Reconstruction method and system for full-view-field hyperspectral reflectivity

Abstract

The invention discloses a reconstruction method and system for full-field-of-view hyperspectral reflectivity. The method comprises the following steps of: firstly, acquiring radiance data of a target scene through hyperspectral imaging, and imaging diffuse reflection standard plates arranged at a plurality of positions under the same illumination and observation geometric conditions to obtain corresponding radiance data of the standard plates; secondly, geometric registration and multi-view-field splicing are carried out on the target scene and the standard board image, and reference radiance distribution covering the range of the target scene is constructed; on the basis, a continuous reference radiance field covering a full view field is reconstructed by utilizing a band-by-band surface fitting technology, and finally, a conversion function is established by combining with the nominal reflectivity of a standard plate, so that pixel-by-pixel reflectivity calculation and output are realized. The system comprises a hyperspectral imaging unit, an illumination and geometric control unit, a standard board laying unit and a data processing unit. The method is suitable for hyperspectral reflectivity consistency reconstruction of a large-scale target scene.

Description

Technical Field

[0001] This invention relates to the field of optical measurement technology, specifically to the field of hyperspectral reflectance reconstruction, and more particularly to a reconstruction method and system for hyperspectral reflectance across the entire field of view. Background Technology

[0002] The core of quantitative applications of remote sensing hyperspectral imaging lies in obtaining the true value of the spectral reflectance of a target surface. In controlled indoor environments (such as laboratory observations, darkroom tests, or simulated scene measurements), artificial light sources and specific observation devices are typically used to perform hyperspectral imaging of complex scenes. The fundamental purpose is to obtain the target reflectance spectral curve after excluding environmental variations and systematic errors, providing a reliable true value reference for subsequent material identification, parameter inversion, and model verification. In addition, in practical projects such as cultural heritage protection and restoration, there is a significant need for detailed spectral recording of large-scale, immovable targets (such as temple murals, stone carvings, and painted components), which often necessitates the use of artificial light sources for large-scale hyperspectral imaging. These objects are large in scale, complex in shape, and sensitive to lighting conditions. Constructing an ideal, uniform lighting environment on-site is usually difficult, leading to more pronounced uneven light intensity distribution and spatial response differences, further increasing the difficulty of quantitative reflectance reconstruction.

[0003] In existing technologies, reflectance calibration methods based on diffuse reflectance standards (DRS) are widely used. This method typically acquires images of the diffuse reflectance standard under illumination and observation geometry conditions consistent with the scene being measured, and establishes a radiance-to-reflectance conversion relationship based on the nominal reflectance of the DRS. Various improvement ideas have been proposed for this type of method, such as: using multiple standard standards with different nominal reflectances to expand or optimize the linear calibration interval and improve calibration accuracy; adjusting the acquisition method of dark fields or reference data in specific object measurements to reduce systematic errors; introducing geometric correction factors for spherical or non-planar targets to compensate for deviations caused by shape and observation geometry; and introducing the concept of a "dynamic standard standard," considering that the same image may be under different illumination conditions at different spatial locations, thereby improving the reliability of calibration under non-uniform illumination conditions. While these methods improve reflectance acquisition to some extent under specific conditions or within a local area, their applicability often depends on the premise that the reference standard has sufficient coverage within the field of view or that the illumination distribution is relatively uniform.

[0004] In large-field-of-view indoor experimental scenarios and hyperspectral imaging applications of large-scale immovable targets, the target scene size is usually significantly larger than the diffuse reflection standard plate size, and the imaging field of view coverage is large. Affected by factors such as the uniformity and layout of the light source, the geometrical variation of illumination incident, and scene occlusion, the illumination intensity and system response within the field of view often exhibit spatial non-uniform distribution. Especially in the field measurement of large-area targets such as temple murals, due to spatial constraints, limited lighting fixture placement and incident angle, as well as wall undulations and local occlusion, it is difficult to achieve uniform illumination, and the non-uniformity is even more significant. In this context, traditional calibration methods based on single-shot diffuse reflection standard plates are prone to the following problems: First, due to the size limitation of the diffuse reflection standard plate, it can usually only cover a local area of ​​the scene in a single imaging, making it difficult to obtain reference data covering the entire field of view, and thus unable to construct a unified full-image reference benchmark. Second, when there is spatial non-uniformity in illumination and system response, if local diffuse reflection standard plate data is still used to uniformly correct the entire image, it will introduce system biases that vary with spatial location, causing inconsistencies in reflectivity reconstruction in different areas, with more prominent errors in areas with significant brightness gradients, field of view edges, or local hot spots. Third, when different experimental batches or system settings change (such as adjustments to the target scene or diffuse reflection standard plate position), it is difficult to maintain consistency in the acquisition location and representativeness of the local reference, causing the reference benchmark to fluctuate with changes in experimental conditions, thereby reducing the comparability between different batches of data and weakening the consistency and reproducibility of repeated measurements. For the reasons mentioned above, there is an urgent need for a reflectivity reconstruction method that can obtain reference information covering the entire imaging range without requiring the diffuse reflection standard plate to cover the entire field of view. This method can achieve stable reconstruction and consistent correction of the reflectivity of the entire image band by band and pixel by pixel, and reduce reflectivity-related errors caused by non-uniform illumination and differences in system spatial response. Summary of the Invention

[0005] The purpose of this invention is to provide a reconstruction method and system for full-field hyperspectral reflectance, addressing the needs of large-field hyperspectral imaging tasks in indoor / semi-enclosed spaces and hyperspectral imaging of large-scale immovable targets. Under conditions where the target size is larger than the diffuse reflection standard plate, it is difficult to obtain full-field reference information in a single imaging, and the illumination and system response space is not uniform, the invention achieves consistent correction and stable reflectance reconstruction of the entire image band by band and pixel by pixel, improving the consistency of results and the comparability and reproducibility of multiple batches of data.

[0006] This invention is achieved by the following technical solution:

[0007] A first aspect of the present invention provides a reconstruction method for hyperspectral reflectance across the entire field of view, comprising the following steps:

[0008] Step 1. Use a hyperspectral imager to image the target scene, acquire raw image data, and perform dark field correction and radiometric calibration on the raw image data to obtain the radiance image data of the target scene.

[0009] Step 2. Under the same lighting and observation geometry as in Step 1, the diffuse reflection standard plate is sequentially placed at multiple locations covered by the target scene and imaged separately to obtain the original image data of the multi-field diffuse reflection standard plate. Dark field correction and radiometric calibration are then performed on the original image data to obtain the radiance image data of the diffuse reflection standard plate. The multiple locations are distributed across the entire range of the target scene, with a number of no less than 5, including at least the central area and four corners of the target scene.

[0010] Step 3. Perform geometric registration on the target scene radiance image data described in Step 1 and the diffuse reflection standard plate radiance image data described in Step 2 to obtain the target scene radiance image data and the standard plate radiance image data for each target scene and standard plate under the same reference coordinate system:

[0011] 1) Detect corner points or extract preset geometric features from the radiance image data of the target scene and the radiance image data of each diffuse reflection standard plate to obtain a set of feature points;

[0012] 2) Determine the correspondence between the feature point sets and solve for the coordinate transformation parameters to obtain the coordinate transformation relationship;

[0013] 3) Use the coordinate transformation relationship to unify the coordinates of the target scene radiance image data and the radiance image data of each diffuse reflection standard plate, so that they are under the same reference coordinate system, and output the target scene radiance image data under the unified coordinate system.

[0014] Step 4. Perform multi-field stitching on the radiance image data of each diffuse reflection standard plate obtained in Step 3 after coordinate unification to obtain a multi-position stitched radiance distribution covering the target scene. Map the radiance image data of each diffuse reflection standard plate under the same reference coordinate system obtained in Step 3 to the same global radiance image data according to their corresponding coordinate positions to obtain a diffuse reflection standard plate stitched radiance distribution covering the entire target scene.

[0015] Step 5. Perform band-by-band radiance surface fitting on the multi-position spliced ​​radiance distribution of the diffuse reflection standard plate described in Step 4 to reconstruct a reference radiance field covering the target scene area:

[0016] 1) Select at least one representative band from the radiance distribution of the diffuse reflection standard plate splicing, perform spatial morphology analysis, and extract morphological features to characterize the spatial distribution of illumination;

[0017] 2) Determine the radiance surface fitting model based on the morphological characteristics, and use the model to fit the radiance distribution of the representative band to obtain the spatial distribution form of the reference radiance field;

[0018] 3) Fit the radiance distribution of each band according to the wavelength to obtain a reference radiance field that varies with wavelength and covers the target scene range.

[0019] Step 6. Based on the target scene radiance image data obtained in Step 3, the reference radiance field reconstructed in Step 5, and the nominal reflectance of the diffuse reflection standard plate, calculate the radiance ratio of the target scene radiance image data and the reference radiance field at that wavelength and pixel position for any wavelength and any pixel position; then scale the ratio with the nominal reflectance of the diffuse reflection standard plate at that wavelength to obtain the corresponding reflectance value, thereby outputting the band-by-band pixel-by-pixel reflectance reconstruction result covering the entire image.

[0020] A second aspect of the present invention provides a reconstruction system for full-field hyperspectral reflectance, comprising:

[0021] The hyperspectral imaging module is used to image the target scene and the diffuse reflection standard plate to acquire raw image data;

[0022] The illumination and geometry control module is used to provide and maintain consistent illumination and observation geometry conditions during the imaging of the target scene and the imaging of the diffuse standard plate.

[0023] The diffuse reflection standard plate deployment module is used to sequentially deploy diffuse reflection standard plates at multiple locations covered by the target scene.

[0024] The data processing module is used to perform dark field correction and radiometric calibration on the original image data to obtain target scene radiance image data and diffuse reflection standard plate radiance image data; to register the target scene radiance image data and the diffuse reflection standard plate radiance image data to achieve coordinate unification; to perform multi-field stitching on the radiance image data of each standard plate to obtain a diffuse reflection standard plate stitched radiance distribution covering the entire target scene; to perform radiance surface fitting on the stitched radiance distribution to reconstruct the reference radiance field; to determine the band-by-band, pixel-by-pixel radiance to reflectance conversion relationship based on the reference radiance field and the nominal reflectance of the diffuse reflection standard plate, and to apply the conversion relationship to the target scene radiance image data to output the reflectance reconstruction result covering the entire image.

[0025] The beneficial effects achieved by this invention are as follows:

[0026] (1) Solved the problem of missing full-field reference caused by limited coverage of diffuse reflection standard plate: In view of the pain point that the target scene is large in scale but the size of diffuse reflection standard plate is limited and a single imaging can only provide local reflectivity, this invention can form a reference benchmark covering the entire target scene range by expanding the diffuse reflection standard plate to the full field of view. It no longer relies on the "single local white board" to approximate the whole field, fundamentally alleviating the problem of unusable or unstable full-image reflectivity caused by insufficient reference coverage.

[0027] (2) Solving the problem of position-related errors caused by non-uniform lighting and system response: In view of the pain point that indoor light sources are difficult to achieve uniform lighting under large field of view conditions, resulting in systematic deviations in reflectance at different spatial locations, this invention performs continuous modeling of the full field of view reference and establishes a band-by-band and pixel-by-pixel transformation relationship, so that the reflectance solution can adapt to the lighting differences at different locations, significantly suppress the error diffusion in the brightness gradient area, edge area and local hot spot area, and achieve a more consistent and reliable reconstruction of the reflectance of the whole image.

[0028] (3) Improve cross-batch comparability and repeatability to meet the needs of long-term comparative evaluation: In response to the pain point that reflectance results are prone to fluctuation and difficult to compare horizontally when repeated experiments or condition fine-tuning, the full field of view reference benchmark and conversion relationship formed by the present invention has stronger stability and reproducibility, making the reflectance results of different batches and different setting conditions more quantitatively comparable and consistent, thus making it more suitable for long-term, repeated and comparative evaluation applications in indoor remote sensing physical simulation scenarios. Attached Figure Description

[0029] Figure 1 It describes the reconstruction method flow for full-field hyperspectral reflectance;

[0030] Figure 2 This is a schematic diagram showing the placement of the diffuse reflection standard plate;

[0031] Figure 3 It is a reconstruction system for depicting hyperspectral reflectance across the entire field of view. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to embodiments. It should be understood that the embodiments described are only for explaining the invention and are not intended to limit the scope of protection of the invention; various modifications or substitutions made by those skilled in the art without departing from the spirit of the invention should fall within the scope of protection of the invention.

[0033] The first aspect of the present invention provides a reconstruction method for hyperspectral reflectance across the entire field of view, comprising the following steps:

[0034] Step 1. Use a hyperspectral imager to image the target scene, acquire raw image data, and perform dark field correction and radiometric calibration on the raw image data to obtain the radiance image data of the target scene;

[0035] 1) In the indoor optical remote sensing physical simulation system, set the lighting state (such as sunlight simulation, skylight simulation or a combination thereof) and observation geometry (sensor height, pitch angle, scan speed / integration time, etc.), and keep this state unchanged in subsequent steps.

[0036] 2) Acquire raw hyperspectral data cube of the target scene ;

[0037] 3) Collect dark field data Dark field correction is performed on the original data:

[0038]

[0039] 4) Perform radiometric calibration on the data after dark calibration: ,in and To obtain the gain and bias that vary with the band obtained from calibration, output the target scene radiance image data. .

[0040] Step 2. Under the same lighting and observation geometry as in Step 1, the diffuse reflection standard plate is sequentially placed at multiple locations covered by the target scene and imaged separately to obtain the original image data of the multi-field diffuse reflection standard plate. Dark field correction and radiometric calibration are then performed on the original image data to obtain the radiance image data of the diffuse reflection standard plate. The multiple locations are distributed across the entire target scene, with a number of no less than 5, including at least the central area and four corners of the target scene.

[0041] 1) Keep the illumination and observation geometry unchanged in step 1 to ensure that the standard plate sampling and the target scene imaging are comparable;

[0042] 2) such as Figure 2 As shown, select 5 positions: the center position and the four corner positions (top left, top right, bottom left, bottom right).

[0043] 3) Deploy diffuse reflection standard plates at each location and image them to obtain the raw data of the standard plates.

[0044] 4) Perform dark field correction and radiometric calibration on the raw data of each standard plate in the same manner as in step 1, and output the diffuse reflection standard plate radiance image data at each location.

[0045] 5) The diffuse reflection standard plate provides the nominal reflectance spectrum. .

[0046] Step 3. Perform geometric registration on the target scene radiance image data described in Step 1 and the diffuse reflection standard plate radiance image data described in Step 2 to obtain the target scene radiance image data and the standard plate radiance image data under the same reference coordinate system.

[0047] 1) Select a reference coordinate system: Position the target scene radiance image... The spatial coordinates are used as the reference coordinate system;

[0048] 2) Detect corner points or extract preset geometric features from the radiance image data of the target scene and the radiance image data of each diffuse reflection standard plate to obtain a set of feature points. and ;

[0049] 3) Determine the correspondence between the feature point sets and solve for the coordinate transformation parameters to obtain the coordinate transformation relationship; establish the correspondence and solve for the transformation parameters. in For the first Standard plate image coordinates Coordinates are reference coordinates;

[0050] 4) Using the aforementioned coordinate transformation relationship, the radiance image data of each diffuse reflection standard plate is unified with the target scene radiance image data as a reference, so that they are under the same reference coordinate system, and the radiance image data of each diffuse reflection standard plate under the unified coordinate system is output. .

[0051] Step 4. Perform multi-field stitching on the coordinate-unified diffuse reflection standard plate radiance image data obtained in Step 3 to obtain a multi-location stitched radiance distribution covering the target scene. Combine the diffuse reflection standard plate radiance image data obtained in Step 3, which are under the same reference coordinate system. By mapping their corresponding coordinate positions to the same global radiance image data, a diffuse standard panel mosaic radiance distribution covering the entire target scene is obtained.

[0052] Step 5. Analyze the radiance distribution of the multi-position splicing of the diffuse reflection standard plate described in Step 4. Perform band-by-band radiance surface fitting to reconstruct a reference radiance field covering the target scene. ;

[0053] 1) Select at least one representative band from the radiance distribution of the diffuse reflection standard plate splicing, perform spatial morphology analysis, and extract morphological features to characterize the spatial distribution of illumination;

[0054] 2) Based on the morphological characteristics, a two-dimensional Gaussian fitting model is selected for fitting. Thus, the spatial distribution of the reference radiance field is obtained;

[0055] 3) Fit the radiance distribution of each band according to the wavelength to obtain a reference radiance field that varies with wavelength and covers the target scene range.

[0056] Step 6. Based on the target scene radiance image data obtained in Step 3, the reference radiance field reconstructed in Step 5, and the nominal reflectance of the diffuse reflection standard plate, determine the band-by-band, pixel-by-pixel radiance to reflectance conversion relationship. The output is a reflectance reconstruction result covering the entire image.

[0057] A second aspect of the present invention provides a reconstruction system for full-field hyperspectral reflectance, such as Figure 3 As shown, it includes:

[0058] The hyperspectral imaging module is used to image the target scene and the diffuse reflection standard plate to acquire raw image data;

[0059] The illumination and geometry control module is used to provide and maintain consistent illumination and observation geometry conditions during the imaging of the target scene and the imaging of the diffuse standard plate.

[0060] The diffuse reflection standard plate deployment module is used to sequentially deploy diffuse reflection standard plates at multiple locations covered by the target scene.

[0061] The data processing module performs dark field correction and radiometric calibration on the original image data to obtain target scene radiance image data and diffuse reflection standard plate radiance image data; registers the target scene radiance image data and the diffuse reflection standard plate radiance image data to achieve coordinate unification; performs multi-field stitching on the radiance image data of each standard plate to obtain a diffuse reflection standard plate stitched radiance distribution covering the entire target scene; performs radiance surface fitting on the stitched radiance distribution to reconstruct a reference radiance field; determines the band-by-band, pixel-by-pixel radiance to reflectance conversion relationship based on the reference radiance field and the nominal reflectance of the diffuse reflection standard plate, and applies the conversion relationship to the target scene radiance image data to output a reflectance reconstruction result covering the entire image.

Claims

1. A reconstruction method for full-field hyperspectral reflectance, characterized in that, Includes the following steps: Step S100: Use a hyperspectral imager to image the target scene, acquire raw image data, and perform dark field correction and radiometric calibration on the raw image data to obtain the radiance image data of the target scene. Step S200: Under the same lighting and observation geometry as in step S100, the diffuse reflection standard plate is sequentially placed at multiple locations covered by the target scene and imaged separately to obtain the original image data of the multi-field diffuse reflection standard plate. Dark field correction and radiometric calibration are then performed on the original image data to obtain the radiance image data of the diffuse reflection standard plate. Step S300: Geometric registration is performed on the target scene radiance image data described in step S100 and the diffuse reflection standard plate radiance image data described in step S200 to obtain target scene radiance image data and standard plate radiance image data under the same reference coordinate system. Step S400: Perform multi-field stitching on the radiance image data of each diffuse reflection standard plate that has been coordinate unified as described in step S300 to obtain the multi-position stitched radiance distribution of the diffuse reflection standard plate covering the target scene range. Step S500: Perform band-by-band radiance surface fitting on the multi-position spliced ​​radiance distribution of the diffuse reflection standard plate described in step S400 to reconstruct a reference radiance field covering the target scene range; Step S600: Based on the target scene radiance image data obtained in step S300, the reference radiance field reconstructed in step S500, and the nominal reflectance of the diffuse reflection standard plate, establish a radiance-reflectance conversion function to perform band-by-band and pixel-by-pixel calculations, and output the reflectance reconstruction result covering the entire image.

2. The reconstruction method for full-field hyperspectral reflectance according to claim 1, characterized in that: The distribution of multiple locations in step S200 on the target scene covers the entire range of the target scene, with no fewer than 5 locations, including at least the central area and the four corners of the target scene.

3. The reconstruction method for full-field hyperspectral reflectance according to claim 1, characterized in that: Step S300, which involves geometric registration of the target scene radiance image data and the radiance image data of each diffuse reflection standard plate, includes: identifying corresponding corner points or preset geometric features in the target scene radiance image data and the radiance image data of each diffuse reflection standard plate, determining the coordinate transformation relationship, and placing the target scene radiance image data and the radiance image data of each diffuse reflection standard plate in the same reference coordinate system.

4. The reconstruction method for full-field hyperspectral reflectance according to claim 1, characterized in that: The multi-field stitching in step S400 includes: mapping the diffuse reflection standard plate radiance image data obtained in step S300 under the same reference coordinate system to the same global radiance image data according to their corresponding coordinate positions, so as to obtain a diffuse reflection standard plate stitched radiance distribution covering the entire target scene.

5. The reconstruction method for full-field hyperspectral reflectance according to claim 1, characterized in that: Step S500, which involves performing band-by-band radiance surface fitting on the multi-position spliced ​​radiance distribution of the diffuse reflection standard plate to reconstruct a reference radiance field covering the target scene, includes the following steps: Step S510: Select at least one representative band from the radiance distribution of the diffuse reflection standard plate splicing, perform spatial morphology analysis, and extract morphological features to characterize the spatial distribution of illumination. Step S520: Determine the radiance surface fitting model based on the morphological characteristics, and use the model to fit the radiance distribution of the representative band to obtain the spatial distribution form of the reference radiance field. Step S530 Fits the radiance distribution of each band according to the wavelength to obtain a reference radiance field that varies with wavelength and covers the target scene range.

6. The reconstruction method for full-field hyperspectral reflectance according to claim 1, characterized in that... In step S600, the band-by-band pixel radiance-reflectance conversion function is generated and calculated according to the following rules: For any wavelength and any pixel position, the ratio of the radiance image data of the target scene to the radiance of the reference radiance field at that wavelength and pixel position is calculated; then, the ratio is scaled proportionally with the nominal reflectance of the diffuse reflection standard plate at that wavelength to obtain the corresponding reflectance value, thereby outputting the band-by-band pixel reflectance reconstruction result covering the entire image.

7. A reconstruction system for full-field hyperspectral reflectance, the system executing the reconstruction method for full-field hyperspectral reflectance as described in claim 1, characterized in that, include: The hyperspectral imaging module is used to image the target scene and the diffuse reflection standard plate to acquire raw image data; The illumination and geometry control module is used to provide and maintain consistent illumination and observation geometry conditions during the imaging of the target scene and the imaging of the diffuse standard plate. The diffuse reflection standard plate deployment module is used to sequentially deploy diffuse reflection standard plates at multiple locations covered by the target scene. The data processing module is used to perform dark field correction and radiometric calibration on the original image data to obtain target scene radiance image data and diffuse reflection standard plate radiance image data; to register the target scene radiance image data and the diffuse reflection standard plate radiance image data to achieve coordinate unification; to perform multi-field stitching on the radiance image data of each standard plate to obtain a diffuse reflection standard plate stitched radiance distribution covering the entire target scene; to perform radiance surface fitting on the stitched radiance distribution to reconstruct the reference radiance field; to determine the band-by-band, pixel-by-pixel radiance to reflectance conversion relationship based on the reference radiance field and the nominal reflectance of the diffuse reflection standard plate, and to apply the conversion relationship to the target scene radiance image data to output the reflectance reconstruction result covering the entire image.

Images