure" algorithm and the "improved support vector machine” algorithm are combined to complete the rapid feature recognition of the special items for entry and exit.
  • the standard spectrum of representative entry and exit special items may be established first, or may be accumulated during use or provided by the inspection and quarantine inspection unit.
  • the Raman signature of the surface package and the actual content may be collected at the site where the content of the special item can be seen without interference from other objects.
  • the graph is then organized to create a Raman spectrum information library to form the standard database.
  • the standard database maintains the overall information of the entry and exit special items.
  • the product name, source company, size, picture, etc. of the standard item are indicated; on the other hand, each entry and exit special item has two corresponding spectra.
  • a spectrum is a spectrum acquisition of the surface of a specific site. The acquired spectrum information of the surface components is collected, and the other is the spectrum acquisition of the interior of the same site. What is collected is the actual internal Spectral information of the ingredients.
  • the spectrum information collected on the surface is compared with the surface standard spectrum in the spectrum library, and the collected internal spectrum information is compared with the internal standard spectrum in the spectrum library. If the threshold is exceeded twice, it is considered that the item to be inspected can be the same item as the standard item, and the matching accuracy is increased.
  • FIG. 1 is a flow chart showing a method for detecting an entry and exit special item inspection according to an embodiment of the present invention.
  • the detection method will be described in detail below with reference to FIG. 1, and the special item for entry and exit according to an embodiment of the present invention is checked.
  • the detection method mainly includes the following steps:
  • Raman spectroscopy acquisition step selecting an item to be inspected on the detecting device, and scanning the item to be inspected to collect a Raman spectrum of the item to be inspected;
  • Raman spectral alignment and determination step using a recognition algorithm according to an embodiment of the present invention, comparing the Raman spectrum of the collected item to be inspected with the standard Raman spectrum of the item indicated in the goods list stored in the standard database To determine whether the two match; wherein the identification algorithm of the embodiment of the present invention will be described in detail below;
  • the data may be saved and the inspection is ended; if the item to be inspected and the item indicated in the manifest are determined If there is no match, continue with the following steps;
  • Chemical substance identification and determination step using a conventional Raman spectroscopy method to detect the actual composition of the item to be inspected to identify the chemical substance contained in the item to be inspected, and then, to identify the chemical substance and the above-mentioned standard database The chemical substances of the items indicated in the goods list are compared to determine whether the two match;
  • the chemical substance identification and determination step in the case where the chemical substance of the article to be inspected can be identified, if it is determined that the item to be inspected is consistent with the information of the item indicated in the manifest, then the data can be saved, and the inspection is ended. If it is determined that the information of the item to be inspected is inconsistent with the item indicated in the goods list, subsequent processing, such as manual unpacking and inspection, is required; if the chemical substance of the item to be inspected cannot be identified, it needs to be checked again. The item is sampled and the collected data is saved for subsequent processing.
  • the Raman spectroscopy technique is combined with the similarity metric algorithm and the support vector machine improvement algorithm.
  • the similarity measure of the article to be inspected and the standard article is calculated by calculating the correlation coefficient;
  • an improved support vector machine algorithm is proposed, which reduces the dimensionality of the high-dimensional vector, and classifies the items to be inspected whose correlation coefficient is lower than the threshold to avoid the Raman spectroscopy equipment.
  • the resulting noise affects the recognition results and improves recognition speed and accuracy.
  • the algorithm model used in the method is described before describing the specific method.
  • the Raman spectral recognition technology is an application technology for classifying and identifying objects to be inspected.
  • key information capable of reflecting the composition of the substance is obtained, and the information contained in the spectral signal is extracted.
  • the spectral information is classified and identified according to the difference of the spectral information, wherein the similarity measurement method is used to calculate the spectral information difference.
  • the spectral preprocessing generally includes denoising, baseline correction, normalization processing, etc.
  • the original spectrum obtained by the acquisition generally needs to be pre-processed, and for the sake of brevity, it will not be described one by one below.
  • the Correlation coefficient is a measure of the degree of linear correlation between variables and is a measure of the relationship between vectors.
  • the feature vectors X(x 1 , x 2 , . . . , x n ), Y(y 1 , y 2 , . . . , y n ) are provided, and the correlation coefficients of the two are defined as follows:
  • n represents the length of the feature vector X
  • Y Represents the mean of the vectors X and Y, respectively
  • i represents the ith data of the vector.
  • the correlation coefficient is selected as the judgment basis of the similarity measure, which can avoid the loss of information by the Euclidean distance and the amplification of the Mahalanobis distance to the small deviation, so that the two can be better judged. Similarity between feature vectors.
  • Raman spectroscopy the Raman spectrum is biased due to differences in sample uniformity, instrument noise, fluorescence background, etc. Denoising, baseline correction, etc., also produce errors during spectral processing. In the identification process, only the correlation coefficient is used to identify the characteristics of the material. The accuracy of the Raman spectrum for the entry and exit special items in the field of inspection and quarantine is introduced. Or the item to be inspected slightly below the threshold for sorting the item.
  • the support vector machine is a two-class model whose basic model is defined as the linear classifier with the largest interval in the feature space. The principle is shown in Figure 2.
  • the geometric meaning of the support vector machine in the linear classifier is that the distance Gap between the hyperplanes l 1 and l 2 takes the maximum value.
  • the optimization function of the linear severable support vector machine is as follows:
  • the Raman spectrum is preprocessed to obtain a high-dimensional vector to represent the substance to be identified, the training learning time process is long in the process of directly using the high-dimensional vector for recognition and classification.
  • the Raman spectrum is firstly subjected to a certain sparse transformation.
  • the target vector c for sparse representation of the Raman spectral information y is calculated using the following objective function:
  • N is the length of the vector c and i is the number of the vector c.
  • the minimum target vector c is taken to represent the spectral information y using the dimensionally reduced target vector c, thereby achieving a sparse representation of the Raman spectrum.
  • the establishment process of the classifier mainly includes three steps, taking the two classifications as an example, and the specific establishment process is as shown in FIG. 3.
  • the appropriate sample is selected as the training sample, and the sample is measured to obtain the Raman spectrum.
  • the test sample data is obtained.
  • the test sample is divided into a positive sample and a negative sample, wherein the positive sample is the spectral information of a certain object to be tested, and the negative sample is the spectral information of the non-test object.
  • the SVM classifier is obtained by sparse representation of the spectrum and using the support vector machine to build the model.
  • a Raman spectroscopy method for item inspection is further described below with reference to FIG. 4, which may include the following steps:
  • Raman spectroscopy acquisition step collecting the Raman spectrum of the item to be inspected.
  • Alignment and determination step comparing the Raman spectrum of the collected item to be inspected with the Raman spectrum of the standard item stored in the standard database to determine whether the item to be inspected matches the standard item.
  • the comparing and determining step specifically includes: calculating a correlation coefficient between a feature vector of a Raman spectrum of the item to be inspected and a feature vector of a Raman spectrum of the standard item stored in the standard database, when the calculated correlation coefficient When the threshold is greater than the preset threshold, it is determined that the item to be inspected matches the standard item.
  • the comparing and determining step further includes: when the calculated correlation coefficient is equal to or lower than a preset threshold, classifying a Raman spectrum of the item to be inspected by using a support vector machine to determine whether the item to be inspected is Standard items match.
  • the step of classifying the Raman spectrum of the item to be inspected by using the support vector machine includes:
  • Establishing a classifier step selecting a training sample, measuring the training sample to obtain a Raman spectrum of the training sample, performing a sparse transform on the Raman spectrum of the training sample, and then constructing a classifier by using a support vector machine algorithm;
  • Classification step The Raman spectrum of the item to be inspected is classified using the established classifier.
  • the Raman spectrum of the training sample can be sparsely changed by the objective function defined by the above formula (6) to obtain a sparse representation of the Raman spectrum to reduce training learning. Time, which makes training and testing faster.
  • Table 1 shows the statistic table of its kind.
  • the instrument used in the experiment was the RT6000 handheld Raman spectrometer of Tongfang Weishi Technology Co., Ltd., with an excitation wavelength of 785 nm; a resolution of 6-9 cm -1 ; and a wavenumber range of 200-3200 cm -1 .
  • the accuracy verification of the improved support vector machine includes two aspects.
  • the substance whose recognition result is the same as the real value is judged as a pass, and the substance different from the real value is judged as a failure.
  • the acquired Raman spectrum is subjected to pre-processing such as baseline correction, denoising, normalization, etc., and the obtained Raman spectrum is compared with the sample in the standard library. If the correlation coefficient is greater than the threshold, it is determined as Matching; Raman spectroscopy equal to or less than the threshold is classified by an improved support vector machine. If the classification result indicates that the two match, it is a match, otherwise, it is a mismatch.
  • test samples were analyzed using an improved support vector machine with a total of six classes of substances, including blood products, viruses, antibodies, and the like.
  • the range of discrimination for various material standard databases is 0.7460-0.7822, and the average discrimination between classes is 0.7623.
  • the matching rate of the conformity check of the substance using the similarity measure, the support vector machine and the improved support vector machine algorithm is counted as shown in Table 2.
  • the correct matching rate using the three methods is shown in Table 2.
  • the comparison shows that when the lower threshold is selected, the correlation coefficient can be used to obtain better detection results.
  • the matching rate is gradually reduced.
  • the matching rate of the items to be inspected is higher than the direct similarity measure. Therefore, increasing the support vector machine classification decision process for the less similar items to be tested can reduce the possibility of missed detection.
  • the improved support vector is improved by the improved sparse expression of the support vector machine algorithm. The machine algorithm obtained the best results during the inspection process.
  • Raman spectroscopy detecting method for item inspection which can directly perform the above ratio using an improved support vector machine algorithm without performing similarity metrics. Pair and decision steps.
  • the method may include the following steps:
  • Raman spectroscopy acquisition step collecting the Raman spectrum of the item to be inspected.
  • the step of classifying the Raman spectrum of the item to be inspected by using the support vector machine includes:
  • Establishing a classifier step selecting a training sample, measuring the training sample to obtain a Raman spectrum of the training sample, performing a sparse transform on the Raman spectrum of the training sample, and then constructing a classifier by using a support vector machine algorithm;
  • Classification step The Raman spectrum of the item to be inspected is classified using the established classifier.
  • the Raman spectrum of the training sample can be sparsely changed by using the objective function defined by the above formula (6) to obtain a sparse representation of the Raman spectrum, Reduce training time.
  • the Raman spectrum of all the items to be tested is classified and identified using the improved support vector machine algorithm, without distinguishing whether the correlation coefficient is greater than or equal to the threshold.
  • the present embodiment may include other steps that are the same as the above-described embodiments, including, for example, various pre-processing steps and the like.
  • FIG. 5 is a block diagram showing an example hardware arrangement of the electronic device 500.
  • the electronic device 500 includes a processor 506 (eg, a microprocessor ( ⁇ P), a digital signal processor (DSP), etc.).
  • processor 506 can be or include a single processing unit or a plurality of processing units for performing different acts of the method steps described herein.
  • the electronic device 500 may also include an input unit 502 for receiving signals from other entities, and an output unit 504 for providing signals to other entities.
  • Input unit 502 and output unit 504 can be arranged as a single entity or as separate entities.
  • electronic device 500 can include at least one computer readable storage medium 508 in the form of a non-volatile or volatile memory, such as an electrically erasable programmable read only memory (EEPROM), flash memory, and/or a hard drive.
  • the computer readable storage medium 508 includes a computer program 510 that includes code/computer readable instructions that, when executed by the processor 506 in the electronic device 500, cause the electronic device 500 to perform, for example, in connection with Figures 1 - 3 above. 4 described process and any variations thereof.
  • Computer program 510 can be configured to have an architecture such as computer program modules 510A-510C Computer program code.
  • the computer program module can essentially perform the various actions in the flows illustrated in Figures 1, 3-4 to simulate a device. In other words, when different computer program modules are executed in processor 506, they may correspond to the different units described above in the device.
  • code means in the embodiment disclosed above in connection with FIG. 5 is implemented as a computer program module that, when executed in processor 506, causes electronic device 500 to perform the actions described above in connection with FIGS. 1-4, however in alternative implementations In an example, at least one of the code means can be implemented at least partially as a hardware circuit.
  • the processor may be a single CPU (Central Processing Unit), but may also include two or more processing units.
  • a processor can include a general purpose microprocessor, an instruction set processor, and/or a related chipset and/or a special purpose microprocessor (eg, an application specific integrated circuit (ASIC)).
  • the processor may also include an onboard memory for caching purposes.
  • the computer program can be carried by a computer program product connected to the processor.
  • the computer program product can comprise a computer readable medium having stored thereon a computer program.
  • the computer program product can be a flash memory, a random access memory (RAM), a read only memory (ROM), an EEPROM, and the computer program modules described above can be distributed to different computer program products in the form of memory in alternative embodiments. .

Landscapes

  • Health & Medical Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

用于物品查验的拉曼光谱检测方法。该方法包括以下步骤 : 拉曼光谱采集步骤 : 采集待检物品的拉曼光谱; 和比对和判定步骤 : 将采集的待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱作比对, 以判定待检物品是否与标准物品匹配。所述比对和判定步骤包括 : 采用支持向量机对待检物品的拉曼光谱进行分类, 以实现待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱的比对。一种用于物品查验的电子设备 (500), 包括处理器 (506), 用于从其他实体接收信号的输入单元 (502), 以及用于向其他实体提供信号的输出单元 (504), 至少一个计算机可读存储介质 (508), 该计算机可读介质 (508) 包括计算机程序 (510), 该计算机程序 (510) 可被配制为具有例如计算机程序模块 (510A-510C) 等架构的计算机程序代码。

Description

用于物品查验的拉曼光谱检测方法和电子设备 技术领域
本发明一般地涉及拉曼光谱检测领域，特别地，涉及用于物品查验的拉曼光谱检测方法，尤其涉及用于出入境特殊物品快速查验的拉曼光谱检测方法。
背景技术
拉曼光谱是一种分子振动光谱，它可以反映分子的指纹特征，可用于对物质的检测。拉曼光谱检测通过检测待测物对于激发光的拉曼散射效应所产生的拉曼光谱来检测和识别物质。拉曼光谱检测方法已经广泛应用于液体安检、珠宝检测、爆炸物检测、毒品检测、药品检测、农药残留检测等领域。
现有的针对出入境特殊物品的现场检测，主要采用现场高清摄像装置的实时监控与监督作用，现场查验人员开箱检测，比对货物信息与申报单信息是否一致，若货物信息一致，则认为货物可放行。
然而，上述方法主要存在以下缺点：(1)现场查验人员工作量大。开箱核对审批单与标签信息是否一致需要现场查验工作人员逐一核对标签信息，人工比对复杂的货物品名、规格、性状及数量等多种信息(大部分为英语等非中文语种信息)是否一致。(2)查验漏洞明显。核对标签信息的查验方式无法验证货物标签信息与实际内容物是否一致，无法有效识别采用低风险或非特殊物品包装储存或运输高风险特殊物品的情况。
发明内容
为了克服或消除现有技术存在的问题和缺陷中的至少一种，提出了本发明。
本发明的至少一个目的是提供用于物品查验的拉曼光谱检测方法和电子设备，尤其涉及用于出入境特殊物品快速查验的拉曼光谱检测方法，其能够增强物品现场查验的准确性，并且可以通过一次检测实现多重信息的同时匹配查验，加快物品的现场查验效率。
根据本发明的一个方面，提供一种用于物品查验的拉曼光谱检测方法，包括以下步骤：
拉曼光谱采集步骤：采集待检物品的拉曼光谱；和
比对和判定步骤：将采集的待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱作比对，以判定待检物品是否与标准物品匹配，
所述的比对步骤包括：采用支持向量机对待检物品的拉曼光谱进行分类，以实现待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱的比对。
根据一些实施例，所述采用支持向量机对待检物品的拉曼光谱进行分类的步骤包括：
选取训练样本，对训练样本进行测量以得到训练样本的拉曼光谱，对训练样本的拉曼光谱进行稀疏变换，然后采用支持向量机算法建立分类器；和
采用建立的分类器，对待检物品的拉曼光谱进行分类。
根据一些实施例，对训练样本的拉曼光谱进行稀疏变换包括如下步骤：
获取训练样本的拉曼光谱信息y的主成分M；
利用公式y＝Mc对训练样本的拉曼光谱信息y进行稀疏表示，得到一系列的稀疏表示的矢量ci；和
利用下列目标函数计算出对拉曼光谱信息y进行稀疏表示的目标矢量c：
Figure PCTCN2017111624-appb-000001
其中，矢量c中的非零元素个数固定为k，N为矢量c的长度，k≤N，i为矢量c的序号。
根据一些实施例，所述比对和判定步骤还包括：
对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行相似性度量；
当相似性度量的结果大于预设的阈值时，判定待测物品与标准物品匹配；
当相似性度量的结果等于或低于预设的阈值时，采用支持向量机对待检物品的拉曼光谱进行分类，以实现待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱的比对。
根据一些实施例，所述的对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行相似性度量包括：
计算待检物品的拉曼光谱的特征向量与标准数据库中存储的标准物品的拉曼光谱的特征向量的相关系数，将计算出的相关系数作为相似性度量的结果。
根据一些实施例，所述拉曼光谱检测方法还包括如下步骤：
建立标准数据库步骤：先后采集标准物品的表面包装和实际内容物的拉曼光谱，以形成包含标准物品的表面成分的拉曼光谱和实际内容物的成分的拉曼光谱的标准数据库。
根据一些实施例，所述拉曼光谱采集步骤包括：对待检物品进行表面和内部扫描，以采集待检物品的表面成分的拉曼光谱和实际内容物的成分的拉曼光谱；
所述对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行相似性度量的步骤包括：分别对待检物品的表面成分的拉曼光谱与标准数据库中存储的标准物品的表面成分的拉曼光谱以及待检物品的实际内容物的成分的拉曼光谱与标准数据库中存储的标准物品的实际内容物的成分的拉曼光谱进行相似性度量；并且
所述相似性度量的结果大于预设的阈值仅包括两个相似性度量的结果均大于预设的阈值的情形。
根据一些实施例，所述建立标准数据库步骤还包括：采集标准物品的品名、来源公司、规格大小和图片信息，并将这些信息存入标准数据库中。
根据本发明的另一方面，还提供一种电子设备，包括：
存储器，用于存储可执行指令；以及
处理器，用于执行存储器中存储的可执行指令，以执行本发明任一方面或实施例中所述的用于物品查验的拉曼光谱检测方法。
根据本发明实施例的检测方法免去了现场查验人员逐一比对待检物品的多项信息的过程，加快了现场查验速度；同时采用特征信息匹配的识别方法，提高了现场查验的准确度。
附图说明
图1示出了根据本发明实施例的用于出入境特殊物品查验的检测方法的流程图；
图2为支持向量机原理的示意图；
图3示出了建立支持向量机分类器的流程图；
图4示出了根据本发明实施例的采用改进支持向量机的检测方法的流程图；和
图5示出了用于执行根据本发明实施例的检测方法的电子设备的示例硬 件布置的框图。
具体实施方式
下面通过实施例，并结合附图，对本发明的技术方案作进一步具体的说明。在说明书中，相同或相似的附图标号表示相同或相似的部件。下述参照附图对本发明实施方式的说明旨在对本发明的总体发明构思进行解释，而不应当理解为对本发明的一种限制。
在本文中，为了描述方便，使用“第一、第二”、“A、B、C”等表述描述方法的步骤，但是，除非有特别说明，这样的表述不应理解为对步骤执行顺序的限制。
在下文中，主要以出入境特殊物品为例对本发明的技术构思进行详细说明，一般地，出入境特殊物品指微生物、人体组织、人类遗传资源、生物制品、血液、血液制品等。由于检验检疫查验现场特殊物品种类繁多、货源多样、包装复杂，同时很多特殊物品具有样品量少、肉眼难以观察、成分不明的特点；另外，特殊物品的流通相比于一般货物，其运输难度和潜在风险更大；对于现场查验，很多特殊物品如微生物、血液等生物制剂均无法实现查验现场的开封检测。这些原因共同决定了出入境特殊物品的检测成为检验检疫查验工作难点。而采用根据本发明实施例的用于物品查验的拉曼光谱检测方法，可以对出入境特殊物品进行快速查验，同时还能提高现场查验的准确度。
根据本发明的实施例，首先建立一个比较完备的出入境特殊物品的整体拉曼谱图信息库，该信息库可以称为标准数据库。然后，在现场查验过程中，根据货物单上表明的物品名，选择已存在于标准数据库中的标准物品，之后对待检物品进行表面和内部扫描，然后对扫描得到的待检物品的谱图与标准数据库中的标准谱图进行比对分析，结合“相似性度量”算法和“改进支持向量机”算法，完成对出入境特殊物品的快速特征识别。
在建立出入境特殊物品标准数据库时，可以先将代表性的出入境特殊物品的标准谱图建立起来，也可在使用过程中累积添加，或由检验检疫查验单位提供。在一个示例中，针对每种出入境特殊物品，在没有其他物体干扰的情况下，可以在可看见该特殊物品的内容物的位点处先后采集其表面包装和实际内容物的拉曼特征谱图，然后整理建立拉曼谱图信息库，以形成该标准数据库。
根据本发明的一个实施例，标准数据库中保存着出入境特殊物品的整体信息。一方面，在对标准数据库中的标准物品命名时，标明标准物品的品名、来源公司、规格大小、图片等信息；另一方面，每种出入境特殊物品均存有对应的两个谱图，一个谱图是对其特定位点的表面进行谱图采集，所采集到的是表面成分的谱图信息，另一个是对同一位点的内部进行谱图采集，所采集到的是其内部实际成分的谱图信息。在待检物品扫描比对时，需将表面采集到的谱图信息与谱图库中的表面标准谱进行比对，将采集到的内部谱图信息与谱图库中的内部标准谱进行比对，两次均超过阈值才认为该待检物品能与标准物品属同一物品，增大匹配准确性。
图1示出了根据本发明实施例的用于出入境特殊物品查验的检测方法的流程图，下面结合图1对该检测方法进行详细说明，根据本发明实施例的用于出入境特殊物品查验的检测方法主要包括如下步骤：
拉曼光谱采集步骤：在检测设备上选择待检物品，扫描待检物品以采集待检物品的拉曼光谱；
拉曼光谱比对与判定步骤：采用根据本发明实施例的识别算法，将采集的待检物品的拉曼光谱与上述标准数据库中存储的货物单中标明的物品的标准拉曼光谱进行比对，以判定二者是否匹配；其中，本发明实施例的识别算法将在下文中详细说明；
在上述拉曼光谱比对与判定步骤中，如果判定待检物品与货物单中标明的物品相匹配，则可以保存数据，并且结束本次查验；如果判定待检物品与货物单中标明的物品不匹配，则继续执行下面的步骤；
化学物质识别与判定步骤：采用常规的拉曼光谱识别方法，检测待检物品的实际成分，以识别出待检物品包含的化学物质，然后，将识别出的化学物质与上述标准数据库中存储的货物单中标明的物品的化学物质进行比对，以判定二者是否匹配；
在上述化学物质识别与判定步骤中，在能够识别出待检物品的化学物质的情况下，如果判定待检物品与货物单中标明的物品的信息一致，那么就可以保存数据，结束本次查验；如果判定待检物品与货物单中标明的物品的信息不一致，那么需要进行后续处理，例如人工开箱查验等操作；在不能够识别出待检物品的化学物质的情况下，需要再次对待检物品进行采样检查，然后将采集的数据保存，以进行后续处理。
在上述方法中的拉曼光谱比对与判定步骤中，需要采用根据本发明实施例的识别算法，将采集的待检物品的拉曼光谱与上述标准数据库中存储的货物单中标明的物品的标准拉曼光谱进行比对，以判定二者是否匹配，下面将结合附图详细说明根据本发明实施例的识别算法。
根据本发明的一个实施例，采用拉曼光谱技术结合相似性度量算法和支持向量机改进算法，首先，通过计算相关系数对待检物品与标准物品进行相似性度量；其次，当待检物品与标准物品的相关系数低于阈值时，针对性地提出了改进的支持向量机算法，对高维向量进行降维处理，并且对相关系数低于阈值的待检物品进行分类，以避免拉曼光谱设备产生的噪音对识别结果的影响，并且提升识别速度和准确性。在描述具体的方法之前，对该方法中所使用的算法模型进行说明。
[算法模型]
(1)相似性度量
拉曼光谱识别技术是对待检物品进行分类和识别的应用技术，根据本发明的一个实施例，完成光谱预处理及特征提取后，得到能够反映物质组成的关键信息，提取光谱信号中所包含的光谱信息，按照光谱信息差异对待检测物品进行分类和识别，其中，采用相似性度量方法来计算光谱信息差异。
需要说明的是，光谱预处理一般包括去噪、基线校正、归一化处理等，在本文中，采集获得的原始光谱一般需要经过预处理，为了简洁，下文不再逐一赘述。
相关系数(Correlation coefficient)是研究变量间线性相关程度的量，是一种衡量向量间相互关系的方法。设有特征向量X(x1，x2，...，xn)，Y(y1，y2，...，yn)，二者的相关系数定义如下：
Figure PCTCN2017111624-appb-000002
其中，n表示特征向量X、Y的长度，
Figure PCTCN2017111624-appb-000003
分别表示向量X、Y的均值，i表示向量的第i个数据。
在本实施例中，选用相关系数作为相似性度量的判断依据，可以避免欧氏距离对信息的丢失和马氏距离对微小偏差的放大作用，从而能够较好地判定两 个特征向量之间的相似性。
(2)支持向量机(SVM)
在拉曼光谱测量中，由于存在样品均匀性差异、仪器噪声、荧光背景等，使得拉曼光谱产生偏差；在光谱处理过程中，去噪、基线校正等也会产生误差。在识别过程中仅采用相关系数进行物质的特征识别的准确率不高，因此，在针对检验检疫领域的出入境特殊物品的拉曼光谱进行特征识别过程中，引入支持向量机对等于和低于或略低于阈值的待检物品进行物品分类。
支持向量机是一个二分类模型，其基本模型定义为特征空间上间隔最大的线性分类器。其原理如图2所示。
设训练样本集X为xi，i＝1，2，...，N，样本分属两类，w1和w2，且线性可分。线性判别函数l的一般形式为：g(x)＝ω·x+b，但是这样的超平面并不唯一。因此，支持向量机的分类识别问题转化为寻找最大间隔的分类超平面问题。
选择合适的ω，b，使得样本类别w1和w2分别满足g(x)＝1和g(x)＝-1，即如下式：
Figure PCTCN2017111624-appb-000004
Figure PCTCN2017111624-appb-000005
支持向量机在线性分类器中的几何含义，就是在超平面l1和l2间的距离Gap取得最大值。在该过程中，对距离超平面l0最近的样本进行归一化处理，这样Gap/2＝2/||ω||，则有线性可分类支持向量机的优化函数如下式所示：
Figure PCTCN2017111624-appb-000006
根据拉格朗日(Lagrange)乘子法及Karush-Kuhn-Tucker条件(KKT条件)可得下式，其中λ为Lagrange乘子。
Figure PCTCN2017111624-appb-000007
(3)改进支持向量机
由于拉曼光谱在经过预处理后，得到高维向量用以表征待识别物质，在直接采用高维向量进行识别分类的过程中，训练学习时间过程较长。
为了改变这一问题，在采用支持向量机进行识别的过程中，首先对于拉曼光谱进行一定的稀疏变换，具体运算包括两步：一、获取训练样本的拉曼光谱信息y的主成分M；二、利用公式y＝Mc对训练样本的拉曼光谱信息y进行稀疏表示，得到一系列的稀疏表示的矢量ci，且矢量c中的非零元素个数固定为k，k≤N；并且利用下列目标函数计算出对拉曼光谱信息y进行稀疏表示的目标矢量c：
Figure PCTCN2017111624-appb-000008
其中，N为矢量c的长度，i为矢量c的序号。
通过使用上述目标函数，寻找使得
Figure PCTCN2017111624-appb-000009
取最小值的目标矢量c，以使用降维后的目标矢量c表示光谱信息y，从而实现拉曼光谱的稀疏表示。
在选用未改进的支持向量进行物质分类的过程中，由于光谱信息的高维性和大部分信息的非零性，需要采用RBF核函数、多项式核函数等映射到更高维进行分类处理；改进后的支持向量机实现了对光谱稀疏表达，增强了光谱信息的可区分性，光谱分类可采用线性核支持向量机实现，使训练和测试速度更快，且所需的存储空间更少，在训练学习过程中减少时间。
分类器的建立过程主要包括三个步骤，以二分类为例，具体建立流程如图3所示。
选取适量样本作为训练样本，对样本进行测量以得到拉曼光谱，经过基线校正、去噪、归一化等预处理后得到测试样本数据。将测试样本分为正样本和负样本，其中正样本即为某种待测物品的光谱信息，负样本为非待测物品的光谱信息。通过对光谱进行稀疏表示，采用支持向量机进行模型建立，得到SVM分类器。
[方法]
基于上面介绍的计算模型，下面结合图4，进一步描述根据本发明实施例的用于物品查验的拉曼光谱检测方法，该方法可以包括如下步骤：
拉曼光谱采集步骤：采集待检物品的拉曼光谱；和
比对和判定步骤：将采集的待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱作比对，以判定待检物品是否与标准物品匹配。
根据一个实施例，上述比对和判定步骤具体包括：计算待检物品的拉曼光谱的特征向量与标准数据库中存储的标准物品的拉曼光谱的特征向量的相关系数，当计算出的相关系数大于预设的阈值时，判定待检物品与标准物品匹配。
进一步地，所述比对和判定步骤还包括：当计算出的相关系数等于或低于预设的阈值时，采用支持向量机对待检物品的拉曼光谱进行分类，以判定待检物品是否与标准物品匹配。
具体地，所述采用支持向量机对待检物品的拉曼光谱进行分类的步骤包括：
建立分类器步骤：选取训练样本，对训练样本进行测量以得到训练样本的拉曼光谱，对训练样本的拉曼光谱进行稀疏变换，然后采用支持向量机算法建立分类器；和
分类步骤：采用建立的分类器，对待检物品的拉曼光谱进行分类。
在对训练样本的拉曼光谱进行稀疏变换的过程中，可以采用上述公式(6)定义的目标函数对训练样本的拉曼光谱进行稀疏变化，以得到拉曼光谱的稀疏表示，以减少训练学习时间，从而使得训练和测试速度更快。
[实验验证]
下面，对根据本发明实施例的用于物品查验的拉曼光谱检测方法进行实验验证。
(1)实验条件和方法
实验随机抽取北京市出入境检验检疫局日常查验过程中的380种入境特殊物品进行拉曼谱图采集和支持向量机算法验证实验。表1表示其种类统计表。
表1 380种出入境特殊物品物质种类统计表
Figure PCTCN2017111624-appb-000010
实验所用仪器为同方威视技术股份有限公司RT6000手持式拉曼光谱仪，激发波长为785nm；分辨率为6～9cm-1；波数范围为200～3200cm-1。
实验所采集到的拉曼光谱图均进行降噪、基线矫正和归一化预处理操作，所有的谱图分析与对比验证均通过Matlab(R2010a)及C++编程方法获得。
改进支持向量机的准确性验证包括两个方面，对于识别结果与真实值相同的物质判断为正确(pass)，与真实值不同的物质判别为错误(fail)。应用测试样本对模型的准确性进行验证，判定算法模型的准确性。具体流程如图4所示。
实验中，对所采集的拉曼光谱进行基线校正、去噪、归一化等预处理，得到的拉曼光谱与标准谱库中的样本进行相似性比较，若相关系数大于阈值，则判定为匹配；对小于等于阈值的拉曼光谱采用改进的支持向量机进行分类，若分类结果表示二者匹配，即为匹配，反之，则为不匹配。
(2)实验结果
运用改进的支持向量机对所有的测试样本进行分析，其中共有6类物质，包括血液制品、病毒、抗体等。各类物质标准数据库的区分度的范围为0.7460-0.7822，类间的平均区分度为0.7623。对采用相似性度量、支持向量机与改进支持向量机算法进行物质的符合性查验的匹配率进行统计，如表2所示。
表2测试结果列表
Figure PCTCN2017111624-appb-000011
在符合性查验过程中，采用三种方法的正确匹配率如表2所示。在依次选取阈值为0.85、0.86、0.87、0.88、0.89、0.90进行查验过程中，对比可知：在选取较低的阈值时，采用相关系数也能得到较好的查验结果，随着阈值进一步增加，匹配率逐渐下降；采用支持向量机与改进支持向量机算法进行查验的过程中，待检物品的匹配率均高于直接采用相似性度量。因此，对于相似性较低的待检物品增加支持向量机分类判定过程，可以降低漏检的可能性，同时，由于改进支持向量机算法的稀疏表达提高了分类器的准确性，使得改进支持向量机算法在查验过程中得到了最优结果。
根据本发明的另一实施例，还提供一种用于物品查验的拉曼光谱检测方法，该方法可以不进行相似性度量，而直接使用改进支持向量机算法进行上述的比 对和判定步骤。
具体地，该方法可以包括如下步骤：
拉曼光谱采集步骤：采集待检物品的拉曼光谱；和
比对和判定步骤：直接采用支持向量机对待检物品的拉曼光谱进行分类，以判定待检物品是否与标准物品匹配。
具体地，所述采用支持向量机对待检物品的拉曼光谱进行分类的步骤包括：
建立分类器步骤：选取训练样本，对训练样本进行测量以得到训练样本的拉曼光谱，对训练样本的拉曼光谱进行稀疏变换，然后采用支持向量机算法建立分类器；和
分类步骤：采用建立的分类器，对待检物品的拉曼光谱进行分类。
同样地，在对训练样本的拉曼光谱进行稀疏变换的过程中，可以采用上述公式(6)定义的目标函数对训练样本的拉曼光谱进行稀疏变化，以得到拉曼光谱的稀疏表示，以减少训练学习时间。
也就是说，在该实施例中，使用改进支持向量机算法对全部待测物品的拉曼光谱进行分类识别，而不区分其相关系数是大于还是小于等于阈值。应该理解，除了该区别之外，本实施例可以包括与上述实施例相同的其它步骤，例如包括各种预处理步骤等。
根据本发明的又一实施例，还提供一种电子设备，图5是示出了该电子设备500的示例硬件布置的框图。电子设备500包括处理器506(例如，微处理器(μP)、数字信号处理器(DSP)等)。处理器506可以是或包括用于执行本文描述的方法步骤的不同动作的单一处理单元或者是多个处理单元。电子设备500还可以包括用于从其他实体接收信号的输入单元502、以及用于向其他实体提供信号的输出单元504。输入单元502和输出单元504可以被布置为单一实体或者是分离的实体。
此外，电子设备500可以包括具有非易失性或易失性存储器形式的至少一个计算机可读存储介质508，例如是电可擦除可编程只读存储器(EEPROM)、闪存、和/或硬盘驱动器。计算机可读存储介质508包括计算机程序510，该计算机程序510包括代码/计算机可读指令，其在由电子设备500中的处理器506执行时使得电子设备500可以执行例如上面结合图1、3-4所描述的流程及其任何变形。
计算机程序510可被配置为具有例如计算机程序模块510A～510C等架构 的计算机程序代码。计算机程序模块实质上可以执行图1、3-4中所示出的流程中的各个动作，以模拟设备。换言之，当在处理器506中执行不同计算机程序模块时，它们可以对应于设备中的上述不同单元。
尽管上面结合图5所公开的实施例中的代码手段被实现为计算机程序模块，其在处理器506中执行时使得电子设备500执行上面结合图1～4所描述的动作，然而在备选实施例中，该代码手段中的至少一项可以至少被部分地实现为硬件电路。
处理器可以是单个CPU(中央处理单元)，但也可以包括两个或更多个处理单元。例如，处理器可以包括通用微处理器、指令集处理器和/或相关芯片组和/或专用微处理器(例如，专用集成电路(ASIC))。处理器还可以包括用于缓存用途的板载存储器。计算机程序可以由连接到处理器的计算机程序产品来承载。计算机程序产品可以包括其上存储有计算机程序的计算机可读介质。例如，计算机程序产品可以是闪存、随机存取存储器(RAM)、只读存储器(ROM)、EEPROM，且上述计算机程序模块在备选实施例中可以用存储器的形式被分布到不同计算机程序产品中。
本领域技术人员应当理解，在本发明的实施例中，虽然以出入境特殊物品的查验为示例详细说明了本发明的技术构思，但是本发明不局限于查验出入境特殊物品，还可以适用于例如药物快速检测、生物样品筛查等领域。
虽然结合附图对本发明进行了说明，但是附图中公开的实施例旨在对本发明优选实施方式进行示例性说明，而不能理解为对本发明的一种限制。
虽然本发明总体构思的一些实施例已被显示和说明，本领域普通技术人员将理解，在不背离本总体发明构思的原则和精神的情况下，可对这些实施例做出改变，本发明的范围以权利要求和它们的等同物限定。

Claims (9)

  1. 一种用于物品查验的拉曼光谱检测方法，包括以下步骤：
    拉曼光谱采集步骤：采集待检物品的拉曼光谱；和
    比对和判定步骤：将采集的待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱作比对，以判定待检物品是否与标准物品匹配，
    其中，所述的比对步骤包括：采用支持向量机对待检物品的拉曼光谱进行分类，以实现待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱的比对。
  2. 根据权利要求1所述的拉曼光谱检测方法，其中，所述采用支持向量机对待检物品的拉曼光谱进行分类的步骤包括：
    选取训练样本，对训练样本进行测量以得到训练样本的拉曼光谱，对训练样本的拉曼光谱进行稀疏变换，然后采用支持向量机算法建立分类器；和
    采用建立的分类器，对待检物品的拉曼光谱进行分类。
  3. 根据权利要求2所述的拉曼光谱检测方法，其中，对训练样本的拉曼光谱进行稀疏变换包括如下步骤：
    获取训练样本的拉曼光谱信息y的主成分M；
    利用公式y＝Mc对训练样本的拉曼光谱信息y进行稀疏表示，得到一系列的稀疏表示的矢量ci；和
    利用下列目标函数计算出对拉曼光谱信息y进行稀疏表示的目标矢量c：
    Figure PCTCN2017111624-appb-100001
    其中，矢量c中的非零元素个数固定为k，N为矢量c的长度，k≤N，i为矢量c的序号。
  4. 根据权利要求1-3中任一项所述的拉曼光谱检测方法，其中，所述比对和判定步骤还包括：
    对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行 相似性度量；
    当相似性度量的结果大于预设的阈值时，判定待测物品与标准物品匹配；
    当相似性度量的结果等于或低于预设的阈值时，采用支持向量机对待检物品的拉曼光谱进行分类，以实现待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱的比对。
  5. 根据权利要求4所述的拉曼光谱检测方法，其中，所述的对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行相似性度量包括：
    计算待检物品的拉曼光谱的特征向量与标准数据库中存储的标准物品的拉曼光谱的特征向量的相关系数，将计算出的相关系数作为相似性度量的结果。
  6. 根据权利要求1-5中任一项所述的拉曼光谱检测方法，其中，还包括如下步骤：
    建立标准数据库步骤：先后采集标准物品的表面包装和实际内容物的拉曼光谱，以形成包含标准物品的表面成分的拉曼光谱和实际内容物的成分的拉曼光谱的标准数据库。
  7. 根据权利要求6所述的拉曼光谱检测方法，其中，
    所述拉曼光谱采集步骤包括：对待检物品进行表面和内部扫描，以采集待检物品的表面成分的拉曼光谱和实际内容物的成分的拉曼光谱；
    所述对待检物品的拉曼光谱与标准数据库中存储的标准物品的拉曼光谱进行相似性度量的步骤包括：分别对待检物品的表面成分的拉曼光谱与标准数据库中存储的标准物品的表面成分的拉曼光谱以及待检物品的实际内容物的成分的拉曼光谱与标准数据库中存储的标准物品的实际内容物的成分的拉曼光谱进行相似性度量；并且
    所述相似性度量的结果大于预设的阈值仅包括两个相似性度量的结果均大于预设的阈值的情形。
  8. 根据权利要求6或7所述的拉曼光谱检测方法，其中，所述建立标准数据库步骤还包括：采集标准物品的品名、来源公司、规格大小和图片信息，并将这些信息存入标准数据库中。
  9. 一种电子设备，包括：
    存储器，用于存储可执行指令；以及
    处理器，用于执行存储器中存储的可执行指令，以执行如权利要求1-8中任一项所述的方法。
PCT/CN2017/111624 2016-12-29 2017-11-17 用于物品查验的拉曼光谱检测方法和电子设备 WO2018121122A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611257547.5 2016-12-29
CN201611257547.5A CN108254351B (zh) 2016-12-29 2016-12-29 用于物品查验的拉曼光谱检测方法

Publications (1)

Publication Number Publication Date
WO2018121122A1 true WO2018121122A1 (zh) 2018-07-05

Family

ID=62706829

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/111624 WO2018121122A1 (zh) 2016-12-29 2017-11-17 用于物品查验的拉曼光谱检测方法和电子设备

Country Status (2)

Country Link
CN (1) CN108254351B (zh)
WO (1) WO2018121122A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111141719A (zh) * 2019-12-27 2020-05-12 梧州市食品药品检验所 抗艾滋病药物的快速无损鉴别方法
CN111426648A (zh) * 2020-03-19 2020-07-17 甘肃省交通规划勘察设计院股份有限公司 一种红外谱图的相似度确定方法及系统
CN112161965A (zh) * 2020-09-22 2021-01-01 复旦大学附属妇产科医院 检测中药药性的方法、装置、计算机设备和存储介质
CN112232673A (zh) * 2020-10-16 2021-01-15 拜耳医药保健有限公司 用于遗传资源的处理的预警方法、装置、设备和介质
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
CN112834451A (zh) * 2021-01-12 2021-05-25 深圳网联光仪科技有限公司 基于红外光谱的样品识别方法、设备及存储介质
CN113466206A (zh) * 2021-06-23 2021-10-01 上海仪电（集团）有限公司中央研究院 一种基于大数据的拉曼光谱分析系统
US11428638B2 (en) * 2018-01-12 2022-08-30 The Regents Of The University Of California Spectroscopic biological material characterization

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110489586A (zh) * 2019-01-07 2019-11-22 公安部第一研究所 一种样品检测数据库分级匹配的方法
TW202113681A (zh) * 2019-06-07 2021-04-01 德商巴斯夫塗料有限責任公司 用以形成用於一物件辨識系統之至少一實況資料庫之裝置及方法
CN111861171B (zh) * 2020-07-08 2021-07-27 京源中科科技股份有限公司 基于物联网的特殊物品监测系统、方法和装置
CN113324967A (zh) * 2020-09-18 2021-08-31 山东农业大学 一种快速鉴定dfd牛肉的方法
CN112686768A (zh) * 2020-12-30 2021-04-20 山东省食品药品检验研究院 一种药品制剂企业物料快速识别系统
CN113092407B (zh) * 2021-04-20 2024-01-30 深圳华创智信科技有限公司 基于近红外光谱的五常稻花香水稻原产地品种识别方法
CN113656559B (zh) * 2021-10-18 2022-01-25 印象(山东)大数据有限公司 一种基于计量平台的数据分析方法、装置及电子设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101403696A (zh) * 2008-10-21 2009-04-08 浙江大学 一种基于拉曼光谱的测量汽油烯烃含量的方法
US20100153323A1 (en) * 2007-03-05 2010-06-17 Kenneth Hennessy ensemble method and apparatus for classifying materials and quantifying the composition of mixtures
CN103217411A (zh) * 2013-04-15 2013-07-24 浙江大学 基于拉曼光谱的食用植物油中油酸、亚油酸和饱和脂肪酸含量检测方法
CN104215623A (zh) * 2013-05-31 2014-12-17 欧普图斯(苏州)光学纳米科技有限公司 面向多行业检测的激光拉曼光谱智能化辨识方法及系统
CN104764734A (zh) * 2015-04-30 2015-07-08 江苏易谱恒科技有限公司 基于拉曼光谱和svm算法的果汁香精香料的鉴别方法
CN105806824A (zh) * 2016-03-22 2016-07-27 武汉轻工大学 基于多源光谱特征融合的食用油种类快速鉴别方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015052551A (ja) * 2013-09-09 2015-03-19 株式会社かんでんエンジニアリング 油中成分測定方法、及び油入電気機器
CN104865279A (zh) * 2014-02-26 2015-08-26 清华大学 一种用于海关监管物项的快速检查设备和方法
CN103903010A (zh) * 2014-03-28 2014-07-02 哈尔滨工程大学 基于稀疏特征和邻域同属性的高光谱图像分类方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100153323A1 (en) * 2007-03-05 2010-06-17 Kenneth Hennessy ensemble method and apparatus for classifying materials and quantifying the composition of mixtures
CN101403696A (zh) * 2008-10-21 2009-04-08 浙江大学 一种基于拉曼光谱的测量汽油烯烃含量的方法
CN103217411A (zh) * 2013-04-15 2013-07-24 浙江大学 基于拉曼光谱的食用植物油中油酸、亚油酸和饱和脂肪酸含量检测方法
CN104215623A (zh) * 2013-05-31 2014-12-17 欧普图斯(苏州)光学纳米科技有限公司 面向多行业检测的激光拉曼光谱智能化辨识方法及系统
CN104764734A (zh) * 2015-04-30 2015-07-08 江苏易谱恒科技有限公司 基于拉曼光谱和svm算法的果汁香精香料的鉴别方法
CN105806824A (zh) * 2016-03-22 2016-07-27 武汉轻工大学 基于多源光谱特征融合的食用油种类快速鉴别方法

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11428638B2 (en) * 2018-01-12 2022-08-30 The Regents Of The University Of California Spectroscopic biological material characterization
US11680907B2 (en) 2018-01-12 2023-06-20 The Regents Of The University Of California Spectroscopic biological material characterization
CN111141719A (zh) * 2019-12-27 2020-05-12 梧州市食品药品检验所 抗艾滋病药物的快速无损鉴别方法
CN111426648A (zh) * 2020-03-19 2020-07-17 甘肃省交通规划勘察设计院股份有限公司 一种红外谱图的相似度确定方法及系统
CN112161965A (zh) * 2020-09-22 2021-01-01 复旦大学附属妇产科医院 检测中药药性的方法、装置、计算机设备和存储介质
CN112232673A (zh) * 2020-10-16 2021-01-15 拜耳医药保健有限公司 用于遗传资源的处理的预警方法、装置、设备和介质
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
CN112834481B (zh) * 2020-12-31 2023-09-05 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
CN112834451A (zh) * 2021-01-12 2021-05-25 深圳网联光仪科技有限公司 基于红外光谱的样品识别方法、设备及存储介质
CN113466206A (zh) * 2021-06-23 2021-10-01 上海仪电（集团）有限公司中央研究院 一种基于大数据的拉曼光谱分析系统

Also Published As

Publication number Publication date
CN108254351B (zh) 2023-08-01
CN108254351A (zh) 2018-07-06

Similar Documents

Publication Publication Date Title
WO2018121122A1 (zh) 用于物品查验的拉曼光谱检测方法和电子设备
CN108362662B (zh) 近红外光谱相似度计算方法、装置和物质定性分析系统
WO2018121121A1 (zh) 用于扣除谱图本底的方法、通过拉曼谱图识别物质的方法和电子设备
US7679740B2 (en) Method and apparatus for multimodal detection
US7990532B2 (en) Method and apparatus for multimodal detection
WO2017019988A1 (en) Gender and race identification from body fluid traces using spectroscopic analysis
CN107818298A (zh) 用于机器学习物质识别算法的通用拉曼光谱特征提取方法
WO2018121082A1 (zh) 基于拉曼光谱的自学习式定性分析方法
Khodabakhshian et al. An evaluation of IR spectroscopy for authentication of adulterated turmeric powder using pattern recognition
Yin et al. Non-destructive detection of foreign contaminants in toast bread with near infrared spectroscopy and computer vision techniques
Li et al. A novel hyperspectral imaging and modeling method for the component identification of woven fabrics
Liu et al. Identification of heat damage in imported soybeans based on hyperspectral imaging technology
Lotter et al. Identifying plastics with photoluminescence spectroscopy and machine learning
CN113310934A (zh) 骆驼奶中掺加奶牛奶及其掺加比例的快速鉴定方法
Anzanello et al. Wavelength selection framework for classifying food and pharmaceutical samples into multiple classes
US10564105B2 (en) Variable reduction method for spectral searching
Dreier et al. Hyperspectral imaging for classification of bulk grain samples with deep convolutional neural networks
WO2018121151A1 (zh) 用于识别拉曼谱图的方法和电子设备
Boodaghidizaji et al. Characterizing viral samples using machine learning for Raman and absorption spectroscopy
Huang et al. The application of wavelet transform of Raman spectra to facilitate transfer learning for gasoline detection and classification
Algazinov et al. Methods of measuring the spectral characteristics and identifying the components of grain mixtures in real-time separation systems
Hasbi et al. Pattern recognition for ultraviolet and fourier transform data: a walkthrough of techniques and direction
Huang et al. Hyperspectral image-based feature integration for insect-damaged hawthorn detection
WO2006019991A2 (en) Method and apparatus for multimodal detection
Sheykhizadeh et al. FTIR spectroscopy coupled with multivariate classification methods to identify different powdered infant formulas adulterated with melamine and cyanuric acid

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: 17888428

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17888428

Country of ref document: EP

Kind code of ref document: A1