Spectroscopy-based composition detection system and method

Abstract

The present application relates to the technical field of spectral analysis, and provides a kind of component detection system and method based on spectrum, and component detection system based on spectrum, comprising: three optical path differential probe, spectrometer and control circuit board, three optical path differential probe, for the initial light signal of the component object to be measured is carried out interference elimination processing, obtains target light signal, and target light signal is sent to spectrometer;Spectrometer is used for obtaining the spectral data of the component object to be measured based on target light signal, and spectral data is sent to control circuit board;Control circuit board is used for determining the component information of the component object to be measured based on spectral data and target component detection model.The present application eliminates strong light interference, improves the accuracy of light signal, without destroying sample when detecting component, shortens the detection duration, improves the stability, accuracy and efficiency of outdoor component detection.

Description

Technical Field

[0001] This invention relates to the field of spectral analysis technology, and in particular to a spectral-based component detection system and method. Background Technology

[0002] Traditional component detection mainly relies on wet chemical analysis methods (such as digestion titration for total nitrogen and colorimetric methods for sugar content) and chromatographic analysis methods (such as high performance liquid chromatography for nicotine content). In practical applications, these methods have efficiency bottlenecks and adaptability limitations: the detection process requires destroying the sample and takes up to several hours, which cannot meet the needs of rapid decision-making at the acquisition site; although conventional spectroscopic equipment improves speed, it is limited by strong light interference (day and night data drift) and temperature and humidity sensitivity (fogging of optical components and baseline distortion), which reduces the stability and accuracy of outdoor component detection. Summary of the Invention

[0003] This invention provides a spectral-based component detection system and method to address the shortcomings of existing technologies, such as limitations caused by strong light interference (day-night data drift) and temperature and humidity sensitivity (fogging of optical components, baseline distortion), which reduce the stability and accuracy of outdoor component detection. The system utilizes a three-path differential probe to eliminate strong light interference from the target component, improving the accuracy of the optical signal. Furthermore, it employs a target component detection model to analyze and detect the spectral data of the target component acquired by the spectrometer, obtaining component information without damaging the sample, thus shortening the detection time and improving the stability, accuracy, and efficiency of outdoor component detection.

[0004] This invention provides a spectrum-based component detection system, comprising: a three-path differential probe, a spectrometer, and a control circuit board, wherein: The three-path differential probe is used to perform interference cancellation processing on the initial optical signal of the component to be measured, obtain the target optical signal, and send the target optical signal to the spectrometer. The spectrometer is used to obtain the spectral data of the component to be measured based on the target light signal, and to send the spectral data to the control circuit board; The control circuit board is used to determine the component information of the target component object based on the spectral data and the target component detection model. The target component detection model is used to detect the spectral data to obtain the component information of the target component object based on the dual-path heterogeneous feature extraction network, the adaptive feature fusion module and the enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local features and global features of the spectral data based on the features of the spectral data, and fuse the local features and global features based on the fusion strategy.

[0005] According to the present invention, a spectral-based component detection system further includes a halogen light source. The three-path differential probe includes a first optical fiber and an optical fiber group. The axis of the first optical fiber and the axis of each optical fiber in the optical fiber group point to the same position on the component to be measured. The optical paths on the first optical fiber and the optical paths of each optical fiber in the optical fiber group intersect at the acquisition plane. The first optical fiber is connected to the halogen light source, and the optical fiber group is connected to the spectrometer. Wherein: The first optical fiber is used to transmit the light provided by the halogen light source to the sample to obtain the initial optical signal; The optical fiber array is used to perform interference cancellation processing on the initial optical signal to obtain the target optical signal, and then send the target optical signal to the spectrometer.

[0006] According to the present invention, a spectral-based component detection system includes an optical fiber assembly comprising a first sub-optical fiber and a second sub-optical fiber, a spectrometer comprising a visible band spectrometer and a near-infrared spectrometer, and spectral data comprising visible band spectral data and near-infrared spectral data. The first sub-optical fiber is connected to the visible band spectrometer, and the second sub-optical fiber is connected to the near-infrared spectrometer. Specifically: the visible band spectrometer is used to acquire the visible band spectral data based on the target light signal transmitted through the first sub-optical fiber and to send the visible band spectral data to the control circuit board; the near-infrared spectrometer is used to acquire the near-infrared spectral data based on the target light signal transmitted through the second sub-optical fiber and to send the near-infrared spectral data to the control circuit board.

[0007] According to the present invention, a spectral-based component detection system further includes a cloud platform, which is used to incrementally train an initial component detection model based on the spectral data to obtain a target component detection model, and then send the target component detection model to the control circuit board.

[0008] According to the present invention, a spectrum-based component detection system includes an operating device, which is equipped with a cloud management system and is used to display the operating interface of the cloud management system.

[0009] According to the present invention, a spectral-based component detection system is provided, wherein the control circuit board includes a preprocessing unit and a detection unit, wherein: the preprocessing unit is used to preprocess the spectral data to obtain target spectral data; and the detection unit is used to input the target spectral data into the target component detection model to obtain the component information of the component to be tested.

[0010] According to the present invention, a spectral-based component detection system is provided, wherein the target component detection model is used to screen the target spectral data to obtain the key wavelengths in the target spectral data, and to detect the key wavelengths to obtain the component information of the component to be tested.

[0011] According to the present invention, a spectral-based component detection system further includes an embedded calibration plate lens cover, which is connected to the three-path differential probe and is used to calibrate the spectral-based component detection system.

[0012] According to the present invention, a spectral-based component detection system further includes a display screen for displaying component information of the component to be detected.

[0013] This invention also provides a spectral-based component detection method, applied to a spectral-based component detection system. The spectral-based component detection system includes a three-path differential probe, a spectrometer, and a control circuit board. The target optical signal is obtained by performing interference cancellation processing on the initial optical signal of the component to be measured through the three-optical-path differential probe. The spectrometer obtains the spectral data of the component to be measured based on the target light signal; The control circuit board determines the component information of the target component object based on the spectral data and the target component detection model. The target component detection model is used to detect the spectral data to obtain the component information of the target component object based on the dual-path heterogeneous feature extraction network, adaptive feature fusion module and enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local features and global features of the spectral data based on the features of the spectral data, and fuse the local features and global features based on the fusion strategy.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the spectral-based component detection method as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the spectral-based component detection method as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the spectral-based component detection method as described above.

[0017] This invention provides a spectral-based component detection system and method. The system comprises a three-path differential probe, a spectrometer, and a control circuit board. The three-path differential probe performs interference cancellation processing on the initial optical signal of the component to be detected, obtaining a target optical signal, which is then sent to the spectrometer. The spectrometer obtains the spectral data of the component based on the target optical signal and sends this data to the control circuit board. The control circuit board determines the component information of the component based on the spectral data and a target component detection model. Thus, by using the three-path differential probe to eliminate strong light interference from the component to be detected, the accuracy of the optical signal is improved. Furthermore, by using the target component detection model to analyze and detect the spectral data of the component acquired by the spectrometer, component information is obtained without damaging the sample, shortening the detection time and improving the stability, accuracy, and efficiency of outdoor component detection. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is one of the structural schematic diagrams of the spectrum-based component detection system provided by the present invention.

[0020] Figure 2 This is the second schematic diagram of the structure of the spectrum-based component detection system provided by the present invention.

[0021] Figure 3 This is a schematic diagram of the structure of the three-path differential probe provided by the present invention.

[0022] Figure 4 This is a schematic diagram of the optical path of the three-path differential probe provided by the present invention.

[0023] Figure 5 This is the third schematic diagram of the structure of the spectrum-based component detection system provided by the present invention.

[0024] Figure 6 This is a schematic diagram of the user interface of the cloud management system provided by the present invention.

[0025] Figure 7 This is a schematic diagram illustrating the display screen provided by the present invention.

[0026] Figure 8 This is a schematic diagram of the structure of the handheld hardware device provided by the present invention.

[0027] Figure 9 This is a schematic diagram of the main body of the device provided by the present invention.

[0028] Figure 10 This is a schematic flowchart of the spectrum-based component detection method provided by the present invention.

[0029] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0030] Figure label: 100: Spectroscopy-based component detection system; 110: Three-path differential probe; 120: Spectrometer; 130: Control circuit board; 140: Halogen light source; 111: First optical fiber; 112: Optical fiber group; 1121: First sub-optical fiber; 1122: Second sub-optical fiber; 150: Cloud; 701: Bluetooth symbol; 702: Setting symbol; 703: Remaining battery symbol; 800: Handheld hardware device; 801: Device body; 802: Display screen; 803: Built-in battery handle; 804: Embedded calibration plate lens cover; 805: Acquisition button; 121: Visible band spectrometer; 122: Near-infrared spectrometer; 1110: Processor; 1120: Communication interface; 1130: Memory; 1140: Communication bus. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0032] Currently, although conventional spectroscopic equipment has improved speed, it is limited by strong light interference (day and night data drift) and temperature and humidity sensitivity (fogging of optical components and baseline distortion), which reduces the stability and accuracy of outdoor component detection.

[0033] Meanwhile, fixed calibration models are difficult to adapt to the differences in tobacco leaf characteristics in different production areas (such as moisture fluctuations or variety variations). After the model fails, it needs to be returned to the factory for calibration, which increases the operation and maintenance costs. Furthermore, since traditional point spectroscopy methods generally only support the detection of 1-2 components at a time, the evaluation of multiple indicators requires repeated operations, which further exacerbates the complexity of the application.

[0034] To address the aforementioned problems, this invention provides a spectral-based component detection system and method. It utilizes a three-path differential probe to eliminate strong light interference from the target component, improving the accuracy of the optical signal. Furthermore, it employs a target component detection model based on a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module to analyze and detect the spectral data of the target component acquired by the spectrometer, obtaining component information. This enhances the model's generalization ability and detection accuracy, eliminates the need to damage the sample, shortens the detection time, and improves the stability, accuracy, and efficiency of outdoor component detection.

[0035] The following is combined Figures 1 to 9 The present invention describes a spectral-based component detection system.

[0036] Figure 1 This is one of the structural schematic diagrams of the spectrum-based component detection system provided by the present invention, such as... Figure 1 As shown, the spectral-based component detection system 100 includes: a three-path differential probe 110, a spectrometer 120, and a control circuit board 130, wherein: The three-path differential probe 110 is used to perform interference cancellation processing on the initial optical signal of the component to be measured, obtain the target optical signal, and send the target optical signal to the spectrometer 120; The spectrometer 120 is used to obtain the spectral data of the component to be measured based on the target light signal, and send the spectral data to the control circuit board 130; The control circuit board 130 is used to determine the component information of the target component object based on the spectral data and the target component detection model. The target component detection model is used to detect the spectral data to obtain the component information of the target component object based on the dual-path heterogeneous feature extraction network, the adaptive feature fusion module and the enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local features and global features of the spectral data based on the features of the spectral data, and fuse the local features and global features based on the fusion strategy.

[0037] Here, the analyte can be any suitable object, such as tobacco, minerals, or plants.

[0038] Here, the component information of the test object refers to its constituent components. For example, the constituent components of tobacco include nicotine, total sugar, total alkaloids, and protein.

[0039] Here, the initial light signal of the component to be measured can be obtained using a halogen light source or a solar light source.

[0040] Here, interference cancellation processing is used to eliminate strong light in the initial optical signal, such as sunlight interference.

[0041] Here, the spectrometer may include a single spectrometer or multiple spectrometers.

[0042] Here, the method of sending the target optical signal can be optical fiber or wireless optical communication.

[0043] Here, the control circuit board can be understood as an embedded processor.

[0044] Here, the target component detection model can be an Artificial Intelligence (AI) spectral big data model. This AI spectral big data model can be obtained by training an initial spectral big data model on the collected spectral data using an AI spectral big data platform in the cloud and then integrating it into the control circuit board; alternatively, an existing spectral big data model can be directly integrated into the control circuit board.

[0045] Here, the spectrum-based component detection system can be understood as a handheld hardware device, and the three-path differential probe, spectrometer, and control circuit board are the components in the handheld hardware device.

[0046] Here, the spectral data can be obtained by scanning the spectra of five representative positions of a single tobacco leaf sample in the spectral range of 400-1700 nm to detect the physical inhomogeneity of the component to be tested (such as tobacco leaves), with each position being measured N times.

[0047] It should be noted that, in addition to the dual-path heterogeneous feature extraction network, adaptive feature fusion module, and enhancement module, the target component detection model may also include a multi-source data preprocessing layer and a sparse activation regression prediction module. The detection process of the target component detection model includes the following: (1) The multi-source data preprocessing layer performs the processing flow. The processing flow averages the spectral data to suppress random noise. Then, multiplicative scattering correction (MSC) is applied to eliminate the influence of physical scattering, and targeted filtering is added. Finally, the spectral data at 5 locations are constructed into a 5×wavelength two-dimensional tensor as the model input, ensuring the richness of spatial dimension information of the input data.

[0048] (2) The dual-path heterogeneous feature extraction network captures local details and global long-range dependencies of spectral data respectively. The first path of the dual-path heterogeneous feature extraction network is for local feature extraction: it uses a two-dimensional convolutional neural network (2D CNN) to efficiently extract local texture and subtle component features in spectral data by utilizing the local receptive field characteristics of the convolution kernel; the second path is for global dependency modeling: it uses a visual state space model (VMamba) to capture global spectral features and long-range dependencies across different bands by utilizing its advantages in long sequence modeling.

[0049] (3) The data-driven adaptive fusion module fuses local and global features to obtain fused features: Through a small meta-network, the model dynamically learns and generates the optimal fusion strategy for local and global features of the spectral data based on the features of the current input data, i.e., the spectral data. During training, the model automatically learns the fusion strategy of "when, where, and how" to fuse the two features, and fuses global and local features according to the fusion strategy, thereby achieving optimal adaptation to different tobacco leaf samples.

[0050] (4) Based on the enhancement module, feature enhancement is performed on the fused features, and a multi-scale attention mechanism is introduced. This mechanism can adaptively focus on the spectral region with the most discriminative power for component detection, while suppressing irrelevant background noise. More importantly, through the "focusing" effect of the attention mechanism, the model can ignore a large amount of redundant computation, thereby significantly reducing the computational load of subsequent computation modules without sacrificing accuracy.

[0051] (5) Component information detection using a sparse activation regression prediction module: A Mixture of Experts (MoE) architecture is introduced to further optimize computational efficiency. Sparse activation mechanism: The MoE architecture consists of multiple "expert" networks. During inference, a sparse activation strategy is adopted, that is, for each input sample, only a small portion (e.g., 1-2) of the most relevant "expert" networks are activated and run to process the data, while the remaining "experts" remain dormant. This sparse activation mechanism allows the model to have a huge total number of parameters (to ensure powerful expressive power), but the computational cost during actual inference is very small, achieving a high efficiency ratio of "large model capability, small model overhead", which is very suitable for deployment on edge devices or in scenarios with real-time requirements.

[0052] (6) Output layer: The features processed by sparse activation are fed into a multilayer perceptron (MLP) regression head to output the predicted content of each target component in the tobacco leaf.

[0053] In this embodiment of the invention, the adaptive feature fusion module is data-driven, replacing the traditional fixed fusion strategy. This allows the model to dynamically adjust the fusion method of local and global features according to the characteristics of the input samples, significantly improving the model's generalization ability and detection accuracy. High-efficiency computational architecture: Focused computation is achieved through a multi-scale attention mechanism, reducing redundant operations; combined with the sparse activation mechanism of MoE, the computational load and energy consumption in the inference stage are significantly reduced while maintaining the model's powerful expressive power, achieving a balance between high accuracy and high efficiency. Multi-path heterogeneous design: Combining a dual-path structure of 2D CNN and VMamba, it can simultaneously capture local details and global dependencies of spectral data, providing a solid data foundation for high-precision detection.

[0054] In this embodiment of the invention, a spectral-based component detection system is constructed using a three-path differential probe, a spectrometer, and a control circuit board. The three-path differential probe is used to eliminate interference in the initial optical signal of the component to be tested, obtaining a target optical signal, which is then sent to the spectrometer. The spectrometer obtains the spectral data of the component to be tested based on the target optical signal and sends the spectral data to the control circuit board. The control circuit board determines the component information of the component to be tested based on the spectral data and a target component detection model. The target component detection model uses a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module to detect the spectral data and obtain the component information of the component to be tested. Thus, the three-path differential probe eliminates strong light interference from the component to be tested, improving the accuracy of the optical signal. Furthermore, the target component detection model analyzes and detects the spectral data of the component to be tested collected by the spectrometer using a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module to obtain component information, improving the model's generalization ability and detection accuracy. This eliminates the need to destroy samples, shortens the detection time, and improves the stability, accuracy, and efficiency of outdoor component detection.

[0055] Figure 2 This is a second schematic diagram of the structure of the spectrum-based component detection system provided by the present invention, as shown below. Figure 2As shown, the spectral-based component detection system 100 further includes a halogen light source 140. The three-path differential probe 110 includes a first optical fiber 111 and an optical fiber group 112. The axis of the first optical fiber 111 and the axis of each optical fiber in the optical fiber group 112 point to the same position on the component to be measured. The optical paths on the first optical fiber 111 and the optical paths of each optical fiber in the optical fiber group 112 intersect at the acquisition plane. The first optical fiber 111 is connected to the halogen light source 140, and the optical fiber group is connected to the spectrometer 120. The first optical fiber 111 is used to transmit the light provided by the halogen light source 140 to the sample to obtain the initial optical signal; The fiber optic group 112 is used to perform interference cancellation processing on the initial optical signal to obtain the target optical signal, and to send the target optical signal to the spectrometer 120.

[0056] Here, the wavelength range of the halogen light source can be 400-1700 nanometers (nm), and the halogen light source can produce a continuous and stable spectrum.

[0057] Here, the types of optical fibers include, but are not limited to, single-mode optical fibers, multimode optical fibers, and quartz optical fibers.

[0058] It should be noted that the first optical fiber is connected to the halogen light source, and the halogen light source transmits the generated optical signal to the component to be measured through the first optical fiber to obtain the initial optical signal corresponding to the component to be measured.

[0059] Here, the initial optical signal can be the light reflected by the component to be measured. The initial optical signal is processed by the optical fiber group to eliminate interference, and the target optical signal is obtained. Then, it is transmitted to the spectrometer through the optical fiber group.

[0060] It should be noted that an optical fiber group may include multiple sub-optical fibers, each of which is connected to a spectrometer, and each spectrometer may be of a different type.

[0061] It should be noted that the optical axes of the first optical fiber and the optical fiber group point to the same position on the object to be measured, and the optical paths on the optical fibers intersect on the acquisition plane. This ensures that the same area is illuminated and received, improves the spatial registration accuracy of multi-band spectral data, and eliminates measurement errors caused by spot offset. This ensures the consistency, accuracy and repeatability of multi-band spectra from the root.

[0062] It should be noted that the divergence angle of the probe on each optical fiber at the point of intersection can be calculated using software simulation.

[0063] In this embodiment of the invention, by setting the axis of the first optical fiber and the axis of each optical fiber in the optical fiber group to point to the same position on the object to be measured, the optical paths on the first optical fiber and the optical paths of each optical fiber in the optical fiber group are converged on the acquisition plane to acquire spectral data, ensuring that the same area is illuminated and received, improving the spatial registration accuracy of multi-band spectral data, and eliminating measurement errors caused by spot offset, thus ensuring the consistency, accuracy and repeatability of multi-band spectra from the root.

[0064] Furthermore, the optical fiber assembly includes a first sub-optical fiber and a second sub-optical fiber, the spectrometer includes a visible band spectrometer and a near-infrared spectrometer, and the spectral data includes visible band spectral data and near-infrared spectral data. The first sub-optical fiber is connected to the visible band spectrometer, and the second sub-optical fiber is connected to the near-infrared spectrometer. Specifically: the visible band spectrometer is used to collect the visible band spectral data based on the target light signal transmitted through the first sub-optical fiber and send the visible band spectral data to the control circuit board; the near-infrared spectrometer is used to collect the near-infrared spectral data based on the target light signal transmitted through the second sub-optical fiber and send the near-infrared spectral data to the control circuit board.

[0065] Here, the visible band spectral data is in the range of 400-800 nm, and the near-infrared band spectral data is in the range of 800-1700 nm.

[0066] Specifically, the first sub-fiber is used to transmit the target optical signal to the visible band spectrometer, and the second sub-fiber is used to transmit the target optical signal to the infrared spectrometer.

[0067] In this embodiment of the invention, by setting up a visible band spectrometer and a near-infrared spectrometer, the range of spectral data is expanded, the diversity of spectral data acquisition is improved, and the accuracy of component detection is enhanced.

[0068] Figure 3 This is a schematic diagram of the structure of the three-path differential probe provided by the present invention, as shown below. Figure 3 As shown, the three-path differential probe 110 includes a first optical fiber 111, a first sub-optical fiber 1121, and a second sub-optical fiber 1122. The first optical fiber 111 transmits the light source of the halogen light source to the object to be measured. The first sub-optical fiber 1121 and the second sub-optical fiber 1122 transmit the target light signal of the object to a visible spectrum analyzer and an infrared spectrum analyzer, respectively. The spatial angle between the first optical fiber 111 and the first sub-optical fiber 1121 is 120 degrees, the spatial angle between the first sub-optical fiber 1121 and the second sub-optical fiber 1122 is 120 degrees, and the spatial intersection distance between the first optical fiber 111, the first sub-optical fiber 1121, and the second sub-optical fiber 1122 is 10 millimeters (mm).

[0069] Figure 4 This is a schematic diagram of the optical path of the three-path differential probe provided by the present invention, as shown below. Figure 4 As shown, green represents the optical path of the first fiber, red represents the optical path of the first sub-fiber, and purple represents the optical path of the second sub-fiber. The fibers are arranged in space at a specific angle so that the axes of the three fibers point to the same point, and the divergence angle of the light emitted from the fibers is calculated. Finally, the divergence angles of the three fibers converge on the acquisition plane.

[0070] Figure 5 This is the third schematic diagram of the structure of the spectrum-based component detection system provided by the present invention. The spectrum-based component detection system 100 also includes a cloud 150. The cloud 150 is used to incrementally train the initial component detection model based on the spectral data to obtain the target component detection model, and send the target component detection model to the control circuit board 130.

[0071] In this context, the spectral-based component detection system can include not only handheld hardware devices but also cloud-based systems.

[0072] It should be noted that after the spectrometer collects spectral data, in addition to sending it to the control circuit board, it can also transmit the spectral data to the cloud, forming a collaborative mechanism with the cloud to update the target component detection model mounted on the control circuit board. The spectral data can be uploaded by the user via cloud operation, or it can be transmitted wirelessly to the cloud by the control circuit.

[0073] It should be noted that incremental training refers to retraining the current component detection model, i.e., the initial component detection model, to obtain the target component detection model.

[0074] Here, the cloud can send the target component detection model or its upgrade package to the control circuit board via wireless or over-the-air update (OTA). After receiving the target component detection model or its upgrade package, the control circuit board updates the initial component detection model.

[0075] In this embodiment of the invention, the handheld hardware device and the cloud platform work together to build a dynamic optimization mechanism. After the spectral data is uploaded, incremental training of the model is triggered and the device-side algorithm is updated to continuously improve the detection accuracy. At the same time, the index parameters can be expanded through the continuous improvement of the AI ​​spectral big data model.

[0076] Furthermore, the cloud includes an operating device, which is equipped with a cloud management system and is used to display the operating interface of the cloud management system.

[0077] Here, users can update the component detection model by clicking the buttons on the operating interface of the device.

[0078] Figure 6 This is a schematic diagram of the user interface of the cloud management system provided by the present invention, as shown below. Figure 6 As shown, the user interface features a sidebar navigation bar on the left, including user management, device management, data management, and an algorithm center. The main content includes buttons for uploading spectral data, uploading verification data, and uploading to the spectral database. The interface also includes buttons for spectral data analysis and model algorithm upgrades, enabling device and data management, analysis, and algorithm upgrades. Furthermore, the interface displays a visual graph of the spectral data, and a list below shows device number, user information, and other details.

[0079] For example, after the spectral data is uploaded to the cloud, it triggers the model algorithm upgrade to perform incremental model updates, and updates the device-side algorithm through OTA technology to continuously improve detection accuracy and expand detection indicators.

[0080] In this embodiment of the invention, device management, data analysis, and dynamic algorithm optimization are achieved by displaying the operation interface of the cloud management system on the cloud-based operating device.

[0081] Furthermore, the control circuit board includes a preprocessing unit and a detection unit, wherein: the preprocessing unit is used to preprocess the spectral data to obtain target spectral data; the detection unit is used to input the target spectral data into the target component detection model to obtain the component information of the component to be tested.

[0082] Optionally, the preprocessing unit can perform preprocessing on the spectral data, such as filtering, interpolation, and reflectance calculation, to improve the reliability and quality of the spectral data, provide high-quality data for subsequent component detection, and further improve the accuracy of component analysis.

[0083] After preprocessing the spectral data, the target spectral data is obtained. The target spectral data is then input into the target component detection model to obtain the component information of the target component.

[0084] In this embodiment of the invention, the spectral data is preprocessed by a preprocessing unit to improve the reliability and quality of the spectral data, providing high-quality data for subsequent component detection, further improving the accuracy of component analysis, and the detection unit is used to complete non-destructive testing and quantitative analysis of the components of the object to be tested.

[0085] Furthermore, the target component detection model is used to filter the target spectral data to obtain key wavelengths in the target spectral data, and to detect the key wavelengths to obtain the component information of the component to be tested.

[0086] Optionally, the screening methods include, but are not limited to, hyperspectral continuous projection algorithms, non-information variable elimination methods, and competitive adaptive reweighted sampling. For example, key wavelengths may include characteristic peaks of nicotine (1410 / 1680 nm), total nitrogen (1510 nm), starch (1450 nm), and reducing sugars.

[0087] Optionally, algorithms for key wavelength detection can employ a fusion of partial least squares regression and random forest algorithms, or neural network classification algorithms. Partial least squares regression is used for feature extraction, while random forest algorithms are used for classification.

[0088] In another embodiment of the present invention, the component information output by the target component detection model may include the component name and component content, i.e., confidence level.

[0089] In this embodiment of the invention, the target component detection model enables rapid quantitative analysis of core crop components (nicotine, total sugar, etc.) outdoors.

[0090] Furthermore, the spectral-based component detection system also includes an embedded calibration plate lens cover, which is connected to the three-optical-path differential probe and is used to calibrate the spectral-based component detection system.

[0091] It should be noted that the initial light signal of the component to be measured can only enter the three-path differential probe for spectral data acquisition when the lens cover of the embedded calibration plate is open. When the lens cover of the embedded calibration plate is closed, the built-in calibration plate can be used to calibrate the spectral component detection system.

[0092] In this embodiment of the invention, calibration is achieved by embedding a calibration plate inside the lens cover, which can adapt to the differences in tobacco leaf characteristics (such as moisture fluctuations or variety variations) in different production areas. After the model fails, there is no need to return it to the factory for calibration, thus reducing operation and maintenance costs.

[0093] Furthermore, the spectral-based component detection system also includes a display screen for displaying component information of the component to be detected.

[0094] Here, the display screen is located behind the handheld hardware device in the spectral-based component detection system. The display screen can also display device information (such as device name, serial number, Bluetooth name, etc.) and algorithm information (algorithm package name, algorithm version, etc.). When the product algorithm is upgraded or new acquisition indicators are added in the future, the screen can adapt to the display of new indicators without the need for secondary development.

[0095] Here, the display screen can display information through a driver.

[0096] Figure 7 This is a schematic diagram illustrating the display screen provided by the present invention, such as... Figure 7 As shown, the display screen includes Bluetooth symbol 701, setting symbol 702, remaining battery symbol 703, and the components and contents of the sample to be tested, including nicotine 2%, total alkaloids 1.5%, total sugar 18%, reducing sugar 20%, total nitrogen 1.7%, and protein 8%, etc.

[0097] In this embodiment of the invention, by feeding back the component information to the display screen to inform the user, the component detection data service is provided to the user, reducing the operational complexity of the component detection device for the user.

[0098] In another embodiment of the present invention, the spectral-based component detection system may further include a built-in battery handle and a data acquisition button. The built-in battery handle serves as a user grip and is equipped with a pluggable battery compartment cover. The built-in battery powers the handheld hardware device in the spectral-based component detection system, and the data acquisition button is used to trigger the handheld hardware device to acquire spectral data.

[0099] Figure 8 This is a schematic diagram of the structure of the handheld hardware device provided by the present invention, as shown below. Figure 8 As shown, the handheld hardware device 800 includes a main body 801, a display screen 802, a built-in battery handle 803, an embedded calibration plate lens cover 804, and a data acquisition button 805. The display screen 802 is located at the rear of the handheld hardware device and directly displays the measured indicators after data acquisition. The built-in battery handle 803 serves as the handle for the user, and is the main part for gripping the device. A pluggable battery compartment cover is located below the handle for installing the device battery. The embedded calibration plate lens cover 804 contains a calibration whiteboard; when the lens cover is closed, the built-in calibration plate can be used for device calibration. The data acquisition button 805 has a conventional structure; when the user presses the button, the device acquires data, and the indicator results are automatically displayed on the display screen 802.

[0100] Figure 9 This is a schematic diagram of the main body of the device provided by the present invention, as shown below. Figure 9As shown, the main body 801 of the device includes a three-path differential probe 110, a visible band spectrometer 121, a near-infrared spectrometer 122, a control circuit board 130, and a built-in light source module. The built-in light source module includes a halogen light source 140. The built-in light source module is equipped with a cooling fan and heat sink structure for overall heat dissipation. The visible band spectrometer 121 is responsible for acquiring visible band spectral data, and the near-infrared spectrometer 122 is responsible for acquiring near-infrared band spectral data. Both spectrometers have built-in optical fibers that can transmit optical signals to the spectrometer's interior. The control circuit board 130 is responsible for overall circuit control and preprocesses the visible band and near-infrared band spectral data, and finally feeds the calculation results back to the screen for display. The three-path differential probe 110 is used to install the optical fibers of the visible band spectrometer, the near-infrared spectrometer, and the halogen light source, and the optical fibers are arranged according to the designed optical path.

[0101] The spectral-based component detection method provided by the present invention will be described below. The spectral-based component detection method described below can be referred to in correspondence with the spectral-based component detection system described above.

[0102] Figure 10 This is a schematic flowchart of the spectrum-based component detection method provided by the present invention, as shown below. Figure 10 This is applied to a spectral-based component detection system, which includes a three-path differential probe, a spectrometer, and a control circuit board. The spectral-based component detection method includes: Step 1001: The initial optical signal of the component to be measured is processed by the three-optical-path differential probe to eliminate interference, thereby obtaining the target optical signal.

[0103] Step 1002: Obtain the spectral data of the component to be measured based on the target light signal using the spectrometer.

[0104] Step 1003: Based on the spectral data and the target component detection model, the control circuit board determines the component information of the target component.

[0105] The target component detection model is used to detect the component information of the target component object by using a dual-path heterogeneous feature extraction network, an adaptive feature fusion module and an enhancement module to detect the spectral data. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local and global features of the spectral data based on the features of the spectral data, and to fuse the local features and global features based on the fusion strategy.

[0106] In this embodiment of the invention, the initial optical signal of the target component is processed by a three-path differential probe to eliminate interference, resulting in a target optical signal, which is then sent to a spectrometer. The spectrometer, based on the target optical signal, obtains the spectral data of the target component and sends it to a control circuit board. The control circuit board, based on the spectral data and a target component detection model, determines the component information of the target component. Thus, by using a three-path differential probe to eliminate strong light interference from the target component, the accuracy of the optical signal is improved. Furthermore, by using the target component detection model to analyze and detect the spectral data of the target component acquired by the spectrometer, component information is obtained without damaging the sample, shortening the detection time and improving the stability, accuracy, and efficiency of outdoor component detection.

[0107] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 11 As shown, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communications bus 1140, wherein the processor 1110, the communications interface 1120, and the memory 1130 communicate with each other through the communications bus 1140. The processor 1110 can call logic instructions in the memory 1130 to execute a spectral-based component detection method, applied to a spectral-based component detection system. The spectral-based component detection system includes a three-path differential probe, a spectrometer, and a control circuit board. The method includes: performing interference cancellation processing on the initial optical signal of the component to be tested through the three-path differential probe to obtain a target optical signal; obtaining spectral data of the component to be tested based on the target optical signal through the spectrometer; and determining the component information of the component to be tested based on the spectral data and a target component detection model through the control circuit board. The target component detection model is used to detect the spectral data to obtain the component information of the component to be tested based on a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy for local and global features of the spectral data based on the features of the spectral data, and fuse the local and global features based on the fusion strategy.

[0108] Furthermore, the logical instructions in the aforementioned memory 1130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0109] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the spectral-based component detection methods provided by the above methods and apply them to a spectral-based component detection system. The spectral-based component detection system includes a three-path differential probe, a spectrometer, and a control circuit board. The system includes: performing interference cancellation processing on the initial optical signal of the component to be tested through the three-path differential probe to obtain a target optical signal; obtaining spectral data of the component to be tested based on the target optical signal through the spectrometer; and determining the component information of the component to be tested based on the spectral data and a target component detection model through the control circuit board. The target component detection model is used to detect the spectral data to obtain the component information of the component to be tested based on a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy for local and global features of the spectral data based on the features of the spectral data, and fuse the local and global features based on the fusion strategy.

[0110] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program performs the spectral-based component detection method provided by the above methods, and is applied to a spectral-based component detection system. The spectral-based component detection system includes a three-path differential probe, a spectrometer, and a control circuit board. The system includes: performing interference cancellation processing on the initial optical signal of the component to be tested through the three-path differential probe to obtain a target optical signal; obtaining spectral data of the component to be tested based on the target optical signal through the spectrometer; and determining the component information of the component to be tested based on the spectral data and a target component detection model through the control circuit board. The target component detection model is used to detect the spectral data to obtain the component information of the component to be tested based on a dual-path heterogeneous feature extraction network, an adaptive feature fusion module, and an enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local and global features of the spectral data based on the features of the spectral data, and fuse the local and global features based on the fusion strategy.

[0111] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0112] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A spectrum-based component detection system, characterized in that, include: The system consists of a three-path differential probe, a spectrometer, and a control circuit board, among which: The three-path differential probe is used to perform interference cancellation processing on the initial optical signal of the component to be measured, obtain the target optical signal, and send the target optical signal to the spectrometer. The spectrometer is used to obtain the spectral data of the component to be measured based on the target light signal, and to send the spectral data to the control circuit board; The control circuit board is used to determine the component information of the target component object based on the spectral data and the target component detection model. The target component detection model is used to detect the spectral data to obtain the component information of the target component object based on the dual-path heterogeneous feature extraction network, the adaptive feature fusion module and the enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local features and global features of the spectral data based on the features of the spectral data, and fuse the local features and global features based on the fusion strategy.

2. The spectrum-based component detection system according to claim 1, characterized in that, The spectral-based component detection system further includes a halogen light source. The three-path differential probe includes a first optical fiber and an optical fiber group. The axis of the first optical fiber and the axis of each optical fiber in the optical fiber group point to the same position on the analyte. The optical paths on the first optical fiber and the optical paths of each optical fiber in the optical fiber group intersect at the acquisition plane. The first optical fiber is connected to the halogen light source, and the optical fiber group is connected to the spectrometer. The first optical fiber is used to transmit the light provided by the halogen light source to the sample to obtain the initial optical signal; The optical fiber array is used to perform interference cancellation processing on the initial optical signal to obtain the target optical signal, and then send the target optical signal to the spectrometer.

3. The spectrum-based component detection system according to claim 2, characterized in that, The optical fiber assembly includes a first sub-optical fiber and a second sub-optical fiber. The spectrometer includes a visible-band spectrometer and a near-infrared spectrometer. The spectral data includes visible-band spectral data and near-infrared spectral data. The first sub-optical fiber is connected to the visible-band spectrometer, and the second sub-optical fiber is connected to the near-infrared spectrometer, wherein: The visible band spectrometer is used to collect visible band spectral data based on the target optical signal transmitted through the first sub-optical fiber, and to send the visible band spectral data to the control circuit board. The near-infrared spectrometer is used to collect near-infrared spectral data based on the target light signal transmitted by the second sub-optical fiber, and to send the near-infrared spectral data to the control circuit board.

4. The spectrum-based component detection system according to any one of claims 1 to 3, characterized in that, The spectral-based component detection system also includes a cloud platform, which is used to incrementally train the initial component detection model based on the spectral data to obtain the target component detection model, and then send the target component detection model to the control circuit board.

5. The spectrum-based component detection system according to claim 4, characterized in that, The cloud includes an operating device, which is equipped with a cloud management system and is used to display the operating interface of the cloud management system.

6. The spectrum-based component detection system according to any one of claims 1 to 3, characterized in that, The control circuit board includes a preprocessing unit and a detection unit, wherein: The preprocessing unit is used to preprocess the spectral data to obtain target spectral data; The detection unit is used to input the target spectral data into the target component detection model to obtain the component information of the component to be tested.

7. The spectrum-based component detection system according to claim 6, characterized in that, The target component detection model is used to filter the target spectral data, obtain the key wavelengths in the target spectral data, and detect the key wavelengths to obtain the component information of the component to be tested.

8. The spectral-based component detection system according to any one of claims 1 to 3, characterized in that, The spectral-based component detection system also includes an embedded calibration plate lens cover, which is connected to the three-optical-path differential probe and is used to calibrate the spectral-based component detection system.

9. The spectrum-based component detection system according to claim 1, characterized in that, The spectral-based component detection system also includes a display screen for displaying component information of the component to be detected.

10. A spectral-based method for component detection, characterized in that, The spectral-based component detection system according to any one of claims 1 to 9, comprising a three-path differential probe, a spectrometer, and a control circuit board, includes: The target optical signal is obtained by performing interference cancellation processing on the initial optical signal of the component to be measured through the three-optical-path differential probe. The spectrometer obtains the spectral data of the component to be measured based on the target light signal; The control circuit board determines the component information of the target component object based on the spectral data and the target component detection model. The target component detection model is used to detect the spectral data to obtain the component information of the target component object based on the dual-path heterogeneous feature extraction network, adaptive feature fusion module and enhancement module. The adaptive feature fusion module is used to dynamically learn and generate a fusion strategy of local features and global features of the spectral data based on the features of the spectral data, and fuse the local features and global features based on the fusion strategy.

Images