WO2018227338A1 - 物质成分检测方法、装置和检测设备 - Google Patents

物质成分检测方法、装置和检测设备 Download PDF

Info

Publication number
WO2018227338A1
WO2018227338A1 PCT/CN2017/087942 CN2017087942W WO2018227338A1 WO 2018227338 A1 WO2018227338 A1 WO 2018227338A1 CN 2017087942 W CN2017087942 W CN 2017087942W WO 2018227338 A1 WO2018227338 A1 WO 2018227338A1
Authority
WO
WIPO (PCT)
Prior art keywords
substance
detected
spectral information
component
learning algorithm
Prior art date
Application number
PCT/CN2017/087942
Other languages
English (en)
French (fr)
Inventor
骆磊
黄晓庆
Original Assignee
深圳前海达闼云端智能科技有限公司
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 深圳前海达闼云端智能科技有限公司 filed Critical 深圳前海达闼云端智能科技有限公司
Priority to US16/320,086 priority Critical patent/US11468264B2/en
Priority to PCT/CN2017/087942 priority patent/WO2018227338A1/zh
Priority to CN201780002587.1A priority patent/CN108064341B/zh
Publication of WO2018227338A1 publication Critical patent/WO2018227338A1/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
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • 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
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • 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
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G01N2021/0106—General arrangement of respective parts
    • G01N2021/0118—Apparatus with remote processing
    • G01N2021/0137—Apparatus with remote processing with PC or the like
    • 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
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G01N2021/0181—Memory or computer-assisted visual determination
    • 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N2021/3129—Determining multicomponents by multiwavelength light
    • 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
    • G01N2201/12—Circuits of general importance; Signal processing
    • G01N2201/129—Using chemometrical methods
    • G01N2201/1296—Using chemometrical methods using neural networks

Definitions

  • the embodiments of the present application relate to the field of substance composition detection, for example, to a method, device and device for detecting a substance component.
  • the inventors have found that at least the following problems exist in the related art: when detecting a mixture using a Raman detecting terminal, the problem of the algorithm itself and the computing power of the device is limited, and the detecting process is slow and inefficient.
  • the embodiment of the present application provides a method for detecting a substance component, where the detecting method is applied to a detecting device, and the method includes:
  • the embodiment of the present application further provides a substance composition detecting device, wherein the detecting device is applied to a detecting device, and the device includes:
  • a spectral measurement module for acquiring spectral information of the substance to be detected
  • the prediction model of the method obtains the composition of the substance to be detected, and the prediction model based on the machine learning algorithm is formed by training the spectral information of the plurality of substances and the composition of the substance.
  • the embodiment of the present application further provides a detecting device, including:
  • At least one processor and,
  • the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described above.
  • the method, device and detection device for detecting a substance component obtained by the embodiments of the present application obtain the spectral information of the substance to be detected, and then match the spectral information to the prediction model based on the machine learning algorithm, because the prediction model based on the machine learning algorithm passes through the input.
  • the spectral information of the substance and the composition of the substance are trained, so that the spectral information of the substance to be detected is matched with the prediction model based on the machine learning algorithm, and the predicted result of the composition of the substance to be detected can be obtained.
  • the embodiment of the present application combines the machine learning algorithm with the spectral recognition, discards the traditional algorithm, improves the recognition speed, and greatly improves the efficiency of the substance detection.
  • 1a is a schematic diagram of an application scenario of the method and apparatus of the present application.
  • 1b is a schematic diagram of an application scenario of the method and apparatus of the present application.
  • FIG. 2 is a flow chart of one embodiment of a detection method of the present application.
  • FIG. 3 is a flow chart of steps of pre-acquiring a prediction model based on a machine learning algorithm in one embodiment of the detection method of the present application;
  • Figure 4 is a schematic diagram of Raman spectrum of a substance
  • FIG. 5 is a flowchart of a step of acquiring a voice command in an embodiment of the detection method of the present application
  • Figure 5 is a flow chart of one embodiment of the detection method of the present application.
  • FIG. 6 is a schematic structural view of an embodiment of a detecting device of the present application.
  • FIG. 7 is a schematic structural view of an embodiment of a detecting device of the present application.
  • Figure 8 is a schematic structural view of an embodiment of the detecting device of the present application.
  • FIG. 9 is a schematic structural diagram of hardware of a detecting device provided by an embodiment of the present application.
  • the embodiment of the present application provides a detection scheme for detecting a component of a substance based on a machine learning algorithm, which is applicable to the application scenario shown in FIG. 1a and FIG. 1b.
  • the substance to be detected 10 the detection terminal 21 and the detection device 20 are included, wherein the detection terminal 21 is used to obtain the spectral information of the substance to be detected 10, and then the spectral information of the substance 10 to be detected is transmitted.
  • the detection device 20 is given.
  • the detecting device 20 is configured to perform component identification based on the spectral information of the substance 10 to be detected.
  • the detecting terminal 21 and the detecting device 20 can communicate with each other through the network 30, wherein the network 30 can be, for example, a home or company local area network, or a specific network or the like.
  • the detecting terminal 21 and the detecting device 20 have at least one network interface to establish a communication connection with the network 30.
  • the detecting device 20 may be a cloud server or other server connected to the detecting terminal 10 via a network.
  • the detecting device 20 can also integrate the function of the detecting terminal into the detecting device 20, and the detecting device 20 separately obtains the spectral information of the substance to be detected 10 from the substance to be detected 10, and acquires the spectrum information through the spectrum information. Detect the composition of the substance.
  • the detecting device 20 pre-acquires a plurality of single substances in a preset single substance spectrum library (predetermined single substance spectrum library is a large database, including a plurality of single substances and corresponding spectral information) according to different compositions and different proportions.
  • the mixture and the corresponding spectral information of the mixture are then input as the input of the spectral information of the components and the mixture of the mixture, and the model is trained based on the machine learning algorithm to obtain the prediction model based on the machine learning algorithm.
  • the prediction model is based on a machine learning algorithm and a large amount of mixture spectral information and component data.
  • the spectral information of the substance to be detected is acquired, the spectral information is matched to the prediction model to obtain a component prediction result of the substance to be detected.
  • This scheme combines machine learning algorithms with spectral recognition, and eliminates the traditional algorithms, so that the efficiency of material detection can be greatly improved.
  • the application scenario may further include more substances to be detected 10 and the detecting device 20 and the detecting terminal 21.
  • the embodiment of the present application provides a material component detecting method, which may be performed by the detecting device 20 in FIG. 1a and FIG. 1b. As shown in FIG. 2, the material component detecting method includes:
  • Step 101 Obtain spectral information of a substance to be detected
  • the spectral recognition method in the embodiment of the present application may adopt a Raman spectrum identification method, an infrared spectrum identification method, or any other spectral recognition method, that is, the spectral information may be a Raman spectrum, an infrared spectrum, or the like.
  • Step 102 Match the spectral information to a pre-obtained prediction model based on a machine learning algorithm, obtain a component of the substance to be detected, and predict the model based on a machine learning algorithm by inputting spectral information and a plurality of substances.
  • the composition of the substance is trained to form.
  • the machine learning algorithm has a self-learning function. It takes the spectral information and components of a large number of different substances as input and trains the prediction model. The prediction model learns the composition of the substance based on the spectral information of the substance through the self-learning function.
  • the prediction result of the composition of the substance to be detected can be obtained.
  • the plurality of substances as the input may be a single substance or a mixture. Considering that the objects in real life are basically in the form of a mixture, it is also possible to perform model training only by taking the spectral information of the mixture and the composition of the mixture as inputs.
  • the predicted result of returning the constituents of the substance to be detected may be one or plural.
  • the returned prediction result is: y+a, y+b, y+c+d, .... If the probability of y+a is the largest and the probability exceeds the preset threshold, the prediction result y+a can be directly used as the final result. That is, as a component of the substance to be detected.
  • the embodiment of the present application combines the machine learning algorithm with the spectral recognition, discards the traditional algorithm, improves the recognition speed, and greatly improves the efficiency of the substance detection.
  • a single substance spectral library containing a large number of single substances and their corresponding spectral information can be preset, a single substance is selected from the single material spectral library, and a mixture is composed according to different compositions and different proportions as an input for model training.
  • the detecting device acquires the prediction model based on the machine learning algorithm in advance, the following steps are performed:
  • Step 201 Obtain spectral information of each single substance from a preset library of single substance spectra
  • the spectral information actually acquired by the detecting device is a set of data corresponding to the spectral curve, and the examples are as follows:
  • the first number of each group corresponds to the abscissa and the second number corresponds to the ordinate.
  • the step of the abscissa is generally 2, and the abscissa is generally 1000-2000 points.
  • Step 202 mixing each single substance according to different compositions and different proportions to obtain spectral information of each mixture;
  • the different compositions that is, the mixture, contain the types of the single substances, and the different proportions are the relative contents of the individual substances in the mixture, that is, the ratio of each single substance to the mixture.
  • the different proportions are for the same composition of the components, that is, for the combination of the single substances containing the same kind, according to the proportion of each single substance, A set of mixtures is obtained. In this way, for a mixture of the same composition, a large number of different data are used as input for model training, which improves the accuracy of composition identification.
  • Step 203 Taking spectral information and composition components of each mixture as input, performing model training based on a machine learning algorithm, and obtaining a prediction model based on a machine learning algorithm.
  • the mixture obtained in step 202 is taken as an input, and model training is performed based on the machine model algorithm.
  • the machine learning algorithm can use artificial neural network algorithm.
  • Artificial neural network is an algorithm developed according to human cognition process. If we only have some input and corresponding output now, and the mechanism of how to get output from input is not clear, then we can between input and output.
  • the unknown process is seen as a "network” that "trains” the network by continuously inputting and corresponding output to the network.
  • the network constantly adjusts the weights between its own nodes according to the input and output to satisfy the input and Output.
  • we give an input and the network will calculate an output based on its adjusted weight. This is the simple principle of neural networks.
  • the spectral information and components of a large number of different mixtures are first input into the artificial neural network, and the network learns according to the self-learning function.
  • the spectral information of the mixture identifies the constituents of the mixture.
  • step 202 of certain embodiments of the method all possible combinations of individual materials can be listed and then the mixture can be obtained based on different ratios, as follows:
  • Step 2021 Obtain a possible combination of each single substance in the preset single substance spectrum library, or obtain a possible combination of the single substances in the preset single substance spectrum library that does not exceed the preset threshold value;
  • the mixture of all possible combinations of the single substances is C(n, x), where x is the species of the single substance in the mixture.
  • the mixing of many substances is rare, or even if it is really a mixture of many substances, users generally only care about the first few kinds of ingredients, and the content of few ingredients is It does not necessarily make sense, so you can set the upper limit xmax of the mixed type, which is the preset threshold.
  • the mixture of all possible combinations of single substances is C(n, x), the value of the preset threshold xmax can be set according to actual needs.
  • Step 2022 According to the number of types of single substances in each combination, according to a preset step value, obtain Multiple mixing ratios for each combination;
  • each type of mixture in step 2021 is obtained as a mixture of different amounts of the same component according to the different contents of the single substances contained therein.
  • the x kinds of single substances are j1, j2...jx, respectively, and their proportions are respectively set to z1%, z2%...zx% according to a certain preset step value in a permutation and combination manner.
  • the initial z1, z2, and z3 can be set to 99.98, 0.01, 0.01, 99.97, 0.02, 0.01, ..., 99.97, 0.01, 0.02, and so on, and all possible permutations and combinations until z1.
  • the value of z2, z3 becomes 0.01, 0.01, 99.98.
  • the preset step value is 0.01, and the preset stepping system can also be other values. This application does not limit this. The smaller the step value, the more samples are trained and the longer the training time is. .
  • Step 2023 Obtain a plurality of mixtures according to a possible combination of each single substance and a plurality of mixing ratios of each combination;
  • Step 2024 The spectral information of each component of each mixture is linearly superimposed according to the mixing ratio to obtain spectral information of the mixture.
  • the proportions of j1, j2...jx are z1%, z2%...zx%, respectively, and the spectral information of the single substances j1, j2...jx are obtained respectively.
  • the ordinate component value ie, the spectral intensity value
  • the ordinate component values are added according to the mixing ratio (ie, the ratio) to obtain the ordinate value of the mixture quality at the abscissa point.
  • the spectral intensity values of j1, j2...jx are b11, b12...b1x, respectively
  • the spectral intensity value of the mixture at abscissa a1 is b11*z1%+b12*z2%+... +b1x*zx%.
  • a single substance 1 and a single substance 2 are mixed, the ordinate of the single substance 1 at the abscissa 950 is 0.4, the ordinate of the single substance 2 at the abscissa 950 is 0.2, and the ratio of the single substance 1 to the single substance 2 is 3.
  • the Raman spectral information of the mixture is obtained by finding the ordinate spectral intensity value of the mixture for each point on the abscissa.
  • the proportion of each component in the substance to be detected may also be obtained according to the composition of the substance to be detected, including the following steps:
  • the composition has a proportion of 100%.
  • the detection result determines that the substance to be detected includes at least two components, obtaining spectral information of each component in the substance to be detected from a preset library of single substance spectra; according to spectral information of the substance to be detected and to be detected The spectral information of each component in the substance acquires the proportion of each component in the substance to be detected.
  • the substance to be detected is a mixture including a plurality of single substances
  • spectral data of each single substance is obtained from a single substance spectrum library, and spectral data of the mixture is fitted to obtain each single substance in the mixture. Proportion.
  • the mixture includes x kinds of single substances, as long as the x-1 features of the mixture spectral data are more obvious abscissa points (the abscissa point of the wave), the solution can be solved (x-1)
  • the elementary equation gives the proportion of each single substance.
  • the method includes:
  • Step 301 Obtain spectral information of each single substance from a preset library of single substance spectra
  • Step 302 Obtain a possible combination of each single substance in the preset single substance spectrum library, or obtain a possible combination of the single substances in the preset single substance spectrum library that does not exceed the preset threshold value;
  • Step 303 Acquire a plurality of mixing ratios of each combination according to a preset step value according to the number of types of the single substances in each combination;
  • Step 304 Obtain a plurality of mixtures according to a possible combination of each single substance and a plurality of mixing ratios of each combination;
  • Step 305 Linearly superimposing the spectral information of each component of each mixture according to the mixing ratio to obtain spectral information of the mixture.
  • Step 306 Taking spectral information and composition components of each mixture as input, performing model training based on a machine learning algorithm, and obtaining a prediction model based on a machine learning algorithm.
  • Step 307 Acquire spectral information of the substance to be detected
  • Step 308 Match the spectral information to a pre-obtained prediction model based on a machine learning algorithm to acquire a component of the substance to be detected.
  • Step 309 Obtain a proportion of each component in the substance to be detected according to the composition of the substance to be detected.
  • the embodiment of the present application further provides a substance composition detecting device, which is used in the detecting device shown in FIG. 1a or 1b.
  • the detecting device 400 includes:
  • a spectrum measuring module 401 configured to acquire spectral information of the substance to be detected
  • a material component obtaining module 402 configured to match the spectral information to a pre-obtained prediction model based on a machine learning algorithm, acquire a component of the substance to be detected, and predict the model based on a machine learning algorithm by inputting a plurality of substances The spectral information and the composition of the substance are trained to form.
  • the embodiment of the present application obtains the spectral information of the substance to be detected, and then matches the spectral information to the prediction model based on the machine learning algorithm, because the prediction model based on the machine learning algorithm inputs the spectral information of the plurality of substances and the composition of the substance. It is trained, so the spectral information of the substance to be detected is matched with the prediction model based on the machine learning algorithm, and the prediction result of the components of the substance to be detected can be obtained.
  • the embodiment of the present application combines the machine learning algorithm with the spectral recognition, discards the traditional algorithm, improves the recognition speed, and greatly improves the efficiency of the substance detection.
  • the substance component obtaining module is specifically configured to:
  • a common component of the prediction result having a relatively large probability of a predetermined number of prediction results is used as a component of the substance to be detected.
  • the detecting apparatus 500 includes: in addition to the spectrum measuring module 502 and the material component acquiring module 503,
  • a prediction model obtaining module 501 configured to acquire a prediction model based on a machine learning algorithm in advance
  • the prediction model obtaining module 501 is specifically configured to:
  • model based on machine learning algorithm Training to obtain a predictive model based on machine learning algorithms.
  • the predictive model obtaining module is further configured to:
  • the spectral information of each component of each mixture was linearly superimposed according to the mixing ratio to obtain spectral information of the mixture.
  • the detecting apparatus 600 includes: a prediction model acquiring module 601, a spectrum measuring module 602, and a substance component acquiring module 603, in addition to:
  • the component content obtaining module 604 is configured to obtain a proportion of each component in the substance to be detected according to a composition component of the substance to be detected;
  • the component content obtaining module 604 is specifically configured to:
  • the substance to be detected includes at least two components, obtaining spectral information of each component in the substance to be detected from a preset library of single substance spectra;
  • the proportion of each component in the substance to be detected is obtained according to the spectral information of the substance to be detected and the spectral information of each component in the substance to be detected.
  • the machine learning algorithm is an artificial neural network learning algorithm.
  • the foregoing detecting apparatus can perform the detecting method provided by the embodiment of the present application, and has the corresponding functional modules and beneficial effects of the executing method.
  • the detection method provided by the embodiment of the present application.
  • FIG. 9 is a schematic diagram showing the hardware structure of the detecting device 20 according to the embodiment of the present application. As shown in FIG. 9, the detecting device 20 includes:
  • processors 22 and memory 23, one processor 22 is exemplified in FIG.
  • the processor 22 and the memory 23 can be connected by a bus or other means, as exemplified by a bus connection in FIG.
  • the memory 23 is used as a non-volatile computer readable storage medium, and can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as program instructions corresponding to the detection methods in the embodiments of the present application.
  • a module for example, the spectral measurement module 401 and the substance composition acquisition module 402 shown in FIG. 6.
  • the processor 22 executes various functional applications and data processing of the server by executing non-volatile software programs, instructions, and modules stored in the memory 23, that is, the detection method of the above method embodiments.
  • the memory 23 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created according to usage of the detecting device, and the like. Further, the memory 23 may include a high speed random access memory, and may also include a nonvolatile memory such as at least one magnetic disk storage device, flash memory device, or other nonvolatile solid state storage device. In some embodiments, memory 23 can optionally include memory remotely located relative to processor 22, which can be connected to the detection device over a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
  • the one or more modules are stored in the memory 23, and when executed by the one or more processors 22, perform the detection method in any of the above method embodiments, for example, performing the above described FIG. Method step 101 to step 102, method step 201 to step 203 in FIG. 3, method step 301 to step 309 in FIG. 5; module 401-402 in FIG. 6, module 501-503 in FIG. 7, and FIG. The function of modules 601-604.
  • the embodiment of the present application provides a non-transitory computer readable storage medium storing computer-executable instructions that are executed by one or more processors, such as in FIG. a processor 22, which may cause the one or more processors to perform the detection method in any of the above method embodiments, for example, to perform the method step 101 to step in FIG. 2 described above.
  • the function implement modules 401-402 in FIG. 6, modules 501-503 in FIG. 7, and modules 601-604 in FIG.
  • the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, ie may be located A place, or it can be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Algebra (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computational Mathematics (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

一种物质成分检测方法、装置和检测设备，检测方法包括：获取待检测物质（10）的光谱信息（101）；将光谱信息匹配预先获取的基于机器学习算法的预测模型，获取待检测物质（10）的组成成分，基于机器学习算法的预测模型，通过输入多种物质的光谱信息和物质的组成成分训练形成（102）。通过获取待检测物质（10）的光谱信息（101），然后将光谱信息匹配基于机器学习算法的预测模型，获得对待检测物质（10）的组成成分的预测结果。检测方法将机器学习算法与光谱识别相结合，摒弃了传统的算法，提高了识别速度，使物质检测的效率得以大幅提升。

Description

物质成分检测方法、装置和检测设备 技术领域
本申请实施例涉及物质成分检测领域，例如涉及一种物质成分检测方法、装置和检测设备。
背景技术
近年来，物质检测设备应用日趋广泛，包括安检中检测可疑物品、药监局检测药品成分、防化部队进行爆炸现场勘查等专业领域，也包括检测农药残留、检测是否存在三聚氰胺、检测地沟油和真假酒等民用领域，尤其在食品安全领域得到广泛应用。目前的检测设备，例如拉曼检测终端，采用拉曼光谱分析的方法，能够比较快速和准确的检测出物质分子成分。
在实现本申请过程中，发明人发现相关技术中至少存在如下问题：在使用拉曼检测终端进行混合物的检测时，限于算法本身和设备计算能力的问题，检测过程较慢，效率低下。
发明内容
本申请实施例的一个目的是提供一种新的物质成分检测方法、装置和检测设备，在用于检测混合物的成分时，能快速检测出混合物的成分。
第一方面，本申请实施例提供了一种物质成分检测方法，所述检测方法应用于检测设备，所述方法包括：
获取待检测物质的光谱信息；
将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
第二方面，本申请实施例还提供了一种物质成分检测装置，所述检测装置应用于检测设备，所述装置包括：
光谱测量模块，用于获取待检测物质的光谱信息；
物质成分获取模块，用于将所述光谱信息匹配预先获取的基于机器学习算 法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
第三方面，本申请实施例还提供了一种检测设备，包括：
至少一个处理器；以及，
与所述至少一个处理器通信连接的存储器；其中，
所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述的方法。
本申请实施例提供的物质成分检测方法、装置和检测设备，通过获取待检测物质的光谱信息，然后将该光谱信息匹配基于机器学习算法的预测模型，由于基于机器学习算法的预测模型通过输入多种物质的光谱信息和所述物质的组成成分训练而成，因此将待检测物质的光谱信息匹配基于机器学习算法的预测模型，就能获得对待检测物质的组成成分的预测结果。本申请实施例将机器学习算法与光谱识别相结合，摒弃了传统的算法，提高了识别速度，使物质检测的效率得以大幅提升。
附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定，附图中具有相同参考数字标号的元件表示为类似的元件，除非有特别申明，附图中的图不构成比例限制。
图1a是本申请方法和装置的应用场景示意图；
图1b是本申请方法和装置的应用场景示意图；
图2是本申请检测方法的一个实施例的流程图；
图3是本申请检测方法的一个实施例中预先获取基于机器学习算法的预测模型步骤的流程图；
图4是物质的拉曼光谱示意图；
图5是本申请检测方法的一个实施例中获取语音命令步骤的流程图；
图5是本申请检测方法的一个实施例的流程图；
图6是本申请检测装置的一个实施例的结构示意图；
图7是本申请检测装置的一个实施例的结构示意图；
图8是本申请检测装置的一个实施例的结构示意图；以及
图9是本申请实施例提供的检测设备的硬件结构示意图。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请实施例提供了一种基于机器学习算法检测物质成分的检测方案，适用于图1a和图1b所示的应用场景。在图1a所示的应用场景中，包括待检测物质10、检测终端21和检测设备20，其中，检测终端21用于获得待检测物质10的光谱信息，然后将待检测物质10的光谱信息传送给检测设备20。检测设备20用于根据待检测物质10的光谱信息进行组成成分识别。检测终端21与检测设备20之间可以通过网络30互相通信，其中，网络30可以是例如家庭或公司的局域网，或一个特定网络等。检测终端21和检测设备20具有至少一个网络接口，与网络30建立通信连接。检测设备20可以是与检测终端10通过网络相连的云端服务器或者其他服务器。如图1b所示，检测设备20也可以将检测终端的功能集成在检测设备20中，由检测设备20单独完成从待检测物质10获取待检测物质10的光谱信息，并通过该光谱信息获取待检测物质的成分。
检测设备20预先将预设的单物质光谱库(预设的单物质光谱库为大数据库，包含多种单物质及其对应的光谱信息)中的单物质按照不同组成和不同占比获取大量的混合物以及混合物对应的光谱信息，然后将大量混合物的组成成分和混合物的光谱信息作为输入，基于机器学习算法进行模型训练，获取基于机器学习算法的预测模型。该预测模型基于机器学习算法和大量混合物光谱信息和组成成分的数据，因此当获取待检测物质的光谱信息后，将该光谱信息匹配该预测模型就能获得待检测物质的组成成分预测结果。本方案将机器学习算法与光谱识别相结合，摒弃了传统的算法，使物质检测的效率得以大幅提升。
需要说明的是，在实际应用过程中，该应用场景还可以包括更多的待检测物质10和检测设备20以及检测终端21。
本申请实施例提供了一种物质成分检测方法，所述物质成分检测方法可由图1a和图1b中的检测设备20执行，如图2所示，所述物质成分检测方法包括：
步骤101：获取待检测物质的光谱信息；
本申请实施例中的光谱识别方法可以采用拉曼光谱识别方法、红外光谱识别方法或者其他任何一种光谱识别方法，即所述光谱信息可以为拉曼光谱、红外光谱等。
步骤102：将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
机器学习算法具有自学习功能，把大量不同物质的光谱信息和组成成分作为输入，训练预测模型，预测模型就会通过自学习功能，学会根据物质的光谱信息识别物质的组成成分。
因此将待检测物质的光谱信息匹配基于机器学习算法的预测模型，就能获得对待检测物质的组成成分的预测结果。其中，作为输入的多种物质可以是单物质也可以是混合物。考虑到现实生活中的物品基本都是以混合物的形式存在，所以也可以只将混合物的光谱信息和混合物的组成成分作为输入进行模型训练。
在实际应用中，返回的对所述待检测物质的组成成分的预测结果有可能为一个，也有可能为多个。预测结果为多个时，需要确认每个预测结果所占的概率。如果某个预测结果的概率最大且超过预设阈值，则确认该预测结果为待检测物质的成分；否则，将各预测结果中预设数量的概率相对较大的预测结果的共有成分作为所述待检测物质的成分。
例如，返回的预测结果为：y+a、y+b、y+c+d、…，如果y+a的概率最大且概率超过预设阈值，则可以直接将预测结果y+a作为最终结果，即作为待检测物质的成分。
如果y+a的概率最大但是概率没有超过预设阈值，如果预设数量为3，且y+a、y+b、y+c+d的概率相对其他预测结果大，则将y+a、y+b、y+c+d的共有成 分y作为最终结果。
本申请实施例将机器学习算法与光谱识别相结合，摒弃了传统的算法，提高了识别速度，使物质检测的效率得以大幅提升。
为了提高模型预测的准确性，该模型训练需要基于大量数据进行训练。因此应用中，可以预设一个包含大量单物质和其对应的光谱信息的单物质光谱库，从该单物质光谱库中选取单物质，按照不同组成和不同比例组成混合物作为输入，进行模型训练。请参照图3，检测设备预先获取基于机器学习算法的预测模型时，执行以下步骤：
步骤201：从预设的单物质光谱库中获取各单物质的光谱信息；
如图4所示，为两种粉末状物质的拉曼光谱示意图，而实际应用中，检测设备实际获取的光谱信息是与光谱曲线对应的一组数据，示例如下：
0 0.002010
2 0.001219
……
12080.012633
12100.003053
12120.000525
……
1998 0.001028
2000 0.001232
每一组的第一个数字对应横坐标，第二个数字对应纵坐标。以拉曼物质检测设备为例，横坐标的步进一般来说为2，横坐标一般取1000-2000个点。
步骤202:将各单物质按照不同组成和不同占比进行混合，获得各混合物的光谱信息；
不同组成即混合物中包含单物质的种类，不同占比即混合物中各单物质的相对含量，也即每种单物质占混合物的比例。不同占比针对的是组成成分相同的混合物，即针对包含相同种类单物质的组合，按照每种单物质的占比不同， 获得一组混合物。这样针对组成成分相同的混合物，有大量占比不同的数据作为输入进行模型训练，提高了组成成分识别的准确性。
步骤203:将各混合物的光谱信息和组成成分作为输入，基于机器学习算法进行模型训练，获得基于机器学习算法的预测模型。
即将步骤202中获取的混合物作为输入，基于机器模型算法，进行模型训练。其中，机器学习算法可以采用人工神经网络算法。人工神经网络是根据人的认识过程而开发出的一种算法，假如我们现在只有一些输入和相应的输出，而对如何由输入得到输出的机理并不清楚，那么我们可以把输入与输出之间的未知过程看成是一个“网络”，通过不断地给这个网络输入和相应的输出来“训练”这个网络，网络根据输入和输出不断地调节自己的各节点之间的权值来满足输入和输出。这样，当训练结束后，我们给定一个输入，网络便会根据自己已调节好的权值计算出一个输出。这就是神经网络的简单原理。
在本申请实施例中为了根据混合物的光谱信息(输入)识别混合物的组成成分(输出)，先把大量不同混合物的光谱信息和组成成分输入人工神经网络，网络就会通过自学习功能，学会根据混合物的光谱信息识别混合物的组成成分。
具体的，在所述方法的某些实施例的步骤202中，可以列出各单物质的所有可能组合再基于不同比例获得混合物，如下：
步骤2021：获取预设单物质光谱库中各单物质的可能组合，或者，获取预设单物质光谱库中单物质组合种类不超过预设阀值的各单物质的可能组合；
获取单物质的所有可能组合的混合物，例如，假设单物质光谱库中有n种物质，则单物质的所有可能组合的混合物种类为
Figure PCTCN2017087942-appb-000001
C(n，x)，其中，x为混合物中单物质的种类。在实际应用中，过多种物质的混合情况并不多见，或者说，即使真的是很多种物质混合，使用者一般仅关心含量多的前若干种，含量很少的成分对使用者并不一定有意义，因此可以设置混合种类的上限xmax，即预设阀值。那么，单物质的所有可能组合的混合物种类为
Figure PCTCN2017087942-appb-000002
C(n，x)，预设阀值xmax的值可以根据实际需要设定。
步骤2022：根据每个组合中单物质的种类数量，按照预设的步进值，获取 每个组合的多个混合比例；
即将步骤2021中每个种类的混合物按其所含单物质的不同含量获得一组成分相同含量不同的混合物。假设x种单物质分别为j1，j2…jx，将其占比按照一定的预设步进值以排列组合方式分别置为z1％，z2％…zx％。假设x为3，则此处初始z1，z2，z3可分别设置为99.98、0.01、0.01，99.97、0.02、0.01，…，99.97、0.01、0.02，以此类推做所有可能排列组合，直到z1，z2，z3的值变为0.01，0.01，99.98。该例中预设步进值为0.01，预设步进制也可以为其他值，本申请对此不做限制，步进值越小，则训练的样本越多，整个训练时间也就越长。
步骤2023：根据各单物质的可能组合和每个组合的多个混合比例，获取多个混合物；
步骤2024：将每个混合物的各个组成成分的光谱信息按照混合比例进行线性叠加，获得混合物的光谱信息。
仍以步骤2022中x种单物质分别为j1，j2…jx为例，j1，j2…jx的占比分别为z1％，z2％…zx％，分别获得单物质j1，j2…jx的光谱信息中某一相同横坐标对应的纵坐标分量值(即光谱强度值)，将该纵坐标分量值按照混合比例(即占比)相加得到混合物质在该横坐标点的纵坐标值。例如，假设在横坐标a1处，j1，j2…jx的光谱强度值分别为b11，b12…b1x，则此混合物在横坐标a1处的光谱强度值为b11*z1％+b12*z2％+…+b1x*zx％。例如单物质1和单物质2混合，单物质1在横坐标950处的纵坐标为0.4，单物质2在横坐标950处的纵坐标为0.2，单物质1和单物质2的占比为3：7，则可以得出该混合物在横坐标950处的纵坐标为0.4*30％+0.2*70％＝0.26。通过对横坐标上的每一个点求混合物的纵坐标光谱强度值，即可得到该混合物的拉曼光谱信息。
进一步的，在所述方法的其他实施例中，还可以根据待检测物质的组成成分获取待检测物质中各组成成分的占比，包括以下步骤:
如果检测结果只包含一种组成成分，则该组成成分的占比为100％。
如果检测结果判定所述待检测物质包括至少两种组成成分，则从预设的单物质光谱库中获取所述待检测物质中各组成成分的光谱信息；根据待检测物质的光谱信息和待检测物质中各组成成分的光谱信息，获取各组成成分在待检测物质中的占比。
即如果所述待检测物质为包括多个单物质的混合物，则从单物质光谱库中获取各单物质的光谱数据，并对该混合物的光谱数据进行拟合，得到各单物质在混合物中的占比。实际计算时，如果该混合物包括x种单物质，则只要取该混合物光谱数据的x–1个特征较明显的横坐标点(有波的横坐标点)，即可通过求解(x-1)元一次方程得到各单物质的占比。
如图5所示，为所述方法的一个实施例的流程示意图，在该实施例中，所述方法包括：
步骤301：从预设的单物质光谱库中获取各单物质的光谱信息；
步骤302：获取预设单物质光谱库中各单物质的可能组合，或者，获取预设单物质光谱库中单物质组合种类不超过预设阀值的各单物质的可能组合；
步骤303：根据每个组合中单物质的种类数量，按照预设的步进值，获取每个组合的多个混合比例；
步骤304：根据各单物质的可能组合和每个组合的多个混合比例，获取多个混合物；
步骤305：将每个混合物的各个组成成分的光谱信息按照混合比例进行线性叠加，获得混合物的光谱信息。
步骤306：将各混合物的光谱信息和组成成分作为输入，基于机器学习算法进行模型训练，获得基于机器学习算法的预测模型。
步骤307：获取待检测物质的光谱信息；
步骤308：将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分。
步骤309：根据待检测物质的组成成分获取待检测物质中各组成成分的占比。
相应的，本申请实施例还提供了一种物质成分检测装置，所述检测装置用于图1a或者图1b所示的检测设备，如图6所示，所述检测装置400包括：
光谱测量模块401，用于获取待检测物质的光谱信息；
物质成分获取模块402，用于将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
本申请实施例通过获取待检测物质的光谱信息，然后将该光谱信息匹配基于机器学习算法的预测模型，由于基于机器学习算法的预测模型通过输入多种物质的光谱信息和所述物质的组成成分训练而成，因此将待检测物质的光谱信息匹配基于机器学习算法的预测模型，就能获得对待检测物质的组成成分的预测结果。本申请实施例将机器学习算法与光谱识别相结合，摒弃了传统的算法，提高了识别速度，使物质检测的效率得以大幅提升。
可选的，在所述装置的其他实施例中，所述物质成分获取模块具体用于：
将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取各针对待检测物质成分的预测结果及预测结果对应的概率；
如果预测结果的概率最大且超过预设阈值，则确认所述预测结果为待检测物质的成分；
否则，将各预测结果中预设数量的概率相对较大的预测结果的共有成分作为所述待检测物质的成分。
可选的，如图7所示，在所述装置的其他实施例中，所述检测装置500除了光谱测量模块502和物质成分获取模块503之外还包括：
预测模型获取模块501，用于预先获取基于机器学习算法的预测模型；
所述预测模型获取模块501具体用于：
从预设的单物质光谱库中获取各单物质的光谱信息；
将各单物质按照不同组成和不同占比进行混合，获得各混合物的光谱信息；
将各混合物的光谱信息和组成成分作为输入，基于机器学习算法进行模型 训练，获得基于机器学习算法的预测模型。
可选的，在所述装置的某些实施例中，所述预测模型获取模块还用于：
获取预设单物质光谱库中各单物质的可能组合，或者，获取预设单物质光谱库中单物质组合种类不超过预设阀值的各单物质的可能组合；
根据每个组合中单物质的种类数量，按照预设的步进值，获取每个组合的多个混合比例；
根据各单物质的可能组合和每个组合的多个混合比例，获取多个混合物；
将每个混合物的各个组成成分的光谱信息按照混合比例进行线性叠加，获得混合物的光谱信息。
可选的，如图8所示，在所述装置的其他实施例中，所述检测装置600除了包括预测模型获取模块601、光谱测量模块602和物质成分获取模块603之外还包括：
成分含量获取模块604，用于根据待检测物质的组成成分获取待检测物质中各组成成分的占比；
所述成分含量获取模块604具体用于：
如果所述待检测物质包括至少两种组成成分，则从预设的单物质光谱库中获取所述待检测物质中各组成成分的光谱信息；
根据待检测物质的光谱信息和待检测物质中各组成成分的光谱信息，获取各组成成分在待检测物质中的占比。
可选的，在所述装置的某些实施例中，所述机器学习算法为人工神经网络学习算法。
需要说明的是，上述检测装置可执行本申请实施例所提供的检测方法，具备执行方法相应的功能模块和有益效果。未在检测装置实施例中详尽描述的技术细节，可参见本申请实施例所提供的检测方法。
图9是本申请实施例提供的检测设备20的硬件结构示意图，如图9所示，该检测设备20包括：
一个或多个处理器22以及存储器23，图9中以一个处理器22为例。
处理器22和存储器23可以通过总线或者其他方式连接，图9中以通过总线连接为例。
存储器23作为一种非易失性计算机可读存储介质，可用于存储非易失性软件程序、非易失性计算机可执行程序以及模块，如本申请实施例中的检测方法对应的程序指令/模块(例如，附图6所示的光谱测量模块401和物质成分获取模块402)。处理器22通过运行存储在存储器23中的非易失性软件程序、指令以及模块，从而执行服务器的各种功能应用以及数据处理，即实现上述方法实施例的检测方法。
存储器23可以包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需要的应用程序；存储数据区可存储根据检测装置的使用所创建的数据等。此外，存储器23可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实施例中，存储器23可选包括相对于处理器22远程设置的存储器，这些远程存储器可以通过网络连接至检测装置。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
所述一个或者多个模块存储在所述存储器23中，当被所述一个或者多个处理器22执行时，执行上述任意方法实施例中的检测方法，例如，执行以上描述的图2中的方法步骤101至步骤102，图3中的方法步骤201至步骤203，图5中的方法步骤301至步骤309；实现图6中的模块401-402、图7中模块501-503，图8中模块601-604的功能。
上述产品可执行本申请实施例所提供的方法，具备执行方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节，可参见本申请实施例所提供的方法。
本申请实施例提供了一种非易失性计算机可读存储介质，所述计算机可读存储介质存储有计算机可执行指令，该计算机可执行指令被一个或多个处理器执行，例如图9中的一个处理器22，可使得上述一个或多个处理器可执行上述任意方法实施例中的检测方法，例如，执行以上描述的图2中的方法步骤101至步 骤102，图3中的方法步骤201至步骤203，图5中的方法步骤301至步骤309；实现图6中的模块401-402、图7中模块501-503，图8中模块601-604的功能。
以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
通过以上的实施方式的描述，本领域普通技术人员可以清楚地了解到各实施方式可借助软件加通用硬件平台的方式来实现，当然也可以通过硬件。本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程是可以通过计算机程序来指令相关的硬件来完成，所述的程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM)或随机存储记忆体(RandomAccessMemory,RAM)等。
最后应说明的是：以上实施例仅用以说明本申请的技术方案，而非对其限制；在本申请的思路下，以上实施例或者不同实施例中的技术特征之间也可以进行组合，步骤可以以任意顺序实现，并存在如上所述的本申请的不同方面的许多其它变化，为了简明，它们没有在细节中提供；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (15)

  1. 一种物质成分检测方法，所述检测方法应用于检测设备，其特征在于，所述方法包括：
    获取待检测物质的光谱信息；
    将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
  2. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    预先获取基于机器学习算法的预测模型；
    所述预先获取基于机器学习算法的预测模型，包括：
    从预设的单物质光谱库中获取各单物质的光谱信息；
    将各单物质按照不同组成和不同占比进行混合，获得各混合物的光谱信息；
    将各混合物的光谱信息和组成成分作为输入，基于机器学习算法进行模型训练，获得基于机器学习算法的预测模型。
  3. 根据权利要求2所述的方法，其特征在于，所述将各单物质按照不同组成和不同占比进行混合，获得各混合物的光谱信息，包括：
    获取预设单物质光谱库中各单物质的可能组合，或者，获取预设单物质光谱库中单物质组合种类不超过预设阀值的各单物质的可能组合；
    根据每个组合中单物质的种类数量，按照预设的步进值，获取每个组合的多个混合比例；
    根据各单物质的可能组合和每个组合的多个混合比例，获取多个混合物；
    将每个混合物的各个组成成分的光谱信息按照混合比例进行线性叠加，获得混合物的光谱信息。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，所述方法还包括： 根据待检测物质的组成成分获取待检测物质中各组成成分的占比；
    所述根据待检测物质的组成成分获取待检测物质中各组成成分的占比，包括：
    如果所述待检测物质包括至少两种组成成分，则从预设的单物质光谱库中获取所述待检测物质中各组成成分的光谱信息；
    根据待检测物质的光谱信息和待检测物质中各组成成分的光谱信息，获取各组成成分在待检测物质中的占比。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，包括：
    将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取各针对待检测物质成分的预测结果及预测结果对应的概率；
    如果预测结果的概率最大且超过预设阈值，则确认所述预测结果为待检测物质的成分；
    否则，将各预测结果中预设数量的概率相对较大的预测结果的共有成分作为所述待检测物质的成分。
  6. 根据权利要求1-5任一项所述的方法，其特征在于，所述机器学习算法为人工神经网络学习算法。
  7. 一种物质成分检测装置，所述检测装置应用于检测设备，其特征在于，所述装置包括：
    光谱测量模块，用于获取待检测物质的光谱信息；
    物质成分获取模块，用于将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取所述待检测物质的组成成分，所述基于机器学习算法的预测模型，通过输入多种物质的光谱信息和所述物质的组成成分训练形成。
  8. 根据权利要求7所述的装置，其特征在于，所述装置还包括：
    预测模型获取模块，用于预先获取基于机器学习算法的预测模型；
    所述预测模型获取模块具体用于：
    从预设的单物质光谱库中获取各单物质的光谱信息；
    将各单物质按照不同组成和不同占比进行混合，获得各混合物的光谱信息；
    将各混合物的光谱信息和组成成分作为输入，基于机器学习算法进行模型训练，获得基于机器学习算法的预测模型。
  9. 根据权利要求8所述的装置，其特征在于，所述预测模型获取模块还用于：
    获取预设单物质光谱库中各单物质的可能组合，或者，获取预设单物质光谱库中单物质组合种类不超过预设阀值的各单物质的可能组合；
    根据每个组合中单物质的种类数量，按照预设的步进值，获取每个组合的多个混合比例；
    根据各单物质的可能组合和每个组合的多个混合比例，获取多个混合物；
    将每个混合物的各个组成成分的光谱信息按照混合比例进行线性叠加，获得混合物的光谱信息。
  10. 根据权利要求7-9任一项所述的装置，其特征在于，所述装置还包括：成分含量获取模块，用于根据待检测物质的组成成分获取待检测物质中各组成成分的占比；
    所述成分含量获取模块具体用于：
    如果所述待检测物质包括至少两种组成成分，则从预设的单物质光谱库中获取所述待检测物质中各组成成分的光谱信息；
    根据待检测物质的光谱信息和待检测物质中各组成成分的光谱信息，获取各组成成分在待检测物质中的占比。
  11. 根据权利要求7-10任一项所述的装置，其特征在于，所述物质成分获取模块具体用于：
    将所述光谱信息匹配预先获取的基于机器学习算法的预测模型，获取各针 对待检测物质成分的预测结果及预测结果对应的概率；
    如果预测结果的概率最大且超过预设阈值，则确认所述预测结果为待检测物质的成分；
    否则，将各预测结果中预设数量的概率相对较大的预测结果的共有成分作为所述待检测物质的成分。
  12. 根据权利要求7-11任一项所述的装置，其特征在于，所述机器学习算法为人工神经网络学习算法。
  13. 一种检测设备，其特征在于，包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求1-6任一项所述的方法。
  14. 一种非易失性计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机可执行指令，当所述计算机可执行指令被检测设备执行时，使所述检测设备执行执行权利要求1-6任一项所述的方法。
  15. 一种计算机程序产品，其特征在于，所述计算机程序产品包括存储在非易失性计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被检测设备执行时，使所述检测设备执行权利要求1-6任一项所述的方法。
PCT/CN2017/087942 2017-06-12 2017-06-12 物质成分检测方法、装置和检测设备 WO2018227338A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US16/320,086 US11468264B2 (en) 2017-06-12 2017-06-12 Substance ingredient detection method and apparatus, and detection device
PCT/CN2017/087942 WO2018227338A1 (zh) 2017-06-12 2017-06-12 物质成分检测方法、装置和检测设备
CN201780002587.1A CN108064341B (zh) 2017-06-12 2017-06-12 物质成分检测方法、装置和检测设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/087942 WO2018227338A1 (zh) 2017-06-12 2017-06-12 物质成分检测方法、装置和检测设备

Publications (1)

Publication Number Publication Date
WO2018227338A1 true WO2018227338A1 (zh) 2018-12-20

Family

ID=62142065

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/087942 WO2018227338A1 (zh) 2017-06-12 2017-06-12 物质成分检测方法、装置和检测设备

Country Status (3)

Country Link
US (1) US11468264B2 (zh)
CN (1) CN108064341B (zh)
WO (1) WO2018227338A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110503141A (zh) * 2019-08-13 2019-11-26 Oppo（重庆）智能科技有限公司 酒类鉴别方法、装置及计算机可读存储介质

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108918426A (zh) * 2018-06-06 2018-11-30 北京云端光科技术有限公司 检测混合物的方法、装置、存储介质及电子设备
CN109073541B (zh) * 2018-08-01 2021-02-02 深圳达闼科技控股有限公司 一种物质检测数据库的数据添加方法、装置和检测设备
CN113196053A (zh) * 2018-12-20 2021-07-30 佳能株式会社 信息处理装置、信息处理装置的控制方法及程序
CN110596020A (zh) * 2019-08-30 2019-12-20 Oppo广东移动通信有限公司 物质检测方法及装置、存储介质、通信终端
WO2021177240A1 (ja) * 2020-03-05 2021-09-10 国立研究開発法人物質・材料研究機構 スペクトル汎化システム及び方法、並びに物質同定システム及び方法
CN113674797B (zh) * 2020-05-15 2024-05-14 复旦大学 基于蛋白质组的数据检测系统、方法、设备及存储介质
CN113358594B (zh) * 2020-06-30 2023-07-28 北京领主科技有限公司 基于光谱检测的物质成分分析系统、方法、装置及介质
CN115885296A (zh) * 2020-08-18 2023-03-31 株式会社岛津制作所 数据解析装置、数据解析方法、学习完毕模型的生成方法、系统以及程序

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101206463A (zh) * 2007-11-30 2008-06-25 深圳先进技术研究院 一种在线监控药物生产混合过程的方法及系统
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
CN101915744A (zh) * 2010-07-05 2010-12-15 北京航空航天大学 物质成分含量的近红外光谱无损检测方法及装置
WO2012045396A2 (de) * 2010-10-04 2012-04-12 Karlsruher Institut für Technologie Selektive proteinquantifizierung mittels multivariater auswertung von uv-absorptionsspektren
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
CN104182251A (zh) * 2014-08-12 2014-12-03 小米科技有限责任公司 农作物特征检测方法、装置及终端
CN106770013A (zh) * 2016-12-28 2017-05-31 天津农学院 基于二维近红外相关光谱不变矩特征判别掺杂尿素牛奶的方法

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6035246A (en) * 1994-11-04 2000-03-07 Sandia Corporation Method for identifying known materials within a mixture of unknowns
CA2331116A1 (en) * 2001-01-15 2002-07-15 Chenomx, Inc. Compound identification and quantitation in liquid mixtures -- method and process using an automated nuclear magnetic resonance measurement system
JP2008536144A (ja) * 2005-04-15 2008-09-04 ケミマジ コーポレーション 混合物をスペクトル分析する方法および装置
WO2007123555A2 (en) * 2005-07-14 2007-11-01 Chemimage Corporation Time and space resolved standoff hyperspectral ied explosives lidar detector
US9103714B2 (en) * 2009-10-06 2015-08-11 Chemimage Corporation System and methods for explosives detection using SWIR
US8271209B2 (en) * 2009-12-18 2012-09-18 General Electric Company Spectral searching method for substance identification
CN102608284B (zh) 2011-12-23 2014-11-05 南京工业大学 一种确定多元混合气体爆炸极限的方法
CN102590175B (zh) * 2012-02-21 2014-04-09 浙江大学 基于拉曼光谱叠加的甲醇汽油甲醇含量快速测定方法
CN102750333B (zh) * 2012-05-31 2014-05-07 华中科技大学 一种用于提取半导体纳米结构特征尺寸的方法
US20160178618A1 (en) * 2014-12-17 2016-06-23 Stc.Unm 3d tissue model for spatially correlated analysis of biochemical, physiological and metabolic micro-environments
US10677713B1 (en) * 2016-08-04 2020-06-09 Hrl Laboratories, Llc Adaptive gas analyzer

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101206463A (zh) * 2007-11-30 2008-06-25 深圳先进技术研究院 一种在线监控药物生产混合过程的方法及系统
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
CN101915744A (zh) * 2010-07-05 2010-12-15 北京航空航天大学 物质成分含量的近红外光谱无损检测方法及装置
WO2012045396A2 (de) * 2010-10-04 2012-04-12 Karlsruher Institut für Technologie Selektive proteinquantifizierung mittels multivariater auswertung von uv-absorptionsspektren
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
CN104182251A (zh) * 2014-08-12 2014-12-03 小米科技有限责任公司 农作物特征检测方法、装置及终端
CN106770013A (zh) * 2016-12-28 2017-05-31 天津农学院 基于二维近红外相关光谱不变矩特征判别掺杂尿素牛奶的方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110503141A (zh) * 2019-08-13 2019-11-26 Oppo（重庆）智能科技有限公司 酒类鉴别方法、装置及计算机可读存储介质

Also Published As

Publication number Publication date
US20190272449A1 (en) 2019-09-05
US11468264B2 (en) 2022-10-11
CN108064341A (zh) 2018-05-22
CN108064341B (zh) 2020-05-12

Similar Documents

Publication Publication Date Title
WO2018227338A1 (zh) 物质成分检测方法、装置和检测设备
CN108780048B (zh) 一种确定检测设备的方法、检测装置及可读存储介质
CN107104978B (zh) 一种基于深度学习的网络风险预警方法
CN108052881A (zh) 一种实时检测施工现场图像中多类实体对象的方法及设备
Abas et al. Agarwood oil quality classification using support vector classifier and grid search cross validation hyperparameter tuning
Barbon Jr et al. Machine learning applied to near‐infrared spectra for chicken meat classification
WO2019100346A1 (zh) 物质检测处理方法、装置和检测设备
US11188822B2 (en) Attendee engagement determining system and method
CN108235733B (zh) 基于拉曼光谱的物质识别方法及云端系统
US11941088B1 (en) Image processing of an environment to select an action to be performed by an agent interacting with the environment
US10408760B2 (en) Raman spectrum detecting method for eliminating package interference and electronic device thereof
CN106323889B (zh) 基于云服务器的质量检测系统
CN108985998A (zh) 一种智慧安检方法
CN111881706B (zh) 活体检测、图像分类和模型训练方法、装置、设备及介质
WO2018232632A1 (zh) 物质检测方法、装置和检测设备
CN108713137B (zh) 一种物质检测方法、检测终端及计算机可读存储介质
Self et al. Determination of decomposition in Salmon products by mass spectrometry with sensory‐driven multivariate analysis
Saar et al. Cross-browser testing in browserbite
Sari et al. Assessment of color levels in leaf color chart using smartphone camera with relative calibration
US20200379890A1 (en) Method for testing software, and computing device and computer-readable storage medium thereof
CN106446902B (zh) 非文字图像识别方法和装置
CN109073541B (zh) 一种物质检测数据库的数据添加方法、装置和检测设备
CN104951490A (zh) 一种适用于元素分析的能谱知识库应用方法
dos Santos et al. Automated detection of grapes and leaves in viticulture with a YOLOv2 neural network
Khaidarov Existence and uniqueness of fixed points of integral operator of Hammerstein type and its applications

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: 17913845

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 04.06.20)

122 Ep: pct application non-entry in european phase

Ref document number: 17913845

Country of ref document: EP

Kind code of ref document: A1