CN107991283B - Raman spectrum detection device and Raman spectrum detection method - Google Patents

Abstract

Embodiments of the present invention provide a Raman spectrum detection device and a Raman spectrum detection method. The Raman spectrum detection equipment includes: an excitation light source configured to emit excitation light to the sample to be tested; an optical device having a spectrum detection light path and a predetermined dangerous area imaging light path, and the spectrum detection light path is configured to collect the excitation light from the sample to be tested. The optical signal of the position, the predetermined dangerous area imaging light path is configured to capture the image of the predetermined dangerous area; the spectrometer is configured to receive the optical signal from the spectrum detection optical path and generate the Raman spectrum of the sample to be measured from the optical signal; the face recognition device is configured To receive an image of a predetermined dangerous area and identify whether a human face is included in the image of the predetermined dangerous area; and a safety controller configured to receive a recognition result of the face recognition device and include the face in the image of the predetermined dangerous area when the face recognition device recognizes In the case of human faces, the excitation light source is turned off.

Description

Raman spectrum detection equipment and Raman spectrum detection method

Technical field

Embodiments of the present invention relate to the field of Raman spectrum detection, and in particular to a Raman spectrum detection equipment and a Raman spectrum detection method.

Background technique

Raman spectroscopic analysis technology is a non-contact spectroscopic analysis technology based on the Raman scattering effect, which can conduct qualitative and quantitative analysis of the components of substances. Raman spectrum is a kind of molecular vibration spectrum, which can reflect the fingerprint characteristics of molecules and can be used to detect substances. Raman spectroscopy detects and identifies substances by detecting the Raman spectrum produced by the Raman scattering effect of the object to be measured on the excitation light. Raman spectroscopy detection methods have been widely used in liquid security inspection, jewelry detection, explosives detection, drug detection, pharmaceutical detection, pesticide residue detection and other fields.

In recent years, Raman spectroscopy analysis technology has been widely used in fields such as dangerous goods inspection and substance identification. In the field of substance identification, due to the different colors and shapes of various substances, people usually cannot accurately judge the properties of substances. The Raman spectrum is determined by the molecular energy level structure of the object being tested, so the Raman spectrum can be used as a "fingerprint" of the substance. ” information for substance identification. Therefore, Raman spectroscopy analysis technology is widely used in customs, public security, food and medicine, environment and other fields.

Since Raman spectroscopy often requires the use of high-power-density lasers as excitation light sources, which may have strong thermal effects, when the sample is unknown, hasty detection may cause the sample to be damaged by laser ablation, or even cause the laser to ignite. Or detonate some flammable and explosive chemicals, causing losses to people and property. Moreover, in some cases, such as the excitation light emitted by a Raman spectrometer detector, it may also harm the human eye.

Contents of the invention

The present invention aims to propose a Raman spectrum detection equipment and a Raman spectrum detection method, which can effectively reduce or avoid the risk of personnel being injured during Raman spectrum detection.

Embodiments of the present invention provide a Raman spectrum detection equipment, including: an excitation light source configured to emit excitation light to a sample to be measured; an optical device having a spectrum detection light path and a predetermined dangerous area imaging light path, so The spectrum detection light path is configured to collect light signals from the position where the sample to be measured is illuminated by the excitation light, and the predetermined dangerous area imaging light path is configured to capture an image of the predetermined dangerous area; the spectrometer is configured to receive signals from the spectrum detection light path. an optical signal and generate a Raman spectrum of the sample to be tested from the optical signal; a face recognition device configured to receive the image of the predetermined dangerous area and identify whether the image of the predetermined dangerous area includes a human face; and safety A controller configured to receive the recognition result of the face recognition device and to turn off the excitation light source when the face recognition device recognizes that the image of the predetermined dangerous area includes a human face.

In one embodiment, the predetermined dangerous area includes an area illuminated by the excitation light or to be illuminated by the excitation light.

In one embodiment, the predetermined danger area includes the surrounding environment area of the optical device or the sample to be measured.

In one embodiment, the spectrum detection light path includes a collection lens, a spectroscope, a Raman filter set and a coupling lens in sequence, and the predetermined dangerous area imaging light path includes a collection lens, a spectroscope, a lens set and an image capture tool, wherein , the collection lens and the beam splitter are common components of the spectrum detection optical path and the predetermined dangerous area imaging optical path.

In one embodiment, the spectral detection light path and the predetermined dangerous area imaging light path are separated from each other.

In one embodiment, the face recognition device includes: an image feature extraction module configured to extract features of the image and generate a feature vector; a classifier model generation module, the classifier model generates a module configured to generate a classifier model based on a reference face image; and a recognition module configured to use the feature vector as an input vector of the classifier model to determine the image collected by the predetermined dangerous area imaging optical path Does it contain facial features?

Embodiments of the present invention also provide a Raman spectrum detection method, which includes: taking an image of a predetermined dangerous area; identifying whether the image of the predetermined dangerous area includes a human face; and identifying the image of the predetermined dangerous area. When a human face is included in the image, the excitation light source is turned off. When it is recognized that the image of the predetermined dangerous area does not include a human face, the excitation light source is started and the Raman spectrum of the sample is collected to analyze the sample. Perform testing.

In one embodiment, before identifying whether the image of the predetermined dangerous area includes a human face, the method further includes: generating a classifier model based on a reference face image.

In one embodiment, identifying whether the image of the predetermined dangerous area includes a human face includes: extracting features of the image and generating a feature vector; and using the feature vector as an input vector of the classifier model to Determine whether the image collected by the predetermined dangerous area imaging light path contains facial features.

With the help of the Raman spectrum detection equipment and the sample safety detection method of Raman spectrum detection according to the above embodiments, detection risks caused by the excitation light igniting, ablating or detonating the sample during the spectrum detection process can be reduced or prevented.

Description of the drawings

Figure 1 shows a schematic diagram of a Raman spectrum detection device according to an embodiment of the present invention;

Figure 2 shows a schematic diagram of a Raman spectrum detection device according to an embodiment of the present invention;

Figure 2a shows a schematic diagram of the equivalent optical path of the predetermined dangerous area imaging optical path in the Raman spectrum detection device according to an embodiment of the present invention;

Figure 3 shows a flow chart of a Raman spectrum detection method according to an embodiment of the present invention;

Figure 4 shows an exemplary flow chart of image feature vector extraction in a Raman spectrum detection method according to an embodiment of the present invention;

Figure 5 shows an exemplary flow chart for establishing a classifier model in a Raman spectrum detection method according to an embodiment of the present invention; and

Figure 6 shows an exemplary flow chart of feature extraction in a Raman spectrum detection method according to an embodiment of the present invention.

The drawings do not show all circuits or structures in the Raman spectrum detection device according to embodiments of the present invention. The same reference numbers throughout the drawings refer to the same or similar parts or features.

Detailed ways

The technical solution of the present invention will be further described in detail below through examples and in conjunction with the accompanying drawings. In the specification, the same or similar reference numerals represent the same or similar components. The following description of the embodiments of the present invention with reference to the accompanying drawings is intended to explain the general inventive concept of the present invention and should not be understood as a limitation of the present invention.

Additionally, in the following detailed description, for convenience of explanation, numerous specific details are set forth to provide a comprehensive understanding of the disclosed embodiments. It will be apparent, however, that one or more embodiments may be practiced without these specific details. In other instances, well-known structures and devices are illustrated in order to simplify the drawings.

An embodiment of the present invention provides a Raman spectrum detection device 100. As shown in FIG. 1 , the Raman spectrum detection equipment 100 includes: an excitation light source 10 , an optical device 20 , a spectrometer 30 , a face recognition device 40 and a security controller 60 . The excitation light source 10 may include, for example, a laser and is configured to emit excitation light to the sample 50 to be measured. The optical device 20 may include a spectrum detection light path 21 and a predetermined dangerous area imaging light path 22 . The spectrum detection light path 21 is configured to collect light signals from the position where the sample to be measured 50 is illuminated by the excitation light. The collected optical signals may be transmitted to spectrometer 30. The spectrometer 30 can be configured to receive the optical signal from the spectrum detection optical path 21 and generate the Raman spectrum of the sample to be measured 50 from the optical signal, thereby realizing the normal process of Raman spectrum detection. The predetermined dangerous area imaging optical path 22 is configured to capture an image of the predetermined dangerous area illuminated or to be illuminated by the excitation light, and the image is transmitted to the face recognition device 40 . The face recognition device 40 may be configured to determine whether there is a face in the predetermined danger area based on the image of the predetermined danger area captured by the predetermined danger area imaging light path 22 .

As mentioned before, using excitation light to collect optical signals is the basic step for Raman spectroscopy detection of the sample to be tested. The excitation light itself has a certain energy. For samples of certain materials to be tested, it may be different from the excitation light. A reaction occurs that results in a change in the composition of the sample. For example, some flammable and explosive substances may be ignited, ablated, exploded, etc. under the action of excitation light. In practice, the composition of the sample to be tested is often unknown, so if a person's face is located close to the sample to be tested (such as an optical device or the surrounding environment area of the sample to be tested), it may be harmed. For another example, the excitation light itself may also cause damage to human eyes. Therefore, before or during the emission of the excitation light, it is also possible to detect whether there is a human face in the area illuminated by the excitation light or that may be illuminated by the excitation light. avoid risk. This is especially important for handheld Raman spectroscopy detection equipment.

In the Raman spectrum detection device according to the embodiment of the present invention, the safety controller 60 may be configured to receive the recognition result of the face recognition device 40 and to include a person in the image in which the face recognition device 40 recognizes the predetermined danger area. In the face case, the excitation light source is turned off. In this way, personal injury can be prevented during the inspection process.

In an example, as shown in FIG. 1 , the spectrum detection optical path 21 may include a collection lens 31 , a beam splitter 32 , a Raman filter set 33 and a coupling lens 34 in sequence. As an example, the Raman filter set 33 may include one or more filters for filtering undesired light such as Rayleigh scattered light and excitation light while retaining the Raman scattered light signal. The coupling lens 34 may be used to couple the optical signal filtered by the Raman filter set 33 to the spectrometer 30 . The predetermined dangerous area imaging optical path 22 may include a lens group 23 and an image capturing tool 24 . The image capturing tool 24 may be, for example, a camera (such as a CCD camera, etc.), a camera, or other device used for capturing images. The lens group 23 may include one or more lenses, which may be separate from the image capturing tool 24 or integrated with the image capturing tool 24 . The lens group 23 can be composed of any number of lenses, or any lens group known in the art that can achieve clear imaging. The predetermined dangerous area imaging light path 22 shown in FIG. 1 is schematic, and the predetermined dangerous area that is desired to be imaged can be selected as needed. The "predetermined dangerous area" mentioned here refers to an area where potential danger may occur if a person's face is in this area during the excitation light signal collection process. For example, the predetermined danger area may include an area illuminated or to be illuminated by the excitation light, or an area of the surrounding environment including the optical device or the sample to be measured. The surrounding environment area of the optical device or the sample to be measured refers to the location in this area where the distance from the optical device or the sample to be measured is less than the safe distance (the safety distance can be determined based on the energy of possible ignition, explosion, and laser radiation). By performing face recognition on images of these predetermined dangerous areas, it can be determined whether there is a risk of personal injury.

As an example, when imaging in a dark environment, it may be necessary to provide a fill light and other devices to provide sufficient light intensity for capturing images of the sample to be tested to improve the clarity of the image.

In the above example shown in FIG. 1 , the spectrum detection light path 21 and the predetermined dangerous area imaging light path 22 are separated from each other. In another example, the spectrum detection light path 21 and the predetermined dangerous area imaging light path 22 may also overlap to a certain extent. For example, as shown in FIG. 2 , the collection lens 31 and the beam splitter 32 are common components of the spectrum detection optical path 21 and the predetermined dangerous area imaging optical path 22 . This helps to improve the compactness of the optical device to save space.

In the Raman spectrum detection device 100' according to another embodiment of the present invention shown in Figure 2, the spectrum detection light path 21 and the predetermined dangerous area imaging light path 22 have a common part close to the sample to be measured 50, and the spectroscope 32 allows The light signal from the sample to be measured 50 is transmitted through to the spectrometer, thereby forming a spectrum detection light path 21 , and the beam splitter 32 reflects the light beam with the image to form a predetermined dangerous area imaging light path 22 . It should be noted that although the spectral detection light path 21 and the predetermined dangerous area imaging light path 22 have a common part, this does not mean that the operations of spectral detection and imaging of the predetermined dangerous area need to be performed at the same time, for example, the operation of imaging the predetermined dangerous area. It is usually carried out before spectrum detection to eliminate safety hazards before spectrum detection to avoid danger during the spectrum detection process. In addition, the spectrum detection light path 21 and the predetermined dangerous area imaging light path 22 share the collection lens 31, which can better image the area that may be illuminated by the excitation light.

It should be noted that FIG. 2 shows the progress of the light beam in the optical device 20 when the excitation light source 10 emits excitation light. However, in practice, when imaging a predetermined dangerous area, the excitation light is often The light source does not emit excitation light, and there is not even a sample 50 to be measured. At this time, the equivalent optical path of the predetermined dangerous area imaging optical path 22 is shown in Figure 2a. The collection lens 31 and the lens group 23 can project the image of the side facing the sample 50 to be tested to the image capturing tool 24, and capture the image. The lens 26 of the tool 24 forms the image on the image plane (such as the plane of the film, CCD, etc.). This can easily obtain an image of the area in front of the collection lens 31 , which can avoid the risk caused by the human eye being located near the detection position of the sample to be tested 50 .

As an example, the optical path through which the excitation light source 10 emits excitation light may partially overlap with the spectrum detection optical path 21. For example, in the examples shown in FIG. 1 and FIG. 2, another beam splitter 35 may be provided, and the beam splitter 35 may be The excitation light emitted by the excitation light source 10 is reflected, and the reflected excitation light is concentrated on the sample to be measured 50 through the collection lens 31 , which helps to simplify the adjustment of the optical path. However, the embodiments of the present invention are not limited thereto. For example, excitation The light emitted by the light source 10 can illuminate the sample to be measured through a light path that is completely independent from the spectrum detection light path. As an example, after the excitation light is emitted from the excitation light source 10, it can also pass through a stray light filter 36, which can be used to remove stray light to improve the signal-to-noise ratio of the excitation light.

The above example only provides an exemplary implementation of a Raman spectrum detection device, but the embodiments of the present invention are not limited thereto. Other Raman spectrum detection methods that those skilled in the art can predict after reading this disclosure Alternative means of equipment are also possible.

As an example, the face recognition device 40 may include: an image feature extraction module 41 configured to extract features of the image and generate a feature vector; a classifier model generation module 42, the classifier model generation module Configured to generate a classifier model based on a reference face image; and a recognition module 43 configured to use the feature vector as an input vector of the classifier model to determine the image collected by the predetermined dangerous area imaging optical path Does it contain facial features?

In the above embodiment, the face recognition device 40 and the security controller 60 can be implemented by a processor, or can also be implemented by other software and hardware structures. As an example, the safety controller 60 can also be implemented by a trigger switch.

An embodiment of the present invention also discloses a Raman spectrum detection method S100. As shown in Figure 3, the method may include:

Step S10: Take images of the predetermined dangerous area;

Step S20: Identify whether the image of the predetermined dangerous area includes a human face; and

Step S30: When it is recognized that the image of the predetermined dangerous area includes a human face, the excitation light source is turned off, and when it is recognized that the image of the predetermined dangerous area does not include a human face, start Excite the light source and collect the Raman spectrum of the sample to detect the sample.

The following is an exemplary introduction to the steps of identifying whether a human face is included in the image of the predetermined dangerous area with reference to Figures 4 to 6:

Figure 4 shows an example of the image feature extraction process. For face recognition, you need to first have training samples, that is to say, known images with faces, here called reference face images. By extracting image features from the training samples and forming a sample feature vector, a classifier model can be established. In practice, the image features collected by the predetermined dangerous area imaging optical path 22 can also be extracted to form a feature vector of the test image. , and calculate the matching degree with the model based on the classifier model to determine whether there are facial features in the actually collected test images.

The flowchart shown in Figure 4 includes step S41, which is the step of collecting training samples. Collecting training samples can randomly collect images through the Raman spectrum detection equipment according to embodiments of the present invention, but in this case, the collected training samples need to contain facial features, that is, in the predetermined dangerous area When collecting images, there should be a face in the predetermined danger zone. Alternatively, training samples can also be collected directly from a known (face image) sample library.

After the training sample images are collected, step S42 is executed, that is, feature extraction is performed. As an example, image dense features can be selected during image feature extraction. In this example, the SIFT feature, which has superior performance and is widely used in the field of image recognition, is used (SIFT feature is known in the art and will not be described again here). The feature extraction steps are as follows, as shown in Figure 6:

① Count the number of collected training sample images. In this example, the number is N ₁ .

② Determine the step length λ in the feature extraction process, and the sliding window size is α (where λ and α are both natural numbers);

③Input the training sample image k, including the length (i.e., the number of pixel rows of the image (img.cols)) and width (i.e., the number of pixel columns (img.rows)) of the recorded image, which are m and n respectively, and the juxtaposition i = 0,j=0;

④Extract SIFT features from (i,j) to (i+α,j+α) image patches (patch) in the image, and record the SIFT features as f;

⑤ Increase j by the step length λ and determine whether j+α≤n is true. If true, repeat steps ②③④, otherwise jump to step ⑥;

⑥Set j to zero, increase i by the step length λ, and determine whether j+α≤m is true. If established, repeat steps ②③④ until i>m completes the extraction of dense features (SIFT features) F _k = [f ₁ , f ₂ ,..., f _Numk ] of the training sample image k, where Numk = ((n-α )/λ)*((m-α)/λ), k=1,2,…,N ₁ .

After completing the dense feature extraction, step S43, namely dictionary construction and feature encoding, needs to be performed. The dictionary construction process can include:

Randomly extract the SIFT features F = [F ₁ , F ₂ ,..., F _N1 ] of the training sample image, where F _k = [f ₁ , f ₂ ,..., f _Numk ], recorded as M = [f ₁ , f ₂ ,…,f _m ], where selected

The image feature M is used for clustering processing, and the number of cluster centers is selected as h. For example, the known K-means clustering algorithm can be used to obtain the feature center vector, thereby generating a dictionary D, D=[d ₁ , d ₂ , ...,d _h ].

The feature encoding process can include:

Input the dense SIFT feature F _k of the training sample image k = [f ₁ , f ₂ ,…, f _Numk ];

According to the following formula (1), the SIFT features of the training sample image k are projected and reconstructed, and c _i is recorded, which is the feature vector of the training sample image, that is, the classifier input vector.

Among them, "st" represents the constraint condition. For N ₁ training sample images, N=N ₁ .

In order to establish a classifier model, the feature vector of the above training sample image is used as the input sample. The linear support vector machine model known in the art can be selected as the classifier for training to obtain the classifier model of the training sample, as shown in Figure 5 . The training sample label in Figure 5 refers to the label corresponding to the training sample image.

As an example, to better confirm the correctness of the classifier model, model validation can be performed. For model verification, you can use a model verification sample image similar to the training sample image. Follow the above steps to extract the feature vector of the training sample image, perform dense feature extraction on the model verification sample image, and project the feature dictionary to form a model verification sample. Feature vector of the image. The feature vector of the model verification sample image can be used as an input vector for classifier model verification to verify whether the classification output is correct, and statistically verify the accuracy of the output, that is, the recall rate of the model. As an example, the ratio of the number of training sample images to the number of model validation sample images can be 5:1. Those skilled in the art should understand that the above model verification process is only an exemplary step provided for optimizing the algorithm, and embodiments of the present invention may not include this model verification process.

As an example, when determining whether the actual image collected by the predetermined dangerous area imaging light path 22 includes a human face, the feature vector of the actual image (such as including Dense features, projection dictionaries, feature encoding, etc.), and then bring the feature vector of the actual image as an input vector into the above classifier model to determine its matching degree with the classifier model, that is, to determine whether the actual image includes a face.

Those skilled in the art should understand that the above steps regarding face recognition are only exemplary, and the embodiments of the present invention are not limited thereto. Other methods known in the art that can achieve face recognition can be used in the embodiments of the present invention. .

In the Raman detection method according to an embodiment of the present invention, before identifying whether the image of the predetermined dangerous area includes a human face, the method may further include:

Step S00: Generate a classifier model based on the reference face image.

The specific exemplary process has been given above and will not be described again here.

In the Raman detection method according to an embodiment of the present invention, the above step S20 may include:

Step S21: Extract features of the image and generate feature vectors; and

Step S22: Use the feature vector as an input vector of the classifier model to determine whether the image collected by the predetermined dangerous area imaging optical path contains facial features.

The specific exemplary process has been given above and will not be described again here.

The step of identifying whether the image of the predetermined dangerous area includes a human face in the Raman detection method according to the embodiment of the present invention can be correspondingly performed by the face recognition device in the Raman detection device according to the embodiment of the present invention. 40 modules to execute.

The above detailed description has explained numerous embodiments of the above-described Raman spectroscopy detection apparatus and methods through the use of schematic diagrams, flow charts, and/or examples. Where such diagrams, flowcharts, and/or examples include one or more functions and/or operations, those skilled in the art will understand that each function and/or operation in such diagrams, flowcharts, or examples may Individually and/or jointly implemented by various structures, hardware, software, firmware or essentially any combination thereof. In one embodiment, portions of the subject matter described in embodiments of the present invention may be implemented via an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), or other integrated formats. However, those skilled in the art will recognize that some aspects of the embodiments disclosed herein may equivalently be implemented, in whole or in part, in an integrated circuit, as one or more computers running on one or more computers. Computer program (e.g., implemented as one or more programs running on one or more computer systems), implemented as one or more programs running on one or more processors (e.g., implemented as one or more programs running on one or more processors) One or more programs running on multiple microprocessors), implemented as firmware, or substantially implemented as any combination of the above methods, and those skilled in the art will have the ability to design circuits and/or write software and /or firmware code capabilities. Furthermore, those skilled in the art will recognize that the mechanisms of the subject matter of the present disclosure can be distributed as a program product in a variety of forms, and that regardless of the specific type of signal-bearing medium actually used to perform the distribution, the mechanisms of the subject matter of the present disclosure can The exemplary embodiments are applicable. Examples of signal bearing media include, but are not limited to: recordable media, such as floppy disks, hard drives, optical disks (CD, DVD), digital tapes, computer memory, etc.; and transmission media, such as digital and/or analog communications media (e.g. , optical fiber cables, waveguides, wired communication links, wireless communication links, etc.).

Unless there are technical obstacles or conflicts, the above-mentioned various embodiments of the present invention can be freely combined to form additional embodiments, and these additional embodiments are all within the protection scope of the present invention.

Although the present invention has been described in conjunction with the accompanying drawings, the embodiments disclosed in the drawings are intended to illustrate preferred embodiments of the present invention and should not be construed as a limitation of the present invention. The dimensional proportions in the drawings are only schematic and are not to be construed as limitations of the present invention.

While some embodiments of the present general inventive concept have been shown and described, those of ordinary skill in the art will understand that changes may be made in these embodiments without departing from the principles and spirit of the present general inventive concept. The scope is defined by the claims and their equivalents.

Claims (8)

1. A Raman spectrum detection equipment, including: An excitation light source configured to emit excitation light to the sample to be measured; Optical device, the optical device has a spectral detection light path and a predetermined dangerous area imaging light path, the spectral detection light path is configured to collect light signals from the position where the sample to be measured is illuminated by the excitation light, the predetermined dangerous area imaging light path is configured To capture images of predetermined dangerous areas; A spectrometer configured to receive an optical signal from the spectrum detection optical path and generate a Raman spectrum of the sample to be measured from the optical signal; a face recognition device configured to receive an image of the predetermined dangerous area and identify whether the image of the predetermined dangerous area includes a human face; and A safety controller configured to receive the recognition result of the face recognition device and to turn off the excitation light source when the face recognition device recognizes that the image of the predetermined dangerous area includes a human face, wherein the person Face recognition devices include: An image feature extraction module, the image feature extraction module is configured to extract features of the image of the predetermined dangerous area and generate a feature vector, the image feature extraction module includes: A device for collecting images of the predetermined danger area; Devices for dense feature extraction; means for constructing projection dictionaries and feature encodings; and means for generating a feature vector of an image of said predetermined danger zone, Wherein, the predetermined dangerous area includes an area illuminated by the excitation light or includes an area to be illuminated by the excitation light. 2. The Raman spectrum detection device according to claim 1, wherein the predetermined danger area includes the optical device or a surrounding environment area including the sample to be measured. 3. The Raman spectrum detection equipment according to claim 1, wherein the spectrum detection light path includes a collection lens, a spectroscope, a Raman filter set and a coupling lens in sequence, and the predetermined dangerous area imaging light path includes a collection lens, Spectroscope, lens group and image capturing tool, wherein the collection lens and spectroscope are common components of the spectrum detection optical path and the predetermined dangerous area imaging optical path. 4. The Raman spectrum detection device according to claim 1, wherein the spectrum detection light path and the predetermined dangerous area imaging light path are separated from each other. 5. The Raman spectrum detection device according to any one of claims 1 to 4, wherein the face recognition device further includes: a classifier model generation module configured to generate a classifier model based on the reference face image; and An identification module configured to use the feature vector as an input vector of the classifier model to determine whether the image collected by the predetermined dangerous area imaging optical path contains facial features. 6. A Raman spectrum detection method, including: Take images of predetermined danger areas; Identify whether the image of the predetermined danger area includes a human face; and When it is recognized that the image of the predetermined dangerous area includes a human face, the excitation light source is turned off; when it is recognized that the image of the predetermined dangerous area does not include a human face, the excitation light source is turned on. , collect the Raman spectrum of the sample to detect the sample, Wherein, identifying whether a human face is included in the image of the predetermined dangerous area includes: Extracting features of the image of the predetermined dangerous area and generating a feature vector. Extracting features of the image and generating a feature vector includes: Collect images of the predetermined danger area; Dense feature extraction; Projection dictionary and feature encoding; and generating a feature vector of the image of the predetermined danger area, Wherein, the predetermined dangerous area includes an area illuminated by the excitation light or includes an area to be illuminated by the excitation light. 7. The method of claim 6, wherein before identifying whether the image of the predetermined danger area includes a human face, the method further includes: Generate a classifier model based on a reference face image. 8. The method according to claim 7, wherein identifying whether the image of the predetermined dangerous area includes a human face further includes: The feature vector is used as an input vector of the classifier model to determine whether the image collected by the predetermined dangerous area imaging optical path contains facial features.