Seed multi-mode microscopic image acquisition method and device, equipment and storage medium

Abstract

The invention relates to the field of image processing, and provides a seed multi-modal microscopic image acquisition method, device and equipment and a storage medium, and the method comprises the steps: carrying out the region segmentation of an obtained hyperspectral image of a plurality of seeds, and obtaining a single-seed region image; performing data extraction on the single seed area image to obtain spectral data of each seed; the spectral data comprises a seed identifier and a reflectivity value of each wave band; segmenting and extracting the acquired scanning images of the multiple seeds to obtain single seed scanning slices; performing three-dimensional reconstruction on the binary mask image to obtain seed three-dimensional physical structure information; the binary mask image is obtained by performing pixel classification processing on the single seed scanning slice. The method provides a basis for comprehensively and deeply analyzing chemical components and internal structures of the seeds, and improves the efficiency of seed detection and analysis.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to a method, apparatus, device, and storage medium for acquiring seed multimodal microscopic images. Background Technology

[0002] The chemical characteristics of seeds (e.g., moisture, protein and oil content) and internal structural characteristics (e.g., embryo morphology and pore distribution) are key factors that determine germination rate, stress resistance and yield.

[0003] In terms of chemical composition analysis, existing near-infrared spectroscopy (NIR) spot detection methods can usually only provide average spectral information of a single location or small area, and cannot obtain information on the spatial distribution of chemical components on or inside the seed surface; in addition, existing NIR spot scanning methods have low throughput and are difficult to adapt to the rapid screening of large batches of seeds.

[0004] In terms of visualizing internal structure, existing methods such as manual dissection combined with microscopic observation can visualize the internal structure of seeds, but their destructive nature makes them unsuitable for testing live seeds or batches of seed samples, and the single testing cycle is long and the testing efficiency is low. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for acquiring multimodal microscopic images of seeds, in order to solve the problem of low efficiency in existing methods for analyzing seed chemical composition and visualizing internal structure.

[0006] This invention provides a method for acquiring multimodal microscopic images of seeds, comprising the following steps: The hyperspectral images of multiple seeds are segmented to obtain single-seed region images. Data is extracted from the image of the individual seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​for each band; The scanned images of multiple seeds are segmented and extracted to obtain scanned slices of a single seed; The three-dimensional physical structure information of the seed is obtained by performing three-dimensional reconstruction on the binarized mask image; the binarized mask image is obtained by pixel classification processing of the scanned slice of the single seed.

[0007] According to the present invention, a method for acquiring multimodal microscopic images of seeds includes, prior to region segmentation of the acquired hyperspectral images of multiple seeds, the following steps: Multiple seeds are scanned using a detector covering the first spectral range to obtain the first spectral information; Multiple seeds are scanned using a detector covering the second spectral range to obtain second spectral information; Based on the first spectral information and the second spectral information, a hyperspectral image of multiple seeds is determined.

[0008] According to the present invention, a method for acquiring multimodal microscopic images of seeds includes performing region segmentation on the acquired hyperspectral images of multiple seeds to obtain single-seed region images, which includes: An initial mask is constructed based on the first waveband; the first waveband is within the first spectral range; Based on the initial mask and target threshold, the effective region of the hyperspectral image is determined; The effective region is filled with holes and filtered for noise to obtain a single seed region image.

[0009] According to the seed multimodal microscopic image acquisition method provided by the present invention, the step of extracting data from the image of a single seed region to obtain the spectral data of each seed includes: Extract the first reflectance of each seed in the first spectral range, and extract the second reflectance of each seed in the second spectral range; Based on the first reflectance and the second reflectance, the spectral data of each seed is determined.

[0010] According to a method for acquiring multimodal microscopic images of seeds provided by the present invention, the method further includes: The single seed scan slice is subjected to pixel classification processing using a trained deep learning model to obtain a binarized mask image.

[0011] According to the seed multimodal microscopic image acquisition method provided by the present invention, the step of performing three-dimensional reconstruction of the binarized mask image to obtain the three-dimensional physical structure information of the seed includes: Determine the target voxels in the binarized mask image; the target voxels belong to the same physical structure and are spatially continuous; the physical structure includes the seed body structure, the internal pore structure, and the seed embryo structure; Based on the target voxels, the binary mask image is reconstructed in three dimensions to obtain the seed's three-dimensional physical structure information.

[0012] The present invention also provides a seed multimodal microscopic image acquisition device, comprising the following modules: The region segmentation module is used to segment the acquired hyperspectral images of multiple seeds into regions to obtain single-seed region images. The spectral data acquisition module is used to extract data from the image of the single seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​of each band; The scanned image segmentation and extraction module is used to segment and extract scanned images of multiple seeds to obtain scanned slices of a single seed. The three-dimensional reconstruction module is used to perform three-dimensional reconstruction on the binarized mask image to obtain the three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single seed scan slice.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the seed multimodal microscopic image acquisition method as described above.

[0014] 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 seed multimodal microscopic image acquisition method as described above.

[0015] This invention provides a method, apparatus, device, and storage medium for acquiring multimodal microscopic images of seeds, constructing a system for seed chemical feature analysis and internal structure visualization. By processing hyperspectral images of multiple seeds, this invention can acquire the spectral data of each seed non-destructively, with high throughput, and automatically, thereby obtaining the spatial distribution information of the seed's chemical composition based on the spectral data. Then, the scanned images of the acquired seeds undergo a series of processes including segmentation and extraction, pixel classification, and 3D reconstruction to obtain the seed's three-dimensional physical structure information. This invention provides a foundation for comprehensive and in-depth analysis of seed chemical composition and internal structure, improving the efficiency of seed detection and analysis. Attached Figure Description

[0016] 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.

[0017] Figure 1 This is one of the flowcharts illustrating the seed multimodal microscopic image acquisition method provided by the present invention.

[0018] Figure 2 This is the second flowchart of the seed multimodal microscopic image acquisition method provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the seed spectral data acquisition process provided by the present invention.

[0020] Figure 4This is a schematic diagram of the seed three-dimensional physical structure information acquisition process provided by the present invention.

[0021] Figure 5 This is a schematic diagram of the structure of the seed multimodal microscopic image acquisition device provided by the present invention.

[0022] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0024] The following is combined with Figures 1-6 This invention describes a method, apparatus, device, and storage medium for acquiring multimodal microscopic images of seeds.

[0025] Figure 1 This is one of the flowcharts illustrating the seed multimodal microscopic image acquisition method provided by the present invention, such as... Figure 1 As shown, the method includes the following: Step 100: Perform region segmentation on the acquired hyperspectral images of multiple seeds to obtain single-seed region images; Specifically, the seed multimodal microscopic image acquisition method provided by this invention uses a specially designed porous seed mold to place seeds, enabling the simultaneous acquisition of hyperspectral images of multiple seeds in a single scan. A dual-detector hyperspectral imaging system covering a wide spectral range (400nm-2500nm) is employed to simultaneously acquire spectral information in the visible and near-infrared (400nm-1000nm) and short-wave infrared (1000nm-2500nm) spectral ranges. The specially designed seed mold can be made of black polyethylene foam; its shape can be a short cylinder with multiple circular through-holes and an external fixing tube.

[0026] Hyperspectral images of multiple seeds are acquired using a dual-detector hyperspectral imaging system. Then, single-seed regions are accurately identified and segmented using automated image processing methods, which is the single-seed region image in this embodiment.

[0027] Step 200: Extract data from the image of the single seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​for each band; Specifically, after segmenting the hyperspectral images of multiple seeds to obtain single-seed region images, data is extracted from the single-seed region images to obtain spectral data for each seed, including seed identification and reflectance values ​​for each band.

[0028] Step 300: Segment and extract the scanned images of multiple seeds to obtain scanned slices of a single seed; Specifically, seeds are placed using the aforementioned specially designed porous seed mold, and multiple seed-loaded molds are stacked one on top of the other, then fitted with an external fixing tube to form a stable scanning unit. This design achieves the effect of acquiring three-dimensional structural data of multiple corn seeds in a single scan, significantly improving the detection throughput. During the scanning process, the mold and fixing tube structure effectively fix the position of the seeds, greatly reducing motion artifacts caused by corn seed displacement and improving the image quality of Micro-CT (microcomputed tomography).

[0029] After the scanning unit acquires scanned images of multiple seeds, the scanned images of multiple seeds are segmented and extracted to obtain scanned slices of single seeds. The specific process of image preprocessing and single seed segmentation and extraction is as follows: 1. Image preprocessing and seed region identification (1) Noise reduction: Gaussian filtering is used to reduce noise in the intermediate layer slices (grayscale image) (Gaussian kernel size is 5x5 pixels, standard deviation can be 0), and noise interference is reduced by blurring operation.

[0030] (2) Threshold segmentation: The global threshold is automatically calculated using the Otsu's Method (OTSU) to binarize the image into a seed region (foreground pixel value 255) and a background (pixel value 0).

[0031] (3) Morphological optimization: The binary image is processed by opening operation (3x3 square kernel, iterated 2 times) to eliminate small noise; the 8-connected region labeling algorithm is used to identify all candidate regions.

[0032] 2. Seed selection and spatial sorting (1) Region filtering: Based on physical size constraints, connected regions with an area ≥ 500 pixels are retained.

[0033] (2) Geometric sorting: Calculate the geometric center point of all seed centroids.

[0034] 3. Dynamic Generation and Expansion of Region of Interest (ROI) (1) Initial ROI determination: The seed minimum bounding rectangle is obtained based on the connected component bounding box.

[0035] (2) Intelligent expansion: Expand by 60 pixels in each direction; automatically constrain the expansion range to not exceed the original image size.

[0036] 4. Batch processing of full slices and standardized output (1) Layered processing: Traverse all BMP (Bitmap, a lossless bitmap image format) files, slice each layer: extract various sub-extended ROI regions; create a 500x500 pixel black canvas; center the original ROI.

[0037] Step 400: Perform three-dimensional reconstruction on the binarized mask image to obtain the three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single seed scan slice.

[0038] Specifically, the collected three-dimensional spatial data of the seed interior is used for three-dimensional reconstruction. Based on the extracted single-seed scan slices, a trained deep learning model is used for segmentation. In this embodiment, the deep learning model can use ResNet18 (a deep residual network) as the backbone network. This model performs pixel-level binarization prediction for the three physical structures: embryo, pores, and the main body of the corn seed. During the prediction stage, the model classifies the input seed scan slices pixel by pixel. Through activation function and thresholding, a single-channel binary mask output image is generated, where the target structure (foreground) pixel value is 255 and the background pixel value is 0. The prediction results are saved to a specified output directory.

[0039] This embodiment provides a seed multimodal microscopic image acquisition method, constructing a seed chemical feature analysis and internal structure visualization system. By processing the acquired hyperspectral images of multiple seeds, this invention can acquire the spectral data of each seed non-destructively, with high throughput, and automatically, thereby obtaining the spatial distribution information of the seed's chemical composition based on the spectral data. Then, the scanned images of the acquired seeds undergo a series of processes including segmentation and extraction, pixel classification, and 3D reconstruction to obtain the seed's 3D physical structure information. This invention provides a foundation for comprehensive and in-depth analysis of seed chemical composition and internal structure, improving the efficiency of seed detection and analysis.

[0040] Figure 2 This is the second flowchart illustrating the seed multimodal microscopic image acquisition method provided by the present invention, as shown below. Figure 2 As shown, the method may further include: Step 10: Scan multiple seeds using a detector covering the first spectral range to obtain the first spectral information; Step 20: Scan multiple seeds using a detector covering the second spectral range to obtain the second spectral information; Step 30: Based on the first spectral information and the second spectral information, determine the hyperspectral image of multiple seeds.

[0041] Specifically, such as Figure 3 As shown, multiple corn seed samples to be tested are sequentially placed on the surface of a customized seed mold. The seed mold can be made of black polyethylene foam, is short and cylindrical, and has multiple through-hole circular placement holes. Multiple seed-loaded molds are placed on the scanning platform of a hyperspectral imaging system. The hyperspectral image acquired by the hyperspectral imaging system includes the corn and the background. The hyperspectral imaging system includes a visible light + near-infrared detector and a short-wave infrared detector. The visible light + near-infrared detector can cover the 400nm-1000nm band (i.e., the first spectral range in this embodiment); the short-wave infrared detector can cover the 1000nm-2500nm band (i.e., the second spectral range in this embodiment).

[0042] Whiteboard calibration plays a crucial role in the processing of hyperspectral reflectance data. Through standardization, it eliminates the influence of light sources and instruments, ensuring the accuracy and consistency of the data. It can obtain high-precision and highly adaptable spectral information of seeds and represent it in the form of hyperspectral images.

[0043] This embodiment uses a dual-detector hyperspectral imaging system to simultaneously acquire spectral information of visible light and near-infrared light, as well as spectral information of short-wave infrared light.

[0044] In one embodiment, the seed multimodal microscopic image acquisition method provided by this invention may further include: Step 110: Construct an initial mask based on the first band; the first band is within the first spectral range; Step 120: Based on the initial mask and target threshold, determine the effective region of the hyperspectral image; Step 130: Fill holes and filter noise in the effective area to obtain a single seed region image.

[0045] Specifically, for the visible light + near-infrared detector (400nm-1000nm band), an initial mask is constructed based on band 118 (around 900nm, i.e., the first band in this embodiment). An adaptive thresholding algorithm (the threshold can be set to 0.15) is used to process only the effective region in the center of the hyperspectral image. After processing, morphological optimization is performed, including hole filling (to eliminate voids inside the seed) and area filtering (e.g., to remove noise regions smaller than 100 pixels). After morphological optimization, spectral feature extraction is performed. Based on the initial mask, the reflectance of each seed in the 400nm-1000nm range is extracted, and the mean spectrum of pixels in various sub-regions is calculated. Numbered spatial distribution labels for the seeds are generated. The labeling order can be sorted from left to right and top to bottom according to the centroid coordinates. The spectral data is output in a specific format, such as Comma-Separated Values ​​(CSV) format, containing the seed ID and reflectance values ​​for all bands.

[0046] For short-wave infrared detectors (1000nm-2500nm band), an initial mask is constructed based on the first band (around 1000nm). An adaptive thresholding algorithm (threshold can be set to 0.15) is used to process only the effective region in the center of the hyperspectral image. After processing, morphological optimization is performed, including hole filling and area filtering. After morphological optimization, spectral features are extracted. Based on the mask, the reflectance of each seed in the 1000nm-2500nm range is extracted, and the mean spectrum of pixels in various sub-regions is calculated. Numbered spatial distribution labels for the seeds are generated. The labeling order can be sorted from left to right and top to bottom according to the centroid coordinates. The spectral data is output in CSV format, containing seed identifiers and reflectance values ​​for all bands.

[0047] This embodiment uses an automated image processing algorithm (based on mask construction for specific bands, adaptive threshold segmentation, morphological optimization, and centroid sorting) to accurately identify and segment single seed regions, thereby obtaining single seed region images.

[0048] In one embodiment, the seed multimodal microscopic image acquisition method provided by this invention may further include: Step 210: Extract the first reflectance of each seed in the first spectral range, and extract the second reflectance of each seed in the second spectral range; Step 220: Determine the spectral data of each seed based on the first reflectance and the second reflectance.

[0049] Specifically, a visible light + near-infrared detector covering the first spectral range (400nm-1000nm band) is used to obtain the first reflectance of each seed within the first spectral range; a short-wave infrared detector covering the second spectral range (1000nm-2500nm band) is used to obtain the second reflectance of each seed within the second spectral range. The continuous spectral information of each seed, i.e., the spectral data of each seed, is determined using the first and second reflectances.

[0050] This embodiment uses a dual-detector hyperspectral imaging system covering a wide spectral range (400nm-2500nm band) to acquire spectral data for each seed.

[0051] In one embodiment, the seed multimodal microscopic image acquisition method provided by this invention may further include: Step 500: Perform pixel classification processing on the single seed scan slice using a trained deep learning model to obtain a binarized mask image.

[0052] Specifically, in Micro-CT data processing, a trained deep learning model (e.g., based on the ResNet18 backbone network) is used to perform end-to-end pixel-level classification prediction on the preprocessed and segmented single-seed CT slices (i.e., the single-seed scan slices in this embodiment) to obtain a binarized mask image.

[0053] In this embodiment, a trained deep learning model is used to perform pixel-level classification and prediction on a single seed scan slice to obtain a binarized mask image.

[0054] In one embodiment, the seed multimodal microscopic image acquisition method provided by this invention may further include: Step 410: Determine the target voxels in the binarized mask image; the target voxels belong to the same physical structure and are spatially continuous; the physical structure includes the seed body structure, the internal pore structure, and the seed embryo structure; Step 420: Perform three-dimensional reconstruction of the binary mask image based on the target voxel to obtain the seed three-dimensional physical structure information.

[0055] Specifically, such as Figure 4 As shown, 3D reconstruction is performed based on the segmented embryo structure, pore structure, and binarized slices of the maize seed body. According to the voxels belonging to the same structure and spatially continuous within the binarized slices, 3D models of each structure are reconstructed, and the volume of each 3D model (in voxels or actual physical dimensions) is automatically calculated. Simultaneously, the percentage (relative proportion) of the volume of the pores and embryo within the seed to the volume of the maize seed body is calculated, i.e., the 3D physical structure information of the seed.

[0056] This embodiment obtains the three-dimensional physical structure information of the seed by performing three-dimensional reconstruction on the binarized slices.

[0057] The seed multimodal microscopic image acquisition device provided by the present invention will be described below. The seed multimodal microscopic image acquisition device described below can be referred to in correspondence with the seed multimodal microscopic image acquisition method described above.

[0058] Please refer to Figure 5 The present invention also provides a seed multimodal microscopic image acquisition device, comprising: The region segmentation module 501 is used to segment the acquired hyperspectral image of multiple seeds into regions to obtain a single seed region image. The spectral data acquisition module 502 is used to extract data from the image of the single seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​of each band; The scanned image segmentation and extraction module 503 is used to segment and extract the scanned image of multiple seeds to obtain a single seed scanned slice; The three-dimensional reconstruction module 504 is used to perform three-dimensional reconstruction on the binarized mask image to obtain the three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single seed scan slice.

[0059] Optionally, the seed multimodal microscopic image acquisition device further includes: The first spectral information determination module is used to scan multiple seeds using a detector covering the first spectral range to obtain the first spectral information; The second spectral information determination module is used to scan multiple seeds using a detector covering the second spectral range to obtain the second spectral information; A hyperspectral image determination module for multiple seeds is used to determine a hyperspectral image of multiple seeds based on the first spectral information and the second spectral information.

[0060] Optionally, the region segmentation module includes: An initial mask construction unit is used to construct an initial mask based on a first waveband; the first waveband is within the first spectral range. An effective region determination unit is used to determine the effective region of the hyperspectral image based on the initial mask and the target threshold. A single-seed region image determination unit is used to fill holes and filter noise in the effective region to obtain a single-seed region image.

[0061] Optionally, the spectral data acquisition module includes: A reflectance extraction unit is used to extract the first reflectance of each seed in the first spectral range and to extract the second reflectance of each seed in the second spectral range. A spectral data determination unit is used to determine the spectral data of each seed based on the first reflectance and the second reflectance.

[0062] Optionally, the seed multimodal microscopic image acquisition device further includes: The pixel classification module is used to perform pixel classification processing on the single seed scan slice using a trained deep learning model to obtain a binarized mask image.

[0063] Optionally, the three-dimensional reconstruction module includes: A target voxel determination unit is used to determine the target voxels in the binary mask image; the target voxels belong to the same physical structure and are spatially continuous; the physical structure includes a seed body structure, an internal pore structure, and a seed embryo structure. The three-dimensional reconstruction unit is used to perform three-dimensional reconstruction of the binary mask image based on the target voxel to obtain the seed three-dimensional physical structure information.

[0064] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 610, a communication interface 620, a memory 630, and a communication bus 640. The processor 610, communication interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a seed multimodal microscopic image acquisition method. This method includes: segmenting the acquired hyperspectral images of multiple seeds to obtain single-seed region images; extracting data from the single-seed region images to obtain spectral data for each seed; the spectral data includes seed identification and reflectance values ​​for each band; segmenting and extracting scanned images of the acquired multiple seeds to obtain single-seed scanned slices; and performing three-dimensional reconstruction on a binarized mask image to obtain three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single-seed scanned slices.

[0065] Furthermore, the logical instructions in the aforementioned memory 630 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, in essence, 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.

[0066] On the other hand, 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 implements a seed multimodal microscopic image acquisition method provided by the above methods. The method includes: performing region segmentation on the acquired hyperspectral image of multiple seeds to obtain a single seed region image; extracting data from the single seed region image to obtain spectral data for each seed; the spectral data includes a seed identifier and reflectance values ​​for each band; segmenting and extracting scanned images of acquired multiple seeds to obtain single seed scan slices; and performing three-dimensional reconstruction on a binarized mask image to obtain three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single seed scan slice.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0068] 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.

[0069] 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 method for acquiring multimodal microscopic images of seeds, characterized in that, include: The hyperspectral images of multiple seeds are segmented to obtain single-seed region images. Data is extracted from the image of the individual seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​for each band; The scanned images of multiple seeds are segmented and extracted to obtain scanned slices of a single seed; The three-dimensional physical structure information of the seed is obtained by performing three-dimensional reconstruction on the binarized mask image; the binarized mask image is obtained by pixel classification processing of the scanned slice of the single seed.

2. The seed multimodal microscopic image acquisition method according to claim 1, characterized in that, The step of performing region segmentation on the acquired hyperspectral image of multiple seeds includes: Multiple seeds are scanned using a detector covering the first spectral range to obtain the first spectral information; Multiple seeds are scanned using a detector covering the second spectral range to obtain second spectral information; Based on the first spectral information and the second spectral information, a hyperspectral image of multiple seeds is determined.

3. The seed multimodal microscopic image acquisition method according to claim 2, characterized in that, The step of performing region segmentation on the acquired hyperspectral images of multiple seeds to obtain single-seed region images includes: An initial mask is constructed based on the first waveband; the first waveband is within the first spectral range; Based on the initial mask and target threshold, the effective region of the hyperspectral image is determined; The effective region is filled with holes and filtered for noise to obtain a single seed region image.

4. The seed multimodal microscopic image acquisition method according to claim 2, characterized in that, The step of extracting data from the image of the individual seed region to obtain the spectral data of each seed includes: Extract the first reflectance of each seed in the first spectral range, and extract the second reflectance of each seed in the second spectral range; Based on the first reflectance and the second reflectance, the spectral data of each seed is determined.

5. The method for acquiring multimodal microscopic images of seeds according to claim 1, characterized in that, The seed multimodal microscopic image acquisition method further includes: The single seed scan slice is subjected to pixel classification processing using a trained deep learning model to obtain a binarized mask image.

6. The seed multimodal microscopic image acquisition method according to claim 5, characterized in that, The step of performing three-dimensional reconstruction on the binary mask image to obtain the seed's three-dimensional physical structure information includes: Determine the target voxels in the binarized mask image; the target voxels belong to the same physical structure and are spatially continuous; the physical structure includes the seed body structure, the internal pore structure, and the seed embryo structure; Based on the target voxels, the binary mask image is reconstructed in three dimensions to obtain the seed's three-dimensional physical structure information.

7. A seed multimodal microscopic image acquisition device, characterized in that, include: The region segmentation module is used to segment the acquired hyperspectral images of multiple seeds into regions to obtain single-seed region images. The spectral data acquisition module is used to extract data from the image of the single seed region to obtain the spectral data of each seed; the spectral data includes the seed identifier and reflectance values ​​of each band; The scanned image segmentation and extraction module is used to segment and extract scanned images of multiple seeds to obtain scanned slices of a single seed. The three-dimensional reconstruction module is used to perform three-dimensional reconstruction on the binarized mask image to obtain the three-dimensional physical structure information of the seed; the binarized mask image is obtained by pixel classification processing of the single seed scan slice.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the seed multimodal microscopic image acquisition method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the seed multimodal microscopic image acquisition method as described in any one of claims 1 to 6.

Images