Snapshot type hyperspectral imaging industrial online sorting system

Abstract

The invention discloses a snapshot type hyperspectral imaging industrial online sorting system which comprises a support assembly, an optical imaging assembly and a motion control assembly. The bracket assembly consists of a bottom plate, and an adjustable bracket and a fixed bracket which are fixed on the bottom plate; the optical imaging assembly comprises a lighting unit, an imaging objective lens and a snapshot type hyperspectral camera, and can synchronously obtain space image information of a measured object and spectral information of a continuous wave band in a single exposure process; the motion control assembly realizes automatic transmission of the measured object through the transmission platform; the data processing module is integrated with an image preprocessing unit, a spectral analysis unit and a spectral database, and is used for end-to-end closed-loop control from original hyperspectral data generation to sorting decision making; the system adopts a metasurface light splitting structure to realize real-time decoding, performs material or category identification by using a pre-trained classification model, and finally automatically generates a sorting decision signal according to an identification result to drive a motion control assembly to complete sorting operation.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial real-time detection, and in particular to a snapshot-type hyperspectral imaging industrial online sorting system. Background Art

[0002] With the advancement of industrial intelligence and automation, companies are placing higher demands on the accuracy and real-time nature of product quality testing. Traditional testing methods primarily include manual observation and point-based sensors. Manual inspection is highly subjective, has poor repeatability, and is inefficient, making it difficult to meet the online monitoring needs of high-speed production lines. While point-based sensors offer high accuracy, they have a limited sampling range, capturing only localized information and prone to misjudgment when dealing with heterogeneous or complex materials.

[0003] To address these issues, spectral imaging technology is increasingly being applied to industrial inspection. This technology fuses the spatial and spectral information of an image to form a "spectral image cube," enabling more comprehensive qualitative and quantitative analysis of the object being inspected. It is widely used in fields such as plastic recycling, agricultural product testing, and mineral identification.

[0004] Currently, mainstream spectral imaging and sorting systems include snapshot multispectral systems and push-broom hyperspectral systems. Snapshot multispectral systems offer fast response speeds and compact structures, but their limited spectral bands make it difficult to distinguish materials with similar spectral characteristics. Push-broom hyperspectral systems provide continuous spectral information and high resolution, but rely on mechanical scanning, which limits their speed and makes them susceptible to vibration and synchronization errors, making them unsuitable for high-speed assembly lines.

[0005] Therefore, it is necessary to provide a snapshot hyperspectral imaging industrial online sorting system to solve the above technical problems. Summary of the Invention

[0006] The present invention overcomes the deficiencies of the prior art and provides a snapshot-type hyperspectral imaging industrial online sorting system.

[0007] To achieve the above-mentioned object, the technical solution adopted by the present invention is: a snapshot hyperspectral imaging industrial online sorting device, comprising: a bracket assembly, an optical imaging assembly and a motion control assembly arranged on the bracket assembly;

[0008] The bracket assembly includes a base plate, an upper surface of which is fixedly connected to an adjustable bracket, an upper surface of which is fixedly connected to a fixed bracket, and the optical imaging assembly includes a lighting unit, an imaging objective lens, and a snapshot hyperspectral camera;

[0009] The motion control assembly includes: a transmission platform, which is fixedly mounted on one side of a fixed bracket; the transmission platform is used to transmit a plurality of objects to be measured;

[0010] One end of the snapshot hyperspectral camera is electrically connected to a data processing module.

[0011] In a preferred embodiment of the present invention, the model of the snapshot hyperspectral camera is xiQ MQ022HG-IM-SM4X4-VIS.

[0012] In a preferred embodiment of the present invention, the lighting unit is fixedly connected to one end of the adjustable bracket, the snapshot hyperspectral camera is fixedly connected to one end of the fixed bracket, and the imaging objective lens is fixedly installed at the bottom of the snapshot hyperspectral camera.

[0013] In a preferred embodiment of the present invention, the data processing module includes: an image preprocessing unit, a spectrum analysis unit and a spectrum database.

[0014] The present invention adopts a snapshot hyperspectral imaging industrial online sorting method, based on the above-mentioned snapshot hyperspectral imaging industrial online sorting device, comprising the following steps:

[0015] S1: The lighting unit provides illumination to the object under test. At the same time, the snapshot hyperspectral camera uses a single-frame exposure method to generate a two-dimensional coded image.

[0016] S2: The spectral reconstruction algorithm chip integrated in the snapshot hyperspectral camera processes the collected two-dimensional encoded image in real time, decodes the spectral information through the metasurface spectroscopic structure, and restores it to a three-dimensional hyperspectral data cube, covering the 400-960nm band;

[0017] S3: The image preprocessing unit in the data processing module performs preprocessing operations on the three-dimensional hyperspectral data cube;

[0018] S4: The spectrum analysis unit extracts the spectrum information of each pixel from the pre-processed hyperspectral data cube;

[0019] S5: Based on the spectral feature database, the spectral features of the measured object are identified and classified using a pre-trained classification model to generate a sorting decision signal; the transmission platform in the motion control component is driven to perform automated sorting operations to complete real-time sorting of the measured object.

[0020] In a preferred embodiment of the present invention, in S1, the snapshot hyperspectral camera synchronously acquires two-dimensional spatial image information and continuous-band spectral information of the object under test during a single imaging process.

[0021] In a preferred embodiment of the present invention, in S3, the pre-processing operation includes dark current correction, noise removal and data normalization processing to eliminate environmental interference and improve data quality.

[0022] In a preferred embodiment of the present invention, in said S4, the spectral information constructs a reflectivity or radiation characteristic curve within the wavelength range of 400 to 960 nm to form a spectral feature database.

[0023] In a preferred embodiment of the present invention, in said S5, the automated sorting operation is: controlling the movement direction of the transport platform according to the sorting decision signal.

[0024] The present invention adopts a snapshot hyperspectral imaging industrial online sorting system, based on the above-mentioned snapshot hyperspectral imaging industrial online sorting method, comprising:

[0025] Signal acquisition and synchronization module, used to obtain the hyperspectral image signal of the object to be measured;

[0026] The signal processing module is used to pre-process the hyperspectral image signal of the object to be measured and analyze the spatial distribution characteristics and spectral characteristics of the signal;

[0027] A machine learning classification model is used to identify the material or category of the pre-processed hyperspectral image signal and to generate a corresponding sorting decision signal based on the identification result;

[0028] Spectral segmentation module, used to segment signals of different materials or categories into independent data sets based on the recognition results;

[0029] The sorting execution module is used to drive the motion control component to complete the automated sorting operation of the objects to be tested according to the sorting decision signal.

[0030] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0031] (1) The present invention provides a snapshot hyperspectral imaging industrial online sorting system, which adopts the single-frame exposure characteristics of the snapshot hyperspectral camera and combines the metasurface spectroscopic structure to realize real-time decoding of spectral information and generation of three-dimensional hyperspectral data cube, significantly improving the real-time performance of the sorting system. This feature synchronously obtains the spatial image and spectral information of the measured object in a single imaging, avoiding the delay problem caused by multiple exposures required for traditional frame-based hyperspectral imaging. The direct effect is to shorten the sorting response time to milliseconds, which is far superior to the inefficient mode of relying on multi-frame splicing or offline processing in the existing technology; further, this feature can adapt to the dynamic needs of high-speed industrial production lines, while ensuring the sorting accuracy, enabling the system to handle the continuous sorting task of thousands of targets per minute, solving the technical bottleneck of the existing technology that is difficult to strike a balance between speed and accuracy.

[0032] (2) The present invention provides a snapshot hyperspectral imaging industrial online sorting system, which realizes end-to-end closed-loop control from raw hyperspectral data to sorting decisions through the setting of an image preprocessing unit, a spectral analysis unit and a spectral database integrated in the data processing module. The data processing module uses preprocessing methods such as dark current correction, noise removal and data normalization to effectively eliminate environmental interference and improve data quality. Compared with the limitations of spectral analysis in the existing technology that relies on manual experience or simple threshold judgment, this case further realizes the accurate recognition of complex materials or categories through the combination of machine learning classification models and spectral databases, significantly reduces the sorting error rate and expands the applicable scenarios of the system.

[0033] (3) The present invention provides a snapshot-type hyperspectral imaging industrial online sorting system, which greatly optimizes the spatial layout and installation efficiency of the system through the miniaturization advantages of adjustable brackets, fixed brackets and transmission platforms and metasurface spectroscopic structures. This design enables the coordinated accuracy of optical imaging components and motion control components to reach the micron level, which directly improves the stability and repeatability of sorting actions, compared with the sorting deviation problems caused by mechanical vibration or optical path offset in traditional systems; furthermore, the modular architecture reduces the equipment maintenance and upgrade costs, enabling the system to be quickly deployed in different industrial scenarios, solving the problems of high maintenance costs and low adaptability caused by complex structures in existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts.

[0035] Figure 1 This is a schematic structural diagram of a snapshot hyperspectral imaging industrial online sorting device of the present invention;

[0036] Figure 2 This is a schematic flow chart of a snapshot hyperspectral imaging industrial online sorting method of the present invention;

[0037] Figure 3 This is a block diagram of a snapshot hyperspectral imaging industrial online sorting system of the present invention.

[0038] In the figure: 101, bracket assembly; 11, adjustable bracket; 12, fixed bracket; 102, lighting unit; 103, snapshot hyperspectral camera; 104, imaging objective lens; 105, measured object; 106, transmission platform; 107, data processing module. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0041] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the scope of protection of the present application. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0042] In the description of this application, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on specific circumstances.

[0043] like Figure 1 As shown, the present invention provides a snapshot hyperspectral imaging industrial online sorting device, comprising: a bracket assembly 101, an optical imaging assembly and a motion control assembly arranged on the bracket assembly 101;

[0044] The bracket assembly 101 includes a base plate, an adjustable bracket 11 is fixedly connected to the upper surface of the base plate, and a fixed bracket 12 is fixedly connected to the upper surface of the base plate. The optical imaging assembly includes an illumination unit 102, an imaging objective lens 104, and a snapshot hyperspectral camera 103. The illumination unit 102 is fixedly connected to one end of the adjustable bracket 11, the snapshot hyperspectral camera 103 is fixedly connected to one end of the fixed bracket 12, and the imaging objective lens 104 is fixedly mounted on the bottom of the snapshot hyperspectral camera 103. The model of the snapshot hyperspectral camera 103 is xiQ MQ022HG-IM-SM4X4-VIS.

[0045] It should be noted that the snapshot hyperspectral camera 103 acquires a hyperspectral coded image covering the entire field of view in a very short time through a single-frame exposure method;

[0046] Furthermore, this means that the spatial information of the object under test 105 and the spectral information of the continuous band can be obtained synchronously within milliseconds, which greatly improves the data acquisition speed and is suitable for the needs of high-speed production lines.

[0047] like Figure 1 As shown, the motion control assembly includes: a transmission platform 106, which is fixedly mounted on one side of the fixed bracket 12; the transmission platform 106 is used to transmit a number of objects 105 to be measured;

[0048] One end of the snapshot hyperspectral camera 103 is electrically connected to a data processing module 107 . The data processing module 107 includes an image pre-processing unit, a spectrum analysis unit, and a spectrum database.

[0049] It should be noted that: based on the pre-trained machine learning classification model, the system can discriminate and classify the measured object 105 according to its spectral characteristics and generate corresponding sorting decision signals;

[0050] The transport platform 106 adjusts its movement direction through the servo motor according to the sorting decision signal, and automatically sorts objects of different categories into corresponding collection areas;

[0051] For example, “qualified products” continue to be transferred to the next process along the production line, while “unqualified products” are separated into the waste area;

[0052] Furthermore, the image pre-processing unit performs dark current correction, noise removal, and data normalization on the three-dimensional hyperspectral data cube to ensure data quality;

[0053] The spectral analysis unit extracts the spectral information of each pixel from the preprocessed data and constructs its reflectivity or radiation characteristic curve to form a spectral feature database, enabling the system to identify and classify materials.

[0054] Figure 2The figure shows a flow chart of a snapshot hyperspectral imaging industrial online sorting method according to an embodiment of the present application.

[0055] A snapshot hyperspectral imaging industrial online sorting method according to an embodiment of the present application includes the following steps:

[0056] S1: The lighting unit 102 provides lighting to the object 105. At the same time, the snapshot hyperspectral camera 103 uses a single-frame exposure method to generate a two-dimensional coded image.

[0057] S2: The spectral reconstruction algorithm chip integrated in the snapshot hyperspectral camera 103 processes the collected two-dimensional coded image in real time, decodes the spectral information through the metasurface spectroscopic structure, and restores it into a three-dimensional hyperspectral data cube, covering the 400-960nm band;

[0058] S3: The image preprocessing unit in the data processing module 107 performs a preprocessing operation on the three-dimensional hyperspectral data cube;

[0059] S4: The spectrum analysis unit extracts the spectrum information of each pixel from the pre-processed hyperspectral data cube;

[0060] S5: Based on the spectral feature database, the spectral features of the object 105 to be measured are identified and classified using a pre-trained classification model to generate a sorting decision signal; the transmission platform 106 in the motion control component is driven to perform automated sorting operations to complete real-time sorting of the object 105 to be measured.

[0061] Below, each step will be described in detail.

[0062] In step S1, first, in the implementation process of the present invention, the lighting unit 102 provides stable and uniform visible light illumination for the object 105 to be measured. The lighting unit 102 uses a high-brightness LED light source covering the 400-960nm band to ensure that the object 105 to be measured has sufficient reflection signals in the full visible light band;

[0063] The objects 105 to be measured are transported by the tracks of the transmission platform 106 and sequentially enter the field of view of the snapshot hyperspectral imaging camera 103. The snapshot hyperspectral imaging camera 103 adopts a single-frame exposure method and can obtain a hyperspectral coded image covering the entire field of view in a very short time.

[0064] Furthermore, the snapshot hyperspectral imaging camera 103 internally integrates a metasurface spectroscopic structure, which is a two-dimensional planar array composed of sub-wavelength-scale nanostructures;

[0065] These nanostructures are typically smaller than or close to the wavelength of the operating light (e.g., 400-960 nm), and are capable of regulating the incident light. Each nanostructure, through its unique geometry and arrangement, can change the phase, amplitude, or polarization state of the incident light, thereby achieving effective modulation of optical signals of different wavelengths.

[0066] Specifically, when light enters the snapshot hyperspectral imaging camera, it first comes into contact with this metasurface. Each nanostructure on the metasurface redistributes light signals of different wavelengths onto a two-dimensional plane according to specific design parameters, forming a two-dimensional encoded image containing spatial position and spectral information.

[0067] Furthermore, the specific design parameters are called the modulation function M(x, y, λ), which determines the change in phase, amplitude, or polarization state of the incident light at a specific position (x, y) and wave λ;

[0068] Geometry: The shape of the nanostructure (e.g., cylindrical, square, or other complex shapes) affects its modulation of the phase and amplitude of light;

[0069] Size and spacing: The size and spacing of the nanostructures are typically smaller than or close to the wavelength of the operating light, ensuring that the behavior of light can be effectively modulated;

[0070] Material properties: Different materials have different refractive indices and absorption characteristics. Choosing the right material can optimize the light regulation effect;

[0071] Furthermore, when light enters the snapshot hyperspectral imaging camera and strikes the metasurface, each nanostructure processes the incident light according to its modulation function M(x, y, λ). Specifically, light signals of different wavelengths are redistributed onto the two-dimensional plane according to the following mechanism:

[0072] Furthermore, by introducing a specific phase delay, light of different wavelengths can be propagated along different paths and ultimately form different spatial distributions on a two-dimensional plane.

[0073] Phase modulation can be expressed as:

[0074] Where: k is the wave number, M(x, y, λ) is the modulation function;

[0075] Furthermore, nanostructures can also change the amplitude of the incident light, thereby affecting its intensity distribution;

[0076] By modulating the amplitudes of light of different wavelengths differently, different intensity patterns can be formed on a two-dimensional plane;

[0077] The final two-dimensional coded image I(x,y) is the sum of all wavelength contributions: I(x,y) =

[0078] ∑ λ∈∧ S(λ)·M(x, y, λ);

[0079] in:

[0080] ∧ is the set of spectral bands (e.g. 400-960nm);

[0081] S(λ) is the spectral reflectance of the measured object 105 at wavelength λλ;

[0082] M(x, y, λ) is the modulation function of the metasurface spectroscopic structure, which represents the coding weight at the position (x, y) and wavelength λ.

[0083] For example, in the 400-960nm band, the metasurface can achieve efficient acquisition of more than 100 spectral channels with a field of view exceeding 55°. This means that each pixel not only contains information about the spatial position, but also implicitly about the wavelength of the incident light, enabling the acquisition of a complete hyperspectral data cube in a single exposure.

[0084] In this way, the metasurface spectroscopic structure efficiently captures and encodes multi-wavelength optical signals, enabling rapid acquisition of multispectral information without mechanical scanning, significantly improving detection speed and accuracy. In short, the metasurface spectroscopic structure is a plane composed of many tiny nanounits that can finely control the behavior of light, enabling the capture of complete hyperspectral data in a single shot.

[0085] In step S2, the spectral reconstruction algorithm chip integrated in the snapshot hyperspectral imaging camera 103 processes the collected two-dimensional coded image in real time. Through the spectral decoding algorithm of the metasurface spectroscopic structure, the two-dimensional coded image is converted into a three-dimensional hyperspectral data cube, covering the 400-960nm band and containing more than 100 spectral channels.

[0086] For example, the two-dimensional coded image I(x, y) is converted into a three-dimensional hyperspectral data cube H(x, y, λ) through the decoding algorithm of the metasurface spectroscopic structure: H(x, y, λ) = D(x, y, λ)·I(x, y);

[0087] Wherein, D(x, y, λ) is a decoding matrix, whose elements are determined by the inverse modulation function of the metasurface spectroscopic structure.

[0088] Finally, the spectral curve of each pixel is:

[0089] Each pixel corresponds to a complete spectral curve, recording detailed information on how its reflectivity or radiation characteristics change with wavelength.

[0090] This processing is completed through the on-chip computing unit without the need for external equipment support, significantly reducing data transmission delays and ensuring the real-time performance of the system.

[0091] In step S3 , the image preprocessing unit in the data processing unit 107 performs a preprocessing operation on the three-dimensional hyperspectral data cube.

[0092] Furthermore, the preprocessing operation specifically includes:

[0093] Dark current correction: Eliminate sensor dark current noise by acquiring dark field images under no-light conditions;

[0094] Dark current correction is performed by subtracting the mean of the dark field image D(x,y); I correctde (x, y, λ)=I (x, y, λ)-μD;

[0095] in, is the total number of pixels;

[0096] Noise removal: Wavelet noise removal algorithm is used to filter noise using wavelet transform to suppress random noise and salt and pepper noise;

[0097] The formula of the wavelet noise removal algorithm is:

[0098] where ψ j,k (x) is the scale function, j and k are the scale and displacement parameters respectively.

[0099] Data normalization: Normalize the spectral data to eliminate the effects of ambient light fluctuations and sensor response differences.

[0100] The formula for normalizing spectral data is:

[0101] The quality of the preprocessed data is significantly improved, providing input data with a high signal-to-noise ratio for subsequent spectral analysis.

[0102] In step S4, the spectrum analysis unit extracts the spectrum information of each pixel from the pre-processed hyperspectral data cube and constructs its reflectivity or radiation characteristic curve in the 400-960 nm band.

[0103] Key features related to the material, composition or state of the object 105 to be measured are extracted through a feature extraction algorithm (such as principal component analysis PCA feature extraction).

[0104] Furthermore, principal component analysis (PCA) is used to extract features, calculate the covariance matrix C, and solve the eigenvalue decomposition:

[0105]

[0106] Among them, X i is the spectrum vector of the i-th sample, is the mean vector;

[0107] Select the first k principal components As features:

[0108] These features are matched with the spectral feature library of known materials to build a dynamically updated spectral feature database.

[0109] In step S5, based on the spectral feature database, the spectral features of the object 105 are identified and classified using a pre-trained classification model;

[0110] further,

[0111] The classification model can be a convolutional neural network (CNN), whose training data includes sample spectral characteristics in industrial scenarios such as plastics, agricultural products, and minerals;

[0112] Furthermore, the convolution operation of CNN can be expressed as:

[0113] Where O(i,j) is the output feature map, is the weight of the Kth convolution kernel, and r is the radius of the convolution kernel.

[0114] The model classification result is a category label of the object 105 being tested (such as “qualified product”, “unqualified product”, “specific material”, etc.).

[0115] According to the classification result, the system generates a sorting decision signal. According to the sorting decision signal, the transmission platform 106 adjusts the movement direction through the servo motor to sort the objects 105 to the corresponding collection area.

[0116] The sorting decision signal is an instruction generated after the object 105 is identified and classified based on the classification model. The sorting decision signal is essentially an electronic or digital signal that instructs the transmission platform 106 how to adjust its movement direction based on the classification result of the object 105, thereby guiding the object to the correct collection area.

[0117] For example:

[0118] Conformable product: If the model classification result indicates that the object is a conforming product, the generated decision signal will instruct the transport platform 106 to continue transporting to the next process along the production line without changing the current path;

[0119] Rejected products: If an object is marked as rejected, the corresponding decision signal will cause the servo motor to adjust the direction of the transport platform 106, making it deviate from the main line and sorting and separating the object into the waste area;

[0120] Specific materials: For specific materials that require special processing, there will also be corresponding decision signals to guide the transmission platform 106 to guide them to a dedicated collection area.

[0121] like Figure 3 As shown, a snapshot hyperspectral imaging industrial online sorting system according to an embodiment of the present application includes:

[0122] Signal acquisition and synchronization module, used to obtain the hyperspectral image signal of the object to be measured;

[0123] The signal processing module is used to pre-process the hyperspectral image signal of the object to be measured and analyze the spatial distribution characteristics and spectral characteristics of the signal;

[0124] A machine learning classification model is used to identify the material or category of the pre-processed hyperspectral image signal and to generate a corresponding sorting decision signal based on the identification result;

[0125] Spectral segmentation module, used to segment signals of different materials or categories into independent data sets based on the recognition results;

[0126] The sorting execution module is used to drive the motion control component to complete the automated sorting operation of the objects to be tested according to the sorting decision signal.

[0127] In one example, the signal acquisition and synchronization module includes an illumination unit 102 and a snapshot hyperspectral camera 103. The illumination unit 102 provides a uniform light source, ensuring the quality of the acquired hyperspectral image. The snapshot hyperspectral camera 103 is used to capture hyperspectral information of the entire scene in a single shot, ensuring data consistency and accuracy.

[0128] In one example, the signal processing module includes: an image preprocessing unit, a spectral analysis unit and a computing unit. The image preprocessing unit is responsible for denoising, correcting and other processing of the original hyperspectral image; the spectral analysis unit extracts the spectral characteristics of each pixel point and constructs a reflectivity or radiation characteristic curve; the computing unit is used to analyze these characteristics and determine the material properties or categories.

[0129] In one example, a machine learning classification model uses a convolutional neural network (CNN) to classify and identify the input hyperspectral image data through a trained model, accurately determine the material or category of the object 105 being measured, and generate a corresponding sorting decision signal accordingly.

[0130] In one example, the spectral segmentation module divides groups of pixels with similar spectral characteristics into different data sets based on the output results of the machine learning classification model. Each data set represents a specific material or category, which facilitates subsequent processing and analysis.

[0131] In one example, the sorting execution module includes: a transmission platform 106 and a servo motor. The transmission platform 106 carries the items to be sorted and moves along a set path. After receiving the sorting decision signal, the servo motor adjusts the direction or speed of the transmission platform 106, thereby achieving efficient and accurate sorting of the objects to be tested.

[0132] Those skilled in the art will understand that other details of the snapshot hyperspectral imaging industrial online sorting system according to the embodiment of the present application are the same as the corresponding details previously described in the machine learning-based radio signal recognition method according to the embodiment of the present application, and will not be repeated here to avoid repetition.

[0133] The above description is based on the ideal embodiment of the present invention. Based on the above description, relevant personnel can make various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the content of the specification and must be determined according to the scope of the claims.

Claims

1. A snapshot hyperspectral imaging industrial online sorting device, comprising: A support assembly (101), an optical imaging assembly and a motion control assembly arranged on the support assembly (101), characterized in that: The support assembly (101) includes a base plate, an upper surface of the base plate is fixedly connected to an adjustable support (11), an upper surface of the base plate is fixedly connected to a fixed support (12), and the optical imaging assembly includes a lighting unit (102), an imaging objective lens (104), and a snapshot hyperspectral camera (103); The motion control component comprises: a transmission platform (106), wherein the transmission platform (106) is fixedly mounted on one side of a fixed bracket (12); the transmission platform (106) is used to transmit a plurality of objects to be measured (105); One end of the snapshot hyperspectral camera (103) is electrically connected to a data processing module (107).

2. The snapshot hyperspectral imaging industrial online sorting device according to claim 1, characterized in that: The model of the snapshot hyperspectral camera (103) is xiQ MQ022HG-IM-SM4X4-VIS.

3. The snapshot hyperspectral imaging industrial online sorting device according to claim 1, characterized in that: The lighting unit (102) is fixedly connected to one end of the adjustable bracket (11), the snapshot hyperspectral camera (103) is fixedly connected to one end of the fixed bracket (12), and the imaging objective lens (104) is fixedly installed at the bottom of the snapshot hyperspectral camera (103).

4. The snapshot hyperspectral imaging industrial online sorting system according to claim 1, characterized in that: The data processing module (107) includes: an image preprocessing unit, a spectrum analysis unit and a spectrum database.

5. A snapshot hyperspectral imaging industrial online sorting method according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1: The lighting unit (102) provides lighting to the object to be measured (105), and at the same time, the snapshot hyperspectral camera (103) uses a single-frame exposure method to generate a two-dimensional coded image; S2: The spectral reconstruction algorithm chip integrated in the snapshot hyperspectral camera (103) processes the collected two-dimensional coded image in real time, decodes the spectral information through the metasurface spectroscopic structure, and restores it into a three-dimensional hyperspectral data cube, covering the 400-960nm band range; S3: The image pre-processing unit in the data processing module (107) performs a pre-processing operation on the three-dimensional hyperspectral data cube; S4: The spectrum analysis unit extracts the spectrum information of each pixel from the pre-processed hyperspectral data cube; S5: Based on the spectral feature database, the spectral features of the object to be measured (105) are discriminated and classified using a pre-trained classification model to generate a sorting decision signal; the transmission platform (106) in the motion control component is driven to perform an automated sorting operation to complete the real-time sorting of the object to be measured (105).

6. The snapshot hyperspectral imaging industrial online sorting method according to claim 1, characterized in that: In the above-mentioned S1, the snapshot hyperspectral camera (103) synchronously acquires two-dimensional spatial image information and continuous-band spectral information of the object to be measured (105) in a single imaging process.

7. The snapshot hyperspectral imaging industrial online sorting method according to claim 1, characterized in that: In S3 , the pre-processing operation includes dark current correction, noise removal, and data normalization processing to eliminate environmental interference and improve data quality.

8. The snapshot hyperspectral imaging industrial online sorting method according to claim 1, characterized in that: In the above S4, the spectral information is used to construct a reflectivity or radiation characteristic curve within the wavelength range of 400 to 960 nm, thereby forming a spectral feature database.

9. The snapshot hyperspectral imaging industrial online sorting method according to claim 1, characterized in that: In said S5, the automated sorting operation is: controlling the movement direction of the transport platform (106) according to the sorting decision signal.

10. A snapshot hyperspectral imaging industrial online sorting system, based on the sorting method according to any one of claims 1 to 9, characterized in that: include: Signal acquisition and synchronization module, used to obtain the hyperspectral image signal of the object to be measured; The signal processing module is used to pre-process the hyperspectral image signal of the object to be measured and analyze the spatial distribution characteristics and spectral characteristics of the signal; A machine learning classification model is used to identify the material or category of the pre-processed hyperspectral image signal and to generate a corresponding sorting decision signal based on the identification result; Spectral segmentation module, used to segment signals of different materials or categories into independent data sets based on the recognition results; The sorting execution module is used to drive the motion control component to complete the automated sorting operation of the objects to be tested according to the sorting decision signal.

Images