Terahertz-Raman integrated detection system and method

Abstract

The invention relates to a terahertz-Raman integrated detection system and method. The detection system comprises a fusion mainframe box and an integrated probe arranged at a sample detection end on the fusion mainframe box, a Raman spectrum module, a terahertz spectrum module, a terahertz optical module and a data analysis module are arranged in the fusion mainframe box; the Raman spectrum module is composed of a Raman laser, a Raman light filter, a Raman grating and a Raman detector which are sequentially connected through a light path; the terahertz spectrum module is composed of a terahertz femtosecond laser, a transmitting photoconductive antenna, a receiving photoconductive antenna and a terahertz detector which are sequentially connected through a light path; the data analysis module is in communication connection with the Raman detector and the terahertz detector and is used for synchronously processing data transmitted by the Raman detector and the terahertz detector. Compared with the prior art, the integrated probe synchronously collects two kinds of spectral data and carries out fusion processing on the two kinds of spectral data through the data analysis module, and the device has the remarkable effects that the device integration degree is high, the detection efficiency and accuracy are greatly improved, and the application range is wide.

Description

Technical Field

[0001] The present invention relates to the field of spectrometer equipment and technology, and in particular to a terahertz-Raman integrated detection system and method. Background Art

[0002] With the rapid development of the global economy and the accelerating pace of urbanization, chemicals and their products have permeated every aspect of human life, becoming inextricably linked to our daily lives, including food, clothing, housing, and transportation. Hazardous chemicals, as specialized commodities, pose numerous safety risks during production, storage, transportation, use, and disposal due to their flammable, explosive, toxic, and corrosive properties. These hazards can easily lead to fires, explosions, poisoning, and leaks, among other hazardous chemical accidents.

[0003] The occurrence of hazardous chemical accidents can be attributed to the lack of information technology for real-time monitoring of the safety status of chemical storage, and the serious lack of dynamic monitoring and early warning capabilities for on-site chemical risks; as well as the serious lack of rapid qualitative analysis of chemicals, precise non-destructive non-contact testing, and trace identification and detection technologies during the on-site disposal of chemical accidents.

[0004] With the continuous advancement of science and technology, hazardous chemical detection technologies and instruments are becoming increasingly common. In recent years, the miniaturization and portability of instruments and equipment have further facilitated the rapid on-site detection of hazardous chemicals. Modern spectroscopic rapid detection methods, such as ion mobility spectrometry, infrared spectroscopy, Raman spectroscopy, and terahertz spectroscopy, can directly analyze the physical and chemical properties of hazardous chemicals, enabling rapid identification. These methods are the simplest and most direct emergency monitoring technologies, offering high accuracy and no contamination or harm to the objects being detected.

[0005] Raman spectroscopy can directly perform non-destructive testing on unknown solid and liquid substances through translucent containers, and obtain the spectrum of the unknown substance and the corresponding test results in seconds. It is suitable for detecting most inorganic and organic hazardous chemicals, but it is easy to produce fluorescence effects during the spectrum acquisition process; Terahertz spectroscopy can penetrate the external packaging materials of most substances and realize on-site non-opening box detection of hazardous chemicals. It has unique advantages such as non-contact, real-time monitoring, and high accuracy, but the existing spectral library data is not perfect.

[0006] Among existing technologies for on-site detection of hazardous chemicals, Raman spectroscopy is the most widely used and has the most comprehensive database. It is suitable for the non-contact, rapid qualitative identification of liquid and solid chemicals. Terahertz spectroscopy, however, has strong penetration and is unaffected by target color, making it suitable for the detection of gaseous and liquid chemicals. It can perform non-contact detection and analysis of liquid substances in containers such as plastic and glass. Therefore, the combined Raman-terahertz spectroscopy method is the most effective universal detection method for various chemicals. This invention combines Raman and terahertz spectroscopy to effectively expand the types of detectable targets and their application scenarios.

[0007] Currently, there are two aspects regarding the combination of terahertz and Raman spectroscopy: (1) Instrument hardware combination. Existing patent document 202011547050.3 is a terahertz Raman spectrometer. The main body of the spectrometer is still a Raman spectrometer, but it carries a terahertz band. It provides a terahertz Raman spectrometer with high spectral resolution and wide-band weak signal detection. Existing patent document 202310500338.2 is a terahertz near-field microscope system based on a Raman spectrometer and its use method. The terahertz near-field microscope is used to process the terahertz near-field signal scattered by the sample to be tested, and to form a terahertz near-field spectrum image. It is not a true integrated combination of terahertz and Raman spectroscopy. (2) Instrument application. References Su Tongfu, Wang Changqing, Zhao Guozhong, et al., Study on terahertz and Raman spectra of L-arabinose fingerprint region [J], Spectroscopy and Spectral Analysis, 2018, 38(9): 2713-2719 and Lu Meihong, Gong Peng, Zhang Fan, et al., Study on terahertz and Raman spectra of disodium ethylenediaminetetraacetate [J], Spectroscopy and Spectral Analysis, 2020, 40(9): 2707-2712. Both used separate terahertz and Raman spectra to characterize L-arabinose and disodium ethylenediaminetetraacetate, and neither used terahertz and Raman spectroscopy in an integrated manner.

[0008] In view of the above analysis, the problems that need to be urgently solved in the existing technology are, on the one hand, that terahertz is only used as an auxiliary device for Raman spectroscopy, and is not used independently, and it is impossible to simultaneously collect terahertz spectral signals and Raman spectral signals of the sample to be tested; on the other hand, terahertz spectroscopy and Raman spectroscopy are used separately to characterize chemicals, and it is not an integrated application of terahertz spectroscopy and Raman spectroscopy. Summary of the Invention

[0009] The purpose of the present invention is to overcome the defects of the above-mentioned existing technologies and provide a terahertz-Raman integrated detection system and method, which realizes the simultaneous acquisition of two spectral data by an integrated probe and the fusion processing by a data analysis module, with the remarkable effects of high equipment integration, greatly improved detection efficiency and accuracy, and a wide range of applications.

[0010] The purpose of the present invention can be achieved by the following technical solutions:

[0011] A first aspect of the present invention provides a terahertz-Raman integrated detection system, comprising a fusion mainframe and an integrated probe provided at a sample detection end on the fusion mainframe;

[0012] The fusion main box is equipped with a Raman spectroscopy module, a terahertz spectroscopy module, a terahertz optical module, and a data analysis module, wherein:

[0013] Raman spectroscopy module: includes a Raman laser, a Raman filter, a Raman grating, and a Raman detector connected in sequence by optical paths;

[0014] Terahertz spectroscopy module: a terahertz femtosecond laser, a transmitting photoconductive antenna, a receiving photoconductive antenna, and a terahertz detector connected in sequence by optical paths;

[0015] Terahertz optical module: includes an off-axis parabolic mirror and a terahertz lens located at the sample detection end, used to guide the terahertz wave to the sample to be tested and receive the return signal;

[0016] Data analysis module: connected to the Raman detector and the terahertz detector for synchronously processing the data transmitted by the Raman detector and the terahertz detector.

[0017] Furthermore, the terahertz optical module includes two off-axis parabolic mirrors and two terahertz lenses;

[0018] An off-axis parabolic mirror and a terahertz lens are matched correspondingly on the transmitting photoconductive antenna side and the receiving photoconductive antenna side.

[0019] Furthermore, the optical path configuration of the terahertz optical module includes:

[0020] The off-axis parabolic mirror reflects the terahertz wave generated by the terahertz femtosecond laser and the transmitting photoconductive antenna, and then passes through the terahertz lens and the integrated probe to the sample to be measured;

[0021] The terahertz wave reflected by the sample to be tested passes through the integrated probe, terahertz lens, off-axis parabolic mirror in sequence to the receiving photoconductive antenna.

[0022] Furthermore, the laser light generated by the Raman laser passes through the integrated probe to reach the sample to be measured, and the Raman scattering signal reflected by the sample to be measured passes through the integrated probe and the Raman filter to the Raman grating.

[0023] Furthermore, the data analysis module synchronously receives and processes the original spectral signals transmitted by the Raman detector and the terahertz detector to generate a dual-spectrum fusion data packet, which contains a time-aligned Raman spectrum and a terahertz absorption spectrum.

[0024] A second aspect of the present invention provides a terahertz-Raman integrated detection method using the above-mentioned detection system, comprising the following steps:

[0025] S1: The laser light generated by the Raman laser irradiates the sample to be measured through the integrated probe. The generated Raman scattering signal returns through the same probe and enters the Raman detector through the Raman filter and Raman grating in sequence.

[0026] S2: The terahertz femtosecond laser drives the transmitting photoconductive antenna to generate terahertz waves. After being reflected by the off-axis parabolic mirror and focused by the terahertz lens, the terahertz waves are incident on the sample to be tested through the integrated probe. The reflected terahertz waves pass through the terahertz lens and the off-axis parabolic mirror, and are captured by the receiving photoconductive antenna and transmitted to the terahertz detector.

[0027] S3: The data analysis module synchronously receives the spectral data from the Raman detector and the terahertz detector and performs time axis alignment;

[0028] S4: The data analysis module performs synchronous feature extraction on Raman spectral data and terahertz absorption spectral data, constructs a time-correlated dual-spectrum fusion data packet, identifies the features in the dual-spectrum fusion data packet based on a pre-trained convolutional neural network, and outputs the joint analysis results.

[0029] Furthermore, in S2, the data analysis module sends a synchronous trigger signal to the Raman detector and the terahertz detector simultaneously, so that the two detectors collect data with the same timestamp.

[0030] Furthermore, in S3, time axis alignment includes timing compensation operations, specifically including:

[0031] Based on the transmission time difference of the terahertz wave in the optical path, the Raman spectrum data is time-shifted and corrected to align the starting time points of the two spectra.

[0032] Furthermore, in S4, the synchronous feature extraction includes:

[0033] Baseline correction and peak position extraction are performed on Raman spectral data, while dielectric constant inversion and characteristic peak fitting are performed on terahertz absorption spectral data to generate a dual-spectrum fusion data package containing molecular bond vibration frequencies and lattice vibration modes.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1) The present invention adopts a highly integrated optical path and module design, integrating the Raman spectroscopy module, terahertz spectroscopy module, terahertz optical module, and data analysis module into a main chassis, and is equipped with an integrated probe. This greatly reduces the size of the overall hardware equipment, eliminates the need for an external computer, enhances the portability and convenience of the equipment, and provides strong support for rapid on-site detection. It is particularly suitable for on-site detection of hazardous chemicals, etc., and can quickly arrive at the scene of the accident and carry out detection work in a timely manner, buying valuable time for emergency response.

[0036] 2) The present invention can synchronously acquire the terahertz spectrum and Raman spectrum of the sample to be tested in one acquisition process, which greatly improves the detection efficiency. In the data analysis stage, the data analysis module is used to synchronously process the original spectral signals of the two, generate a dual-spectrum fusion data packet containing a time-aligned spectrum graph, and use a pre-trained convolutional neural network to perform feature recognition and analysis, thereby achieving accurate detection of unknown chemicals. For samples that are applicable to both terahertz and Raman, the two test results can be compared and verified with each other, effectively reducing the risk of misjudgment and significantly improving the accuracy and reliability of the test results; for samples that are only applicable to one spectrum, the tedious process of using two instruments for separate detection is avoided, saving time and effort, further highlighting the huge application advantages of the present invention in the field of hazardous chemicals detection, and providing all-round technical support for rapid qualitative analysis of chemicals, accurate non-destructive non-contact detection, and micro-trace identification and detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the structure of the terahertz-Raman integrated detection system in the present invention.

[0038] In the figure: 1: sample to be tested, 2: integrated probe, 3: fusion mainframe, 4: Raman laser, 5: Raman filter, 6: Raman grating, 7: Raman detector, 8: data analysis module, 9: fusion power supply module, 10: terahertz detector, 11: terahertz femtosecond laser, 12: transmitting photoconductive antenna, 13: receiving photoconductive antenna, 14: off-axis parabolic mirror, 15: terahertz lens. DETAILED DESCRIPTION

[0039] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. Component models, material names, connection structures, circuit structures, control methods, algorithms, and other features not explicitly described in this technical solution are considered common technical features disclosed in the prior art.

[0040] Example 1

[0041] The terahertz-Raman integrated detection system in the embodiment includes a fusion mainframe 3 and an integrated probe 2 provided at the sample detection end of the fusion mainframe 3. Figure 1The fusion mainframe 3 is provided with a Raman spectroscopy module, a terahertz spectroscopy module, a terahertz optical module, and a data analysis module 8, which specifically include:

[0042] The Raman spectroscopy module includes a Raman laser 4, a Raman filter 5, a Raman grating 6, and a Raman detector 7 which are optically connected in sequence;

[0043] The terahertz spectroscopy module is optically connected in sequence to a terahertz femtosecond laser 11, a transmitting photoconductive antenna 12, a receiving photoconductive antenna 13, and a terahertz detector 10;

[0044] The terahertz optical module includes an off-axis parabolic mirror 14 and a terahertz lens 15 provided at the sample detection end, which are used to guide the terahertz wave to the sample to be measured 1 and receive the return signal;

[0045] The data analysis module 8 is in communication with the Raman detector 7 and the terahertz detector 10 respectively. The data analysis module 8 is used for synchronously processing the data transmitted by the Raman detector 7 and the terahertz detector 10 .

[0046] In a specific implementation, the terahertz optical module includes two off-axis parabolic mirrors 14 and two terahertz lenses 15;

[0047] An off-axis parabolic mirror 14 and a terahertz lens 15 are respectively matched on the side of the transmitting photoconductive antenna 12 and the side of the receiving photoconductive antenna 13 .

[0048] In a specific implementation, the optical path configuration of the terahertz optical module includes:

[0049] The off-axis parabolic mirror 14 reflects the terahertz wave generated by the terahertz femtosecond laser 11 and the transmitting photoconductive antenna 12, and then passes through the terahertz lens 15 and the integrated probe 2 to the sample 1 to be tested;

[0050] The terahertz wave reflected by the sample to be tested 1 passes through the integrated probe 2 , the terahertz lens 15 , the off-axis parabolic mirror 14 in sequence to reach the receiving photoconductive antenna 13 .

[0051] In specific implementation, the laser light generated by the Raman laser 4 passes through the integrated probe 2 to reach the sample 1 to be tested, and the Raman scattering signal reflected by the sample 1 to be tested passes through the integrated probe 2 and the Raman filter 5 to the Raman grating 6 .

[0052] In specific implementation, the data analysis module 8 synchronously receives and processes the original spectral signals transmitted by the Raman detector 7 and the terahertz detector 10 to generate a dual-spectrum fusion data packet, which contains a time-aligned Raman spectrum and a terahertz absorption spectrum.

[0053] Specifically, the Fusion Host Box 3, serving as the core of the entire detection system, integrates a Raman spectroscopy module, a terahertz spectroscopy module, a terahertz optical module, and a data analysis module 8. Its compact design makes the device easy to carry and use on-site. The touch screen integrated directly into the Fusion Host Box 3 for human-computer interaction further enhances its independence and ease of operation, eliminating the need for external computer connection and providing strong support for rapid on-site testing.

[0054] Specifically, the integrated probe 2, located within the fusion mainframe 3, is a key component for acquiring and transmitting spectral signals. It guides Raman laser light and terahertz waves toward the sample 1 under test, collects the Raman scattered signals and terahertz wave signals reflected by the sample, and transmits them back to the corresponding detectors within the fusion mainframe 3, ensuring efficient acquisition and transmission of spectral signals.

[0055] Specifically, the Raman laser 4 generates laser light of a specific wavelength as an excitation light source for Raman spectroscopy detection. The laser light generated by the Raman laser 4 is irradiated onto the sample 1 to be tested through the integrated probe 2, causing the sample to generate a Raman scattering signal.

[0056] Specifically, the Raman filter 5 performs filtering processing on the Raman scattering signal reflected from the sample, removes interference light such as Rayleigh scattered light, and retains pure Raman scattering signal.

[0057] Specifically, the Raman grating 6 spectrally disperses the Raman scattered signal after filtering, separating the different wavelengths of light according to a specific pattern to form a spectral image. The grating's dispersion effect decomposes the Raman scattered signal into spectral components of different wavelengths, enabling subsequent detailed analysis of the sample's molecular vibrational modes, enabling more accurate identification of the sample's molecular structure and chemical composition.

[0058] Specifically, Raman detector 7 receives the Raman spectrum signal after dispersion by Raman grating 6 and converts it into an electrical signal for subsequent processing and analysis by data analysis module 8. It accurately detects the intensity and distribution of the Raman spectrum signal and is a key component in converting optical signals into analyzable electrical signals.

[0059] Specifically, the terahertz femtosecond laser 11 generates femtosecond-level terahertz wave pulses, providing a high-intensity terahertz wave source for terahertz spectrum detection.

[0060] Specifically, the transmitting photoconductive antenna 12 emits terahertz waves in response to the femtosecond laser pulses generated by the terahertz femtosecond laser 11. It efficiently converts the laser energy into terahertz wave energy and transmits the generated terahertz waves, which are then guided to the sample 1 under test via the terahertz optical module, thereby irradiating the sample with terahertz waves.

[0061] Specifically, the receiving photoconductive antenna 13 captures the terahertz waves reflected by the sample 1 and converts them into electrical signals. It works in conjunction with the transmitting photoconductive antenna 12 to achieve the transmission and reception of terahertz waves, ensuring the effective acquisition of terahertz spectrum signals.

[0062] Specifically, the terahertz detector 10 receives the electrical signal from the receiving photoconductive antenna 13 , and processes and amplifies it so that the data analysis module 8 can further analyze it and perform high-sensitivity detection and processing on the weak electrical signal of the terahertz wave signal.

[0063] Specifically, there are two off-axis parabolic mirrors 14, one located on the transmitting photoconductive antenna 12 and the other on the receiving photoconductive antenna 13. Their primary function is to reflect and focus terahertz waves. On the transmitting side, they reflect and collimate the terahertz waves generated by the terahertz femtosecond laser 11, allowing them to efficiently enter the subsequent optical transmission path. On the receiving side, they collect and focus the terahertz waves reflected from the sample, improving the efficiency of the receiving photoconductive antenna 13.

[0064] Specifically, there are two terahertz lenses 15, which are used in conjunction with the off-axis parabolic mirror 14. In the transmitting optical path, the terahertz waves reflected by the off-axis parabolic mirror 14 are further focused, allowing them to accurately pass through the integrated probe 2 and be incident on the sample 1 to be measured. In the receiving optical path, the terahertz waves reflected from the sample are focused, increasing the received signal strength of the receiving photoconductive antenna 13, ensuring the energy concentration and efficient utilization of the terahertz waves during transmission, and guaranteeing the quality and stability of the terahertz spectral signal.

[0065] Specifically, the data analysis module 8 communicates with both the Raman detector 7 and the terahertz detector 10, serving as the brains of the entire detection system. Specifically, it utilizes an x86-based microprocessor coupled with a powerful GPU, RAM, and ROM. The module is responsible for synchronously receiving and processing the raw spectral signals transmitted from both detectors. Using data processing algorithms, it performs operations such as time-alignment and simultaneous feature extraction on the Raman and terahertz spectral data, generating a dual-spectrum fusion data packet containing the time-aligned Raman and terahertz absorption spectra. A pretrained convolutional neural network is then used to identify and analyze features within the dual-spectrum fusion data packet, ultimately outputting a combined analysis result. This enables comprehensive, accurate, and rapid detection and identification of unknown samples.

[0066] Example 2

[0067] This embodiment is a terahertz-Raman integrated detection method constructed using the above-mentioned detection system, comprising the following steps:

[0068] S1: The laser light generated by the Raman laser 4 irradiates the sample 1 to be measured through the integrated probe 2. The generated Raman scattering signal returns through the same probe 2 and enters the Raman detector 7 through the Raman filter 5 and the Raman grating 6 in sequence.

[0069] S2: The terahertz femtosecond laser 11 drives the transmitting photoconductive antenna 12 to generate a terahertz wave. After being reflected by the off-axis parabolic mirror 14 and focused by the terahertz lens 15, the terahertz wave is incident on the sample to be tested 1 through the integrated probe 2. The reflected terahertz wave passes through the terahertz lens 15 and the off-axis parabolic mirror 14, and is captured by the receiving photoconductive antenna 13 and transmitted to the terahertz detector 10.

[0070] During specific implementation, in S2 , the data analysis module 8 sends a synchronous trigger signal to the Raman detector 7 and the terahertz detector 10 simultaneously, so that both of them collect data with the same timestamp.

[0071] Specifically, the ultrashort laser pulses generated by the terahertz femtosecond laser 11 excite the photoconductive antenna 12. Based on the photoconductive effect, the antenna releases electrons in an ultrashort time, generating high-intensity terahertz waves. The generated terahertz waves are reflected by the off-axis parabolic mirror 14 for direction adjustment and collimation, and then focused by the terahertz lens 15, increasing the energy density of the terahertz waves and ensuring that they efficiently enter the integrated probe 2 and accurately enter the sample 1 to be measured. The terahertz waves reflected by the sample return through the same optical path, are again focused by the terahertz lens 15 and reflected by the off-axis parabolic mirror 14, and are guided to the receiving photoconductive antenna 13. The receiving photoconductive antenna 13 converts the captured terahertz wave signals into electrical signals and transmits them to the terahertz detector 10 for subsequent processing. Simultaneously, the data analysis module 8 sends a synchronization trigger signal to the Raman detector 7 and the terahertz detector 10, so that both collect data at the same timestamp, achieving precise synchronization of the two spectral data.

[0072] S3: The data analysis module 8 synchronously receives the spectral data from the Raman detector 7 and the terahertz detector 10 and performs time axis alignment.

[0073] In specific implementation, in S3, time axis alignment includes timing compensation operations, specifically including:

[0074] Based on the transmission time difference of the terahertz wave in the optical path, the Raman spectrum data is time-shifted and corrected to align the starting time points of the two spectra.

[0075] Specifically, due to the time difference between the generation and transmission of Raman spectral signals and terahertz spectral signals in their respective optical paths, the time axes of the two are not completely aligned. In specific implementation, in order to achieve accurate synchronization of the two spectral data, the time axis alignment operation becomes a key link. Among them, the timing compensation operation is the core, which is mainly based on the time difference of the transmission of the terahertz wave in the optical path to perform time shift correction on the Raman spectral data. Through the precise measurement and calculation of the transmission time difference, the time starting point of the Raman spectral data is adjusted accordingly, so that the starting time points of the two spectra are aligned, and the precise synchronization of the time axis is achieved, which provides a reliable data basis for the subsequent joint analysis and fusion processing of the two spectral data, and ensures the accuracy and reliability of the analysis results.

[0076] S4: The data analysis module 8 performs synchronous feature extraction on the Raman spectrum data and the terahertz absorption spectrum data, constructs a time-correlated dual-spectrum fusion data packet, identifies the features in the dual-spectrum fusion data packet based on the pre-trained convolutional neural network, and outputs the joint analysis results.

[0077] In specific implementation, in S4, the synchronous feature extraction includes:

[0078] Baseline correction and peak position extraction are performed on Raman spectral data, while dielectric constant inversion and characteristic peak fitting are performed on terahertz absorption spectral data to generate a dual-spectrum fusion data package containing molecular bond vibration frequencies and lattice vibration modes.

[0079] Specifically, for Raman spectral data, baseline correction is performed to remove background interference and stabilize the spectral baseline. A peak extraction algorithm is then used to precisely locate the Raman characteristic peaks, the wavelengths of which are related to the specific vibrational modes of the sample molecules. For terahertz absorption spectral data, based on the sample's absorption characteristics for terahertz waves, dielectric constant inversion is used to calculate the sample's dielectric constant, thereby obtaining information such as the polarization and conductivity within the sample. At the same time, characteristic peak fitting is used to determine the position and intensity of the terahertz absorption peak, which are closely related to the sample's lattice vibrational modes. After completing the synchronous feature extraction, the features extracted from the two spectral data sets are fused to construct a time-correlated dual-spectrum fusion data package containing information on molecular bond vibration frequencies and lattice vibrational modes.

[0080] Finally, the data packet is input into a pre-trained convolutional neural network, which uses its pattern recognition and feature learning capabilities to analyze and classify the fused features, identify the specific spectral characteristics of the sample, and output the joint analysis results based on the two spectral technologies, thereby achieving accurate detection and identification of unknown samples.

[0081] The pre-trained convolutional neural network in the present invention is obtained by training a large number of Raman spectra and terahertz spectra data of known chemicals. First, a large number of representative chemical samples are collected, and their corresponding Raman spectra and terahertz spectra data are obtained to construct a training data set. Then, these spectral data are preprocessed, including normalization, feature extraction and other operations to meet the input requirements of the convolutional neural network. Next, a convolutional neural network model is constructed, which usually includes multiple convolutional layers, pooling layers and fully connected layers. The convolutional layer is used to extract local features in the spectral data, the pooling layer is used to reduce the data dimension and reduce the amount of calculation, and the fully connected layer is used to comprehensively analyze and classify the extracted features. During the training process, the training data set is input into the network, and the loss function (such as cross entropy loss) between the network output and the true label is calculated by forward propagation. Then, the network weights are updated using the backpropagation algorithm and optimization algorithm (such as gradient descent method), and the network parameters are continuously adjusted to minimize the loss function. After multiple iterative training, when the accuracy of the network on the validation set reaches a high level and tends to be stable, the training process ends and a pre-trained convolutional neural network model is obtained. The model can effectively identify features in dual-spectrum fusion data packets and achieve accurate classification and identification of unknown chemicals.

[0082] The above description of the embodiments is intended to facilitate understanding and use of the invention by those skilled in the art. It will be apparent that those skilled in the art can readily make various modifications to these embodiments and apply the general principles described herein to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above-described embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention, without departing from the scope of the present invention, should be within the scope of protection of the present invention.

Claims

1. A terahertz-Raman integrated detection system, characterized in that: It comprises a fusion mainframe box (3) and an integrated probe (2) provided at a sample detection end on the fusion mainframe box (3); The fusion mainframe box (3) is provided with: Raman spectroscopy module: comprising a Raman laser (4), a Raman filter (5), a Raman grating (6), and a Raman detector (7) which are optically connected in sequence; Terahertz spectroscopy module: a terahertz femtosecond laser (11), a transmitting photoconductive antenna (12), a receiving photoconductive antenna (13), and a terahertz detector (10) connected in sequence by optical paths; A terahertz optical module comprises an off-axis parabolic mirror (14) and a terahertz lens (15) provided at a sample detection end, and is used to guide the terahertz wave to the sample to be measured (1) and receive a return signal; A data analysis module (8) is respectively connected to the Raman detector (7) and the terahertz detector (10) for communication, and the data analysis module (8) is used for synchronously processing data transmitted by the Raman detector (7) and the terahertz detector (10).

2. The terahertz-Raman integrated detection system according to claim 1, characterized in that: The terahertz optical module includes two off-axis parabolic mirrors (14) and two terahertz lenses (15); An off-axis parabolic mirror (14) and a terahertz lens (15) are respectively matched on one side of the transmitting photoconductive antenna (12) and one side of the receiving photoconductive antenna (13).

3. The terahertz-Raman integrated detection system according to claim 2, characterized in that: The optical path configuration of the terahertz optical module includes: The off-axis parabolic mirror (14) reflects the terahertz waves generated by the terahertz femtosecond laser (11) and the transmitting photoconductive antenna (12), and then passes through the terahertz lens (15) and the integrated probe (2) to the sample to be tested (1); The terahertz wave reflected by the sample to be tested (1) passes through the integrated probe (2), the terahertz lens (15), the off-axis parabolic mirror (14) in sequence to the receiving photoconductive antenna (13).

4. The terahertz-Raman integrated detection system according to claim 1, characterized in that: The laser light generated by the Raman laser (4) passes through the integrated probe (2) to reach the sample to be measured (1), and the Raman scattering signal reflected by the sample to be measured (1) passes through the integrated probe (2) and the Raman filter (5) to the Raman grating (6).

5. The terahertz-Raman integrated detection system according to claim 1, characterized in that: The data analysis module (8) synchronously receives and processes the original spectrum signals transmitted by the Raman detector (7) and the terahertz detector (10), and generates a dual-spectrum fusion data packet, wherein the dual-spectrum fusion data packet contains a time-aligned Raman spectrum and a terahertz absorption spectrum.

6. A terahertz-Raman integrated detection method using the detection system according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1: The laser light generated by the Raman laser (4) is irradiated on the sample to be measured (1) through the integrated probe (2). The generated Raman scattering signal is returned through the same probe (2) and then passes through the Raman filter (5) and the Raman grating (6) to enter the Raman detector (7); S2: The terahertz femtosecond laser (11) drives the transmitting photoconductive antenna (12) to generate terahertz waves, which are reflected by the off-axis parabolic mirror (14) and focused by the terahertz lens (15). The terahertz waves are incident on the sample to be tested (1) through the integrated probe (2). The reflected terahertz waves pass through the terahertz lens (15) and the off-axis parabolic mirror (14), and are captured by the receiving photoconductive antenna (13) and transmitted to the terahertz detector (10). S3: The data analysis module (8) synchronously receives the spectral data of the Raman detector (7) and the terahertz detector (10), and performs time axis alignment; S4: The data analysis module (8) performs synchronous feature extraction on the Raman spectrum data and the terahertz absorption spectrum data, constructs a time-correlated dual-spectrum fusion data packet, identifies the features in the dual-spectrum fusion data packet based on the pre-trained convolutional neural network, and outputs the joint analysis results.

7. The terahertz-Raman integrated detection method according to claim 6, characterized in that: In S2, the data analysis module (8) simultaneously sends a synchronous trigger signal to the Raman detector (7) and the terahertz detector (10), so that both of them collect data with the same time stamp.

8. The terahertz-Raman integrated detection method according to claim 6, characterized in that: In S3, timeline alignment includes timing compensation operations, specifically: Based on the transmission time difference of the terahertz wave in the optical path, the Raman spectrum data is time-shifted and corrected to align the starting time points of the two spectra.

9. The terahertz-Raman integrated detection method according to claim 6, characterized in that: In S4, the synchronous feature extraction includes: Baseline correction and peak position extraction are performed on Raman spectral data, while dielectric constant inversion and characteristic peak fitting are performed on terahertz absorption spectral data to generate a dual-spectrum fusion data package containing molecular bond vibration frequencies and lattice vibration modes.

10. The terahertz-Raman integrated detection method according to claim 6, characterized in that: In S4, the specific process of constructing the time-correlated dual-spectrum fusion data packet includes: The Raman spectrum and the terahertz absorption spectrum at the same timestamp are superimposed into a three-dimensional spectrum, where the X-axis is the wave number, the Y-axis is the time, and the Z-axis is the signal intensity.

Images