Segmentation and optimization method and system based on multispectral spectral image structure

Abstract

The application discloses a segmentation and optimization method and system based on multispectral spectral image structure, relates to the technical field of spectral image processing and image optimization, and collects multispectral spectral data of a to-be-measured slice; the multispectral spectral data is preprocessed to obtain preprocessed multispectral spectral data; the preprocessed multispectral spectral data is fused to obtain tower spectral data; the tower spectral data is divided into a plurality of spectral subsegments according to wavelength intervals, a spectral anomaly evaluation score is calculated for each spectral subsegment, an abnormal spectral subsegment is screened based on an adaptive threshold, and a primary spectral structure region is formed based on the abnormal spectral subsegment; based on the primary spectral structure region, adaptive fine tuning is performed on collection parameters, and multispectral spectral data is re-collected; the above steps are repeated on the new multispectral spectral data until the region boundary converges, and a final spectral region of interest is obtained, so that dynamic enhancement and segmentation optimization of the spectral image are realized.

Description

Segmentation and Optimization Method and System Based on Multi-band Spectral Image Structure Technical Field

[0001] This invention relates to the field of spectral image processing and image optimization technology, and in particular to a segmentation and optimization method and system based on multi-band spectral image structure. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Current pathological section examination mainly relies on optical microscopy imaging and staining assessment. The image quality is prone to fluctuation with changes in section thickness, staining depth, and lighting conditions, making it difficult to guarantee spatial consistency. In vivo tissue spectral detection typically uses single-band or fixed-parameter fiber optic acquisition methods, which cannot optimize image quality based on changes in tissue optical properties, resulting in insufficient signal-to-noise ratio or low local contrast in some tissue areas.

[0004] Traditional spectral imaging methods are mostly one-time acquisitions, lacking the ability to jointly model the structure of multi-band spectral images. This makes it difficult to improve image contrast and enhance spectral structure differences across the entire spectrum, and also makes it impossible to dynamically adjust acquisition parameters based on local spectral image features, ultimately affecting the stability of image segmentation and subsequent analysis. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies in terms of spectral image quality, image structure stability, and parameter unadjustability, this invention provides a segmentation and optimization method and system based on multi-band spectral image structure, aiming to achieve dynamic enhancement and segmentation optimization of spectral images, thereby obtaining more stable and clearer tissue structure regions.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, the present invention provides a segmentation and optimization method based on multi-band spectral image structure, including:

[0008] Collect multi-band spectral data of the slice to be tested;

[0009] The multi-band spectral data is preprocessed to obtain preprocessed multi-band spectral data;

[0010] The preprocessed multi-band spectral data are fused to obtain tower-shaped spectral data;

[0011] The tower-shaped spectral data is divided into multiple spectral segments according to wavelength range. A spectral anomaly evaluation score is calculated for each spectral segment. Anomaly spectral segments are then filtered based on an adaptive threshold, and an initial spectral structure region is formed based on the anomaly spectral segments.

[0012] Based on the initial spectral structure region, the acquisition parameters are adaptively fine-tuned, and multi-band spectral data are reacquired; the above steps are repeated on the new multi-band spectral data until the region boundary converges, and the final spectral region of interest is obtained.

[0013] In a further technical solution, the multi-spectral data includes visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images.

[0014] In a further technical solution, the preprocessing includes baseline correction and smoothing.

[0015] A further technical solution involves fusing the preprocessed multi-band spectral data as follows: using the visible light image as a base, reconstructing the preprocessed hyperspectral image, fluorescence image, and near-infrared image, and then fusing and superimposing them with the visible light image to obtain tower-shaped spectral data.

[0016] Further technical solutions, specifically the integration, are as follows:

[0017] Principal component analysis was used to vectorize and reduce the dimensions of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images, respectively.

[0018] Statistical features of spectral sub-segments of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images are calculated respectively to obtain structural features of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images.

[0019] Calculate the Pearson correlation coefficients between the structural features of visible light images and the structural features of near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images, respectively.

[0020] Using visible light images as a base, hyperspectral reflectance images, fluorescence images, and near-infrared reflectance images are normalized to obtain processed image data.

[0021] The processed image data are stacked based on the Pearson correlation coefficient to obtain tower-shaped spectral data.

[0022] A further technical solution is that the anomaly evaluation score is represented as:

[0023]

[0024] in, For abnormal evaluation scores, For spectral deviation, The rate of change of spectral morphology. Due to differences in spectral energy, , , These are the weighting factors for spectral deviation, spectral shape change rate, and spectral energy difference, respectively.

[0025] A further technical solution, which forms the initial spectral structure region based on the abnormal spectral segments, specifically involves mapping the selected abnormal spectral segments to a spatial region and performing regional aggregation processing to generate the spectral structure region.

[0026] Secondly, the present invention provides a segmentation and optimization system based on multi-band spectral image structure, comprising:

[0027] The data acquisition module is configured to acquire multi-band spectral data of the slice to be tested;

[0028] The preprocessing module is configured to preprocess the multi-band spectral data to obtain preprocessed multi-band spectral data.

[0029] The spectral fusion module is configured to fuse preprocessed multi-band spectral data to obtain tower-shaped spectral data.

[0030] The anomaly detection module is configured to: divide the tower-shaped spectral data into multiple spectral segments according to wavelength range, calculate the spectral anomaly evaluation score for each spectral segment, filter out abnormal spectral segments based on an adaptive threshold, and form an initial spectral structure region based on the abnormal spectral segments.

[0031] The dynamic optimization module is configured to: adaptively fine-tune the acquisition parameters based on the initial spectral structure region, and reacquire multi-band spectral data; repeat the above steps on the new multi-band spectral data until the region boundary converges, and obtain the final spectral region of interest.

[0032] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the segmentation and optimization method based on multi-band spectral image structure as described in the first aspect.

[0033] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the segmentation and optimization method based on multi-band spectral image structure as described in the first aspect.

[0034] The above one or more technical solutions have the following beneficial effects:

[0035] This invention constructs a "tower-shaped spectral image" structure that integrates multiple spectral bands, and combines spectral structure clustering, spectral anomaly measurement, and image region segmentation to achieve dynamic enhancement and segmentation optimization of spectral images, thereby obtaining more stable and clearer tissue structure regions.

[0036] This invention achieves full-process optimization of spectral images from acquisition and fusion to regional structured representation through three innovations: multi-band spectral image fusion, spectral structure clustering, and spectral image segmentation. This can significantly improve the accuracy and stability of regional segmentation of spectral images.

[0037] This invention integrates visible light, near-infrared, hyperspectral, and fluorescence imaging channels to extract representative bands from the staining channels of tissue sections and the endogenous or exogenous fluorescence of living tissues. It innovatively proposes a "tower"-shaped spectral data structure, which contains spatial information and multi-band spectral information from tissue sections or living tissues, thereby achieving joint imaging of the entire cellular spectrum. Attached Figure Description

[0038] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0039] Figure 1 is a flowchart of the segmentation and optimization method based on multi-band spectral image structure according to an embodiment of the present invention;

[0040] Figure 2 is a schematic diagram of multi-band spectral data fusion according to an embodiment of the present invention;

[0041] Figure 3 is a flowchart of the dynamic spectral anomaly region division according to an embodiment of the present invention. Detailed Implementation

[0042] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0044] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0045] Example 1

[0046] As shown in Figures 1 and 3, this embodiment discloses a segmentation and optimization method based on multi-band spectral image structure. The method includes the following steps:

[0047] S1: Acquire multi-band spectral data of the slice to be tested;

[0048] In this embodiment, a hyperspectral camera, a fluorescence excitation / imaging module, and a near-infrared imager are arranged on the slicing stage to simultaneously acquire multi-band spectral data of the collected slices or living tissue, including visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images, with a frame rate of 10–30 Hz, to ensure seamless coverage of the absorption peaks and intrinsic fluorescence characteristics of each component of the tissue.

[0049] S2: Preprocess the multi-band spectral data to obtain preprocessed multi-band spectral data;

[0050] Preprocessing includes baseline correction and smoothing. The acquired spectral data is first subjected to baseline correction and Savitzky-Golay smoothing to reduce noise.

[0051] Baseline correction obtains the baseline fitting curve and calculates the residual using the least squares method. The baseline calibration formula is as follows:

[0052]

[0053] in, The residual value, The original spectrum, To fit the spectrum, This refers to the number of bands.

[0054] The Savitzky–Golay smoothing formula is as follows:

[0055]

[0056] in, For the window fitting value matrix, For a variable matrix, The coefficient matrix is ​​the polynomial fitting matrix. The residual matrix is... To fit the maximum number of times, This refers to the window size.

[0057] S3: The preprocessed multi-band spectral data are fused to obtain tower-shaped spectral data (tower-shaped spectral image structure, containing multi-band information).

[0058] In this embodiment, the present invention proposes a data reconstruction method for multi-band spectral data fusion. As shown in Figure 2, the visible light image is used as a base to provide a clear reference for the new data structure. The preprocessed hyperspectral image, fluorescence image and near-infrared image are then reconstructed according to bands and fused and superimposed with the visible light image to obtain a new data structure, which is called the "tower" type data structure.

[0059] The specific steps of data fusion are as follows:

[0060] The algorithm principle is as follows: First, principal component analysis is used to vectorize and reduce the dimensionality of the four spectral matrices. The statistical characteristics of the spectral sub-segments of the visible light image are denoted as... The statistical characteristics of spectral segments in a hyperspectral image are denoted as... The statistical characteristics of the spectral sub-segments of the fluorescence image are denoted as... The statistical characteristics of the spectral sub-segments of near-infrared images are denoted as... ; then calculate separately and , , Pearson correlation coefficient , , The calculation formula is:

[0061]

[0062] in, This represents the Pearson correlation coefficient. ; This represents the number of data points in the statistical characteristics of the spectral sub-segment. for The first vector One data point, for , , The first vector Data points.

[0063] Then, using the visible light image as a base, the hyperspectral image, fluorescence image, and near-infrared image are normalized using the following formula:

[0064]

[0065] in, This represents the normalized image data. Image data representing visible light, Image data representing hyperspectral images, fluorescence images, and near-infrared images, respectively.

[0066] Finally, the processed image data are stacked to obtain a "tower"-shaped data structure resulting from the fusion of multi-band spectral data. The formula is as follows:

[0067]

[0068] in, This represents "tower" shaped data that integrates spatial information and multi-band spectral information.

[0069] The tower-shaped spectral structure was obtained by fusing visible light, near-infrared, hyperspectral, and fluorescence spectra using a correlation-weighted method, with the correlation determined by the Pearson correlation coefficient. Multi-band data fusion achieved complementarity and enhancement of information from different spectral channels through multi-layer stacking and weighted normalization.

[0070] S4: Divide the tower-shaped spectral data into multiple spectral segments according to wavelength ranges, calculate the spectral anomaly evaluation score for each spectral segment, and filter out anomalous spectral segments based on an adaptive threshold. Based on these anomalous spectral segments, form the initial spectral structure region. Specifically, this includes the following steps:

[0071] 1. Divide the tower-shaped spectral structure into multiple spectral segments according to a preset wavelength window:

[0072]

[0073] in, For a set of spectral segments, For the first Each spectral segment, Each segment is a multi-wavelength spectral sequence. The spectral segments are divided according to a preset wavelength window or band boundaries adaptively determined based on the spectral rate of change.

[0074] 2. To assess the degree of deviation of spectral segments from normal spectral modes, this invention constructs a comprehensive scoring function composed of multiple spectral anomaly evaluation indicators:

[0075]

[0076] in, For abnormal evaluation scores, This represents the spectral deviation (integral of absolute difference). This represents the rate of change in spectral morphology (difference in derivatives). Due to differences in spectral energy, , , These are the weighting factors for spectral deviation, spectral shape change rate, and spectral energy difference, respectively. , , These are variables that are either predefined constants based on spectral quality indicators or adaptively updated based on spectral score distributions.

[0077] The three evaluation indicators are defined as follows:

[0078] (1) Spectral deviation :

[0079] Used to measure the overall deviation of a spectral segment from the global mean spectrum:

[0080]

[0081] in, It is a global average spectral sequence.

[0082] (2) Spectral morphology change rate :

[0083] Used to evaluate the anomaly of spectral shape variation trends:

[0084]

[0085] in, For the first Each spectral segment, For the first A global average spectral sequence.

[0086] (3) Spectral energy difference :

[0087] Used to assess whether there is a significant shift in the overall energy distribution of the spectrum:

[0088]

[0089] in, For the first The square of each spectral segment, This represents the average energy across all spectral segments.

[0090] 3. Adaptive threshold construction and updating.

[0091] An adaptive threshold is constructed based on the anomaly evaluation score distribution of the spectral segments:

[0092]

[0093] in, As the new anomaly threshold, The old anomaly threshold, For adjustment coefficients, The mean of the sub-segments with scores higher than the median. The mean of the sub-segments whose scores are below the median.

[0094] This method enables adaptive compensation for tissue spectral heterogeneity.

[0095] 4. Abnormal segment filtering and spatial mapping.

[0096] When a certain segment satisfies:

[0097]

[0098] When the anomaly evaluation score of a certain segment is greater than the new anomaly threshold, it is identified as an anomalous spectral segment. The spatial position of the wavelength corresponding to the anomalous segment in the tower-shaped spectral structure is mapped to the coordinate plane, and aggregation and morphological closing operations are performed on the adjacent regions. In other words, the anomalous spectral segment is mapped to the spatial coordinates of the tower-shaped spectral structure, and morphological closing operations are used to perform region fusion to form the initial spectral structure region (spectral anomaly region).

[0099] S5: Based on the initial spectral structure region, the acquisition parameters are adaptively fine-tuned, and multi-band spectral data is reacquired; the steps of spectral sub-segment division, spectral structure clustering, and abnormal region segmentation are repeated on the new multi-band spectral data until the region boundary converges, and the final spectral region of interest is obtained.

[0100] After obtaining the initial spectral structure region, the system adaptively adjusts the acquisition parameters (including integration time, light source power, sampling density, etc.) based on the spectral signal-to-noise ratio, spectral energy distribution, and local contrast index of that region. The adjusted acquisition parameters are used to regenerate multi-band spectral data and repeat the spectral structure analysis and image region segmentation process.

[0101] Through a closed-loop mechanism of acquisition, processing, and feedback, the system automatically corrects the segmentation threshold of the spectral region in each iteration and dynamically refines the spectral structure boundary for unstable or low signal-to-noise ratio regions, so that the segmentation result of the final spectral image region gradually stabilizes.

[0102] The aforementioned closed-loop optimization process is executed cyclically within a preset period. Through the coupling of parameter fine-tuning and regional structure updates, the system can maintain an adaptive response to changes in tissue optical properties, differences in new samples, or environmental disturbances, ultimately obtaining a converged and stable spectral region of interest.

[0103] In some implementations, this embodiment provides an implementation combining spectral structure analysis and image segmentation in the spectral structure region segmentation and dynamic optimization based on tower-shaped spectral images, illustrating the specific implementation process of the method of the present invention on multi-band spectral images. This process does not require image classification or medical diagnosis; it only performs structured processing, region division, and image enhancement on the spectral image, and belongs entirely to the field of image processing technology.

[0104] (1) Construction of tower-shaped spectral images

[0105] After preprocessing visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images, a multi-spectral fused image is generated through band reconstruction, spatial alignment, and intensity normalization. Subsequently, using the visible light image as a base, different band spectra are superimposed in wavelength order to construct a tower-shaped spectral image structure.

[0106] Each layer in the tower-shaped structure corresponds to different spectral information, and the spectral differences between different layers reflect the changes in the spectral structure of the tissue region.

[0107] (2) Spectral segment division and structural quantization

[0108] The tower-shaped spectral image is divided into multiple spectral segments according to wavelength or spectral energy distribution, and spectral structure statistics for each segment are calculated, such as: mean spectral intensity, energy spectrum distribution, local gradient change, spectral deviation, and rate of change of the first and second derivatives.

[0109] These quantitative metrics are used to describe structural differences in spectral segments, rather than for any identification purpose.

[0110] (3) Region clustering based on spectral structure similarity

[0111] Unsupervised clustering is used to group the structural statistics of each spectral segment to form a set of segments with similar spectral structures, i.e., candidate spectral regions.

[0112] The purpose of clustering is to obtain structurally consistent regions between spectral layers, which facilitates subsequent image segmentation, without involving classification or recognition.

[0113] (4) Image region segmentation driven by spectral anomaly evaluation function

[0114] For candidate spectral regions obtained by clustering, calculate spectral anomaly evaluation scores, including but not limited to: spectral deviation, spectral energy difference, and spectral morphology change (derivative difference).

[0115] The evaluation results are mapped to the spatial location corresponding to the tower-shaped spectral image, and the initial spectral structure region (initial spectral anomaly region) is generated through threshold segmentation and regional connectivity analysis. Threshold segmentation involves filtering spectral anomaly segments through thresholds and mapping them to the image space to form continuous spectral structure regions.

[0116] To improve the continuity of regional boundaries, morphological closing operations and boundary smoothing processes can be performed on the segmentation results to make the spectral structure regions more stable and coherent.

[0117] (5) Closed-loop optimization of acquisition parameters based on spectral region quality

[0118] For the initial spectral structure region, the system dynamically adjusts the acquisition parameters (such as integration time, light source power, or scanning step) based on its spectral signal-to-weight ratio, spectral contrast, and spectral consistency within the region.

[0119] The adjusted parameters are used to reacquire spectral images and repeat: spectral segmentation, spectral structure quantization, spectral clustering, and region segmentation.

[0120] In each iteration, the spectral region boundary gradually converges, eventually yielding a stable spectral region of interest.

[0121] (6) Iterative convergence criterion

[0122] The iteration stopping conditions can be selected as follows: the change of the region boundary is lower than the set threshold for two consecutive rounds, the change of the spectral anomaly evaluation score tends to be stable, and the spectral signal-to-noise ratio reaches the preset standard in the local area.

[0123] The final output is a stable region structure obtained based on multi-band spectral images, rather than any classification or diagnostic results.

[0124] In other implementations, for candidate spectral regions, a spectral image transformation network based on multi-scale convolution is used to perform structural enhancement and region segmentation on the spectral image to obtain an initial region of interest. Based on the spectral structure consistency of the initial region, closed-loop adjustments are performed on the acquisition parameters, and the spectral image is regenerated to achieve dynamic optimization of the tower-shaped spectral image. The above steps are iterated until the region segmentation results converge, ultimately obtaining a stable region of interest.

[0125] Example 2

[0126] This embodiment discloses a segmentation and optimization system based on multi-band spectral image structure, including:

[0127] The data acquisition module is configured to acquire multi-band spectral data of the slice to be tested;

[0128] The preprocessing module is configured to preprocess the multi-band spectral data to obtain preprocessed multi-band spectral data.

[0129] The spectral fusion module is configured to fuse preprocessed multi-band spectral data to obtain tower-shaped spectral data.

[0130] The anomaly detection module is configured to: divide the tower-shaped spectral data into multiple spectral segments according to wavelength range, calculate the spectral anomaly evaluation score for each spectral segment, filter out abnormal spectral segments based on an adaptive threshold, and form an initial spectral structure region based on the abnormal spectral segments.

[0131] The dynamic optimization module is configured to: adaptively fine-tune the acquisition parameters based on the initial spectral structure region, and reacquire multi-band spectral data; repeat the preprocessing, fusion, spectral segment division, abnormal segment screening and structural region segmentation steps on the new multi-band spectral data until the region boundary converges to obtain the final spectral region of interest.

[0132] Example 3

[0133] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method of Embodiment 1.

[0134] Example 4

[0135] The purpose of this embodiment is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method of Embodiment 1.

[0136] The steps and methods involved in the apparatuses of Embodiments 3 and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0137] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0138] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0139] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A segmentation and optimization method based on multi-band spectral image structure, characterized in that, include: Collect multi-band spectral data of the slice to be tested; The multi-band spectral data is preprocessed to obtain preprocessed multi-band spectral data; The preprocessed multi-band spectral data are fused to obtain tower-shaped spectral data. Specifically, the visible light image is used as the base, and the preprocessed hyperspectral image, fluorescence image and near-infrared image are reconstructed and fused with the visible light image to obtain tower-shaped spectral data. The tower-shaped spectral data is divided into multiple spectral segments according to wavelength ranges. A spectral anomaly evaluation score is calculated for each spectral segment, and anomaly spectral segments are filtered based on an adaptive threshold. An initial spectral structure region is then formed based on these anomaly spectral segments. The anomaly evaluation score is expressed as: in, For abnormal evaluation scores, For spectral deviation, The rate of change of spectral morphology. Due to differences in spectral energy, 、 、 These are the weighting factors for spectral deviation, spectral shape change rate, and spectral energy difference, respectively. Based on the initial spectral structure region, the acquisition parameters are adaptively fine-tuned, and multi-band spectral data are reacquired; the above steps are repeated on the new multi-band spectral data until the region boundary converges, and the final spectral region of interest is obtained.

2. The segmentation and optimization method based on multi-band spectral image structure as described in claim 1, characterized in that, The multi-spectral data includes visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images.

3. The segmentation and optimization method based on multi-band spectral image structure as described in claim 1, characterized in that, The preprocessing includes baseline correction and smoothing.

4. The segmentation and optimization method based on multi-band spectral image structure as described in claim 1, characterized in that, The fusion process involves: using principal component analysis to vectorize and reduce the dimensions of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images; calculating the statistical characteristics of the spectral sub-segments of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images to obtain the structural features of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images. Pearson correlation coefficients were calculated for the structural features of visible light images, near-infrared reflectance images, hyperspectral reflectance images, and fluorescence images, respectively. Using the visible light image as a basis, the hyperspectral reflectance image, fluorescence image, and near-infrared reflectance image were normalized to obtain processed image data. The processed image data were then stacked based on the Pearson correlation coefficients to obtain tower-shaped spectral data.

5. The segmentation and optimization method based on multi-band spectral image structure as described in claim 1, characterized in that, The process of forming the initial spectral structure region based on anomalous spectral segments is as follows: the selected anomalous spectral segments are mapped to a spatial region and then aggregated to generate the spectral structure region.

6. A segmentation and optimization system based on multi-band spectral image structure, employing the segmentation and optimization method based on multi-band spectral image structure as described in any one of claims 1-5, characterized in that, include: The data acquisition module is configured to acquire multi-band spectral data of the slice to be tested; The preprocessing module is configured to preprocess the multi-band spectral data to obtain preprocessed multi-band spectral data. The spectral fusion module is configured to fuse preprocessed multi-band spectral data to obtain tower-shaped spectral data. The anomaly detection module is configured to: divide the tower-shaped spectral data into multiple spectral segments according to wavelength ranges, calculate the spectral anomaly evaluation score for each spectral segment, and filter out abnormal spectral segments based on an adaptive threshold, forming an initial spectral structure region based on the abnormal spectral segments; the dynamic optimization module is configured to: adaptively fine-tune the acquisition parameters based on the initial spectral structure region, and reacquire multi-band spectral data; repeat the above steps on the new multi-band spectral data until the region boundary converges, obtaining the final spectral region of interest.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the segmentation and optimization method based on multi-band spectral image structure as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the segmentation and optimization method based on multi-band spectral image structure as described in any one of claims 1-5.

Images