49 results about "Spectral reconstruction" patented technology

The application discloses a kind of light field regulation detection method and device based on computing spectral reconstruction, belong to spectral analysis technical field.The method is by introducing medium-metal composite nanostructure, integrates metal nanoisland array and silicon nitride nanocolumn array on quartz substrate, realizes electric field enhancement using the local surface plasmon resonance effect excited by metal nanoisland array, simultaneously realizes wide spectral range phase coding by medium nanocolumn array;Super surface chip is integrated with single point detector, combines pre-calibrated super surface sensing matrix A and enhancement factor matrix E, and reconstructs spectrum using compressed sensing or Tikhonov regularization algorithm.The application keeps the advantages of miniaturization and no moving parts, realizes directional high sensitivity detection to weak signal, and signal-to-noise ratio is improved.

This invention discloses an underwater target reflectance inversion method based on an underwater transmission model, comprising: acquiring the original digital values ​​of the underwater target collected by a hyperspectral camera; combining the built-in radiometric calibration parameters of the hyperspectral camera; converting the original digital values ​​into radiance based on the sensor pre-calibration response characteristics; physically modeling the water absorption effect, scattering effect, and path radiation effect during underwater light propagation to determine the linear mapping relationship between radiance and target reflectance; constructing a training radiative transmission model using a non-negative linear regression model; using a calibrated target with known reflectance as a sample to train the radiative transmission model between the original digital values ​​and the target reflectance, and solving for the mapping parameters between the two; and based on the original digital values ​​and mapping parameters of the underwater target to be reconstructed, using the trained radiative transmission model to invert and obtain the true reflectance of the target, thus completing the spectral reconstruction.

This invention discloses an adaptive ambient light sensing method based on a multi-source broadband photoelectric sensor, relating to the fields of broadband photoelectric detection and intelligent environmental sensing technology. The invention includes the following steps: S1: Acquiring ambient light signals through a multi-source broadband photoelectric sensor array, wherein the array includes a visible light channel sensor, an ultraviolet light channel sensor, and an infrared light channel sensor, to obtain the original light intensity-voltage sequence V. raw (t); S2: Perform non-uniformity correction and noise filtering on the original light intensity voltage sequence to obtain the corrected spectral response matrix X. This invention integrates a multi-source sensor array of UV, VIS, and NIR, combined with a spectral reconstruction algorithm, to achieve accurate sensing over a wide spectral range of 380nm-1100nm, significantly improving the accuracy of color temperature and color rendering calculations under complex light sources; employing multi-exposure fusion and bio-inspired gain control algorithms, the system's dynamic range is extended to over 140dB, and convergence can be completed within 10ms during sudden changes in illumination, eliminating the hysteresis of traditional sensors.

This invention proposes a method and system for modeling and analyzing facial physiological parameters based on the spectrum of real skin color. First, the spectral reflectance of local facial skin is measured by capturing facial images and extracting image data from the measurement locations. Second, the spectral reflectance of the facial skin region is obtained by combining the captured and measured skin data with a spectral reconstruction algorithm. Subsequently, a two-layer optical model including the epidermis and dermis is constructed, and a power-law scattering model is introduced to describe tissue scattering behavior. Based on this, a positive mapping from physiological parameters to reflectance spectra is established using the theory of semi-transparent turbid media combined with Fresnel reflection correction at the air-skin and epidermal-dermal interfaces. Finally, a weighted optimization objective function with physiological constraints is designed for the measured spectrum, and an algorithm for boundary-constrained optimization problems is used to simultaneously invert 10 key physiological parameters, achieving high-precision, non-invasive, and quantitative assessment of skin pigmentation, blood oxygenation status, and nutritional levels.

The application discloses a kind of tunable diode absorption spectrum wide-range detection saturated spectrum processing method, it is related to spectral gas detection technical field.The tunable diode absorption spectrum wide-range detection saturated spectrum processing method, by the original absorption spectrum signal is carried out denoising and baseline correction, obtain effective spectrum signal, and extract multiple spectral feature parameters to calculate saturated quantization value, according to the quantization value, using spectral reconstruction algorithm reconstructs complete absorption spectral line profile, and the weight of absorption spectrum and wavelength modulation spectrum is calculated by S function, in combination with ambient temperature, pressure and compensation model, preliminary concentration value is corrected, and accurate final gas concentration is output.The application is by constructing saturated quantization evaluation system, stratified adaptive spectral reconstruction strategy, dual-mode detection fusion mechanism and working condition compensation model, without additional hardware, realize wide-range high-precision gas detection, eliminate environmental interference and saturation influence.

The application provides a multi-scale hyperspectral image reconstruction method fusing an adaptive guidance mechanism, which is used for realizing efficient and high-fidelity hyperspectral reconstruction and adapting to multi-scene application requirements. Specifically, the method constructs a spatial-spectral dual-branch interaction structure for decoupling modeling, and combines spectral relative position information to enhance relationship expression, thereby solving the problem of insufficient spatial-spectral coupling modeling. Further, the application proposes a spectral adaptive sparse mechanism combined with a mask, which adaptively selects and focuses on key spectral information, while suppressing redundant spectral information and improving feature discriminability, thereby adapting to spectral extraction requirements in different scenes. In addition, aiming at the problem that excessive smoothing in the spatial dimension leads to loss of detail information, the method designs a structure-guided feature refinement module, which enhances the recovery ability of edge and texture details through structure prior, thereby meeting the reconstruction requirements of high-precision scenes. Experimental results show that the method is superior to existing methods on multiple public data sets, and significantly improves spatial structure fidelity and spectral consistency, which is more suitable for actual application scene requirements.

This application relates to the field of medical image processing technology, specifically to a hemoglobin emphasis device and method for endoscopes. The device includes a light source unit; an imaging unit for acquiring reflected light signals from a target area under multispectral illumination and outputting a raw Bayer format image or an RGB color image; and an image processing unit for receiving and processing the output Bayer format image or RGB color image, comprising at least: a spectral reconstruction module for calculating and reconstructing a single-band image signal using a spectral reconstruction model; a hemoglobin concentration calculation module configured to calculate a relative hemoglobin concentration based on the image signal; and an image enhancement module configured to perform pixel-level gain adjustment on the raw color image or the processed image, convert it to a set color space for contrast mapping, and finally output a hemoglobin-emphasized image. The image output by this invention can significantly improve the contrast between blood vessels and surrounding tissues.

The application provides a hyperspectral super-resolution reconstruction method and system fusing phase priori and degradation perception, the method comprising: acquiring an observed image and a spectral response function matrix, and constructing an optimization objective function comprising a data fidelity term and a space-spectrum joint priori term; using a semi-quadratic splitting algorithm to decouple the objective function into three constraint terms of physical fidelity, spectral low rank and depth priori; using a Fourier phase-amplitude mixed processing model, a truncated singular value decomposition model and a double-branch proximal mapping network model to optimize and solve the three constraint terms respectively, to obtain and update a first auxiliary variable, a second auxiliary variable and a target hyperspectral image; through cyclic iteration until the iteration number reaches a number threshold, the target hyperspectral image output in the last iteration cycle is taken as a high-resolution hyperspectral reconstruction image. The application effectively improves the spatial texture recovery capability and spectral fidelity of the image.

This application proposes a near-field three-dimensional imaging method for sparse MIMO arrays based on efficient spectrum reconstruction. The method includes: establishing a two-dimensional MIMO array imaging model, assuming that the receiving antennas of the two-dimensional MIMO array are fully arranged, and analyzing the imaging process of a single transmitter; further assuming that the receiving antennas of the two-dimensional MIMO array are sparsely arranged along the X-axis, and determining the actual recoverable spatial scattering signal based on the imaging process of a single transmitter; performing a Fourier transform on the actual recoverable spatial scattering signal first along the Y-axis and then along the X-axis to obtain a uniform wavenumber spectrum; performing a two-dimensional inverse Fourier transform on the uniform wavenumber spectrum to obtain the target spatial scattering signal, and then converting it to the frequency domain; extracting the component corresponding to the frequency of the target distance in the frequency domain, and performing phase compensation to recover the sub-image corresponding to a single transmitter; and finally accumulating the sub-images corresponding to all transmitters to obtain a complete three-dimensional imaging result.

This invention discloses a snapshot-type hyperspectral three-dimensional skin imaging method and device, belonging to the field of biomedical imaging technology. The method includes: constructing a snapshot-type spectral-light field coupled imaging system containing a light source and an acquisition model; achieving full-domain skin illumination with broadband illumination; synchronously coupling and acquiring spectral and three-dimensional spatial information through the synergistic action of a lens group and a hyperspectral sensor to generate spectral coupled light field data; establishing a correlation dataset between the spectral coupled light field image and the hyperspectral light field image; training a spectral reconstruction network and solving the data; reconstructing the hyperspectral light field image and obtaining multi-view images through macro-pixel sampling; using a light field super-resolution network to improve image resolution; selecting the optimal skin penetration band image for stereo matching to obtain a target band depth map; and fusing the high-resolution center-view images from each channel with the target band depth map to complete the generation of a hyperspectral three-dimensional image. This invention achieves single-exposure hyperspectral three-dimensional imaging of the superficial skin layer.

This invention belongs to the field of optical technology and spectral analysis technology, and particularly relates to a bandpass sampling infrared spectrometer based on a horizontal roof mirror and a mosaic grating. Along the beam propagation direction, it sequentially includes a cylindrical beam shaping system, a dispersion and compensation integrated system, a static interferometry system, a post-beam shrinking system, and an area array detector. The dispersion and compensation integrated system includes a differential blaze angle mosaic grating and a horizontal roof mirror. The static interferometry system is located at the output path of the differential blaze angle mosaic grating. The post-beam shrinking system is located at the output end of the static interferometry system, and the area array detector is located at the image plane of the post-beam shrinking system. This invention also discloses a staged spectral reconstruction method. This scheme achieves miniaturization of the mid-infrared spectrometer, high-efficiency detection across the entire spectrum, and effective suppression of spectral aliasing under fixed sampling.

This invention provides a spectral baseline correction method based on sparse Bayesian learning, consisting of four core modules: a data normalization and downsampling module, a sparse baseline weighted fitting module, an adaptive residual quantization and weight allocation module, and a baseline smoothing and constraint module. The baseline extracted through Bayesian inference combined with mode coupling priors exhibits high physical continuity and feature peak integrity, effectively improving the quality of spectral reconstruction and reducing the risk of local distortion and peak truncation. The dynamic weight allocation technique using minimum error entropy effectively suppresses large error outliers, enhancing the algorithm's adaptability to non-Gaussian noise and providing more accurate data support for quantitative analysis. This method can output high-quality, distortion-free spectral baselines that conform to real physical characteristics, while also adapting to noise interference in complex environments and demonstrating good numerical stability.

This invention discloses a hyperspectral image classification method based on multi-scale self-supervised contrastive learning. The method comprises a network structured as follows: a contrastive learning feature extraction module suppresses unimportant features in image patches and filters out useful features; when the number of labels is small, a self-supervised mode is used to learn features; a deep and shallow multi-scale feature learning module constructs hierarchical scales to extract multi-scale information from hyperspectral images; a spatial-spectral reconstruction module constructs spectral and spatial branches, using separation, filtering, and reconstruction to combine deep and shallow features, reducing model parameters and complexity, and enhancing feature representation capabilities; the network output is then average-pooled and fed into a classifier to obtain the result. This invention acquires rich spatial-spectral features through the multi-scale feature learning module, spatial-spectral reconstruction module, and contrastive learning feature extraction module, broadening the application scenarios for images with few labels, accelerating network convergence, and thus improving classification accuracy.

This invention proposes a system for inverting forest canopy physiological parameters using canopy photon re-collision probability. The system includes: a digital hemispherical photographic image processing module, which extracts vegetation regions from captured images and generates a binary mask image; a canopy photon re-collision probability acquisition module, which calculates the canopy photon re-collision probability using the canopy directional porosity distribution on the binary mask image; a canopy spectral reconstruction module based on spectral invariance theory, which combines leaf-scale spectral information and reconstructs the canopy-scale spectral response according to spectral invariance theory, establishing a physical mapping relationship from canopy structural parameters to the canopy spectrum; and a canopy physiological parameter inversion module, which reconstructs the canopy spectrum based on the physical mapping relationship and obtains forest canopy physiological parameters using a canopy radiative transfer inversion model. This invention achieves low-cost, rapid, and stable estimation of canopy photon re-collision probability, providing reliable input for the physical model inversion of forest canopy physiological parameters.

The application discloses a kind of soil organic carbon and spectral coupling generation model method and system, belong to soil remote sensing monitoring technical field.The method includes five core steps of data preprocessing, dry soil spectrum fitting, SOC mapping model construction, morphological parameter generation, spectral reconstruction and bidirectional verification.Brightness-shape-humidity (BSM) theoretical framework separates spectral brightness and shape effect, fuses random forest algorithm and double-stage hybrid optimization model, and constructs the bidirectional reversible mapping system between spectral morphological parameter and SOC content.The application breaks through the limitation of traditional technology one-way inversion, can be based on soil spectrum high-precision inversion SOC content, and can also generate high-fidelity soil spectrum according to target SOC content, has physical explainability and data-driven flexibility, solves the technical problems of cross-regional spectral deviation and lack of explainability, and provides reliable tool for soil carbon cycle remote sensing monitoring.

An image reconstruction method, apparatus, and electronic device based on metasurface imaging are disclosed, relating to the fields of optical imaging and image processing; capable of improving image quality in complex environments. The image reconstruction method based on metasurface imaging includes: acquiring a multispectral polarization image using a metasurface imaging device; inputting the multispectral polarization image into a spectral reconstruction module to output a reconstructed hyperspectral image; inputting the hyperspectral image into a polarization reconstruction module, whereby the polarization reconstruction module employs a polarization total variation reconstruction sensing algorithm to obtain a polarization image of the target polarization direction.

This invention discloses a high-resolution dual-energy spectral CT reconstruction method based on prior information, comprising the following steps: Step 1, a high-resolution dual-energy spectral CT scanning model; Step 2, an iterative dual-energy spectral reconstruction algorithm based on BPF; Step 3, an iterative CT reconstruction algorithm based on prior information; Step 4, numerical experiments and result analysis. This high-resolution dual-energy spectral CT reconstruction method based on prior information designs a hybrid dual-detector CT scanning system, studies the combination of the material differentiation advantages of dual-energy spectral CT with the high-resolution advantages of microscopic CT, and proposes an iterative CT reconstruction algorithm based on prior information. This algorithm can perform high-resolution energy spectral reconstruction on line-to-line cards. First, the energy spectral projection data is reconstructed to obtain a low-resolution base image. Then, the base image is used as prior information and optimized and combined with the high-resolution projection data to finally obtain a high-resolution base image, thereby avoiding the missed detection of minute fractures.

This invention provides a single-stage hyperspectral image dehazing and reconstruction method based on joint polarization spectral priors, comprising the following steps: Step 1, haze imaging modeling; Step 2, acquiring joint polarization spectral compressed observation data; Step 3, constructing a single-stage polarization spectral reconstruction neural network; Step 4, network training and optimization. The integrated reconstruction architecture proposed in this invention effectively overcomes the limitations of traditional "two-stage" methods. By directly establishing an end-to-end mapping from compressed observations to clear hyperspectral images, it avoids the cumbersome computational steps involved in dehazing followed by reconstruction (or vice versa), significantly improving inference efficiency.

This invention discloses a computational spectral measurement system based on spatial and angular coupling modulation. The system introduces the electromagnetic radiation to be measured through a point light source, which is then subjected to secondary modulation by a spectroscopic filtering unit and a phase interferometry unit with different positional combinations. The spectroscopic filtering unit restricts the wavelength passband in the spatial dimension, while the phase interferometry unit generates interference fringes that vary highly with wavelength by utilizing the coupling relationship between the incident angle and phase delay. A composite coded image containing the spatial wavelength characteristics and the angular phase characteristics is acquired by a detection unit. Then, by establishing a forward physical model and employing a computational reconstruction algorithm, ultra-high resolution spectral reconstruction is achieved across the entire wavelength range. This system effectively solves the problem of aliasing in high-order interference through dual modulation at the physical level, and has advantages such as simple structure, extremely high resolution, and strong band versatility.

This invention belongs to the field of marine acoustic detection and remote sensing technology, and discloses a method and system for inverting the abundance of deep-sea mixed sediments based on bottom echo data. The method involves preprocessing the bottom echo waveforms of multibeam bathymetry data to obtain a bottom echo sequence; extracting features from the bottom echo sequence waveforms and concatenating these features to the corresponding bottom echo waveforms to form a new waveform sequence; fusing the enhanced waveform sequences to construct an observation matrix Y; decomposing the observation matrix Y into an endmember waveform matrix H′ and an abundance distribution matrix W′, and introducing a minimum volume constraint during the decomposition process; adding a spectral reconstruction error term based on dynamic time warping to the loss function; iteratively optimizing H′ and W′ using gradient descent or multiplicative update rules until convergence; and obtaining the endmember waveform H′, which represents the acoustic response waveform of typical sediment types, and the abundance matrix W′, which reflects the mixing ratio of each endmember at different sampling points, demonstrating high classification accuracy.

This invention belongs to the field of micro-nano optics technology, specifically disclosing a method for generating modulation units in a computational spectral sensing chip. First, a dielectric metasurface structure based on a magnetic dipole resonance mechanism is fabricated on a substrate. By adjusting geometric parameters, ultra-narrowband, differentiated transmittance curves within multiple operating wavelength bands are obtained. Subsequently, a protective layer is fabricated to form a preliminary structure. Finally, by calculating the correlation coefficients between the curves, effective units with low correlation are screened and determined to constitute the spectral modulation units. This invention utilizes the sharp spectral responses with varying peak positions, linewidths, and line shapes generated by magnetic dipole resonance to ensure low correlation between filter matrix units, thereby improving the spectral reconstruction robustness and resolution accuracy of the computational spectral sensing chip.

The invention relates to the field of spectrum reconstruction, in particular to a spectrum reconstruction method for a broadband filtering modulation spectrum detection system, which comprises the following steps of: converting a hyperspectral data set into a training data set according to parameters of the broadband filtering modulation spectrum detection system; building a spectrum reconstruction network model based on deep learning and carrying out training; and pixel splitting point-by-point scanning reconstruction is carried out on a reconstruction target, a hyperspectral image is constructed according to original arrangement, space-spectrum joint optimization is carried out, and finally high-precision spectral image reconstruction is realized. According to the method, the demand of a spectrum reconstruction process on a data set scale is reduced through a pixel splitting point-by-point scanning reconstruction strategy, a Transform architecture is introduced into a reconstruction network, the feature information extraction capability of the network is improved, the strong nonlinear characterization capability of a KAN architecture is combined, the spectrum curve reconstruction precision is improved, and spatial spectrum combined processing is combined, so that the spectrum reconstruction efficiency is improved. The error is further reduced, and high-quality reconstruction of the spectral image is realized.

The application provides an intelligent imaging system based on electrochromic and multispectral cooperation, and belongs to the technical field of multispectral imaging, and is used for solving the problems that the filter wheel is large in volume, slow in switching, and difficult to balance miniaturization, real-time performance and high spectral accuracy in the related art. The system comprises a miniaturized switchable filter wheel, an image sensor and a controller. The filter wheel is integrated with an electrochromic narrow-band filter station and a multi-band synchronous acquisition filter area. The controller constructs an implicit spectral radiation field model and updates parameters thereof online through variational inference, outputs a posterior mean and a posterior variance, calculates expected information gain of each candidate acquisition action based on the posterior variance, and executes an optimal action, so as to realize adaptive cooperation of narrow-band high-precision acquisition and wide-band rapid acquisition. The system significantly improves compactness, real-time performance and spectral reconstruction quality of multispectral imaging.

The application provides a kind of based on perovskite tunable filter array computing reconstruction type spectral detector, belongs to spectral detector technical field, including perovskite filter array, photosensitive camera and the host computer with spectral computing reconstruction module;The light spectrum to be measured is incident to perovskite filter array and is selectively absorbed, and then directly to photosensitive camera, converts the transmitted light signal into electrical signal, and the output gray image is acquired and processed by the host computer, and the spectral reconstruction is completed by the spectral computing reconstruction module.The application utilizes the specific absorption characteristics of perovskite materials with different components to different wavelengths of light, and uses perovskite filter array to realize the function of spectrometer.Combining the system response matrix and the system response result vector, the spectral reconstruction is realized by the spectral computing reconstruction algorithm.The application can greatly reduce the size of the spectrometer to micron level, and reduce the difficulty and cost of preparation, and can change the perovskite filter array for different detection needs, and improve the compatibility of spectral detector.

The application discloses a kind of computing spectrometer and spectral reconstruction method based on random multimode waveguide grating.It includes silicon substrate layer, silicon dioxide buried oxygen layer, lithium niobate waveguide layer, silicon dioxide coating layer and interdigital electrode arranged in turn from bottom to top.The lithium niobate waveguide layer is formed with mode multiplexer, multimode waveguide spiral and multimode waveguide grating, the mode multiplexer and multimode waveguide spiral are communicated, the multimode waveguide spiral includes at least one multimode straight waveguide, and the multimode straight waveguide is provided with a plurality of non-periodically distributed multimode waveguide gratings, each multimode waveguide grating unit adopts Bragg grating structure, and each multimode waveguide grating unit includes two groups of grating tooth structures located on both sides of the multimode straight waveguide.The application can simplify the electrical control logic, and the unordered resonance of different modes in the spiral can provide high spectral reconstruction resolution, and lithium niobate electro-optic modulation can realize fast and low-power measurement channel scanning.

This invention discloses a method and system for identifying the authenticity of Pinellia ternata and its processed products based on hyperspectral imaging and transfer learning networks. The method involves acquiring hyperspectral image data of Pinellia ternata and its processed products, performing black-and-white correction on the hyperspectral image data, and extracting the spectral sequence from the source domain data. An S-STN transfer learning framework is constructed and trained using the spectral sequence from the source domain data. The S-STN transfer learning framework includes a spectral encoder and a capsule classifier. A masked spectral reconstruction strategy is used to perform self-supervised pre-training of the spectral encoder, learning device-independent general spectral representations. The parameters of the pre-trained spectral encoder are frozen, and the capsule classifier is trained in a supervised manner on the source domain task to obtain a preliminary identification model. This preliminary identification model is then transferred to the target domain task, the spectral encoder is frozen, and only the capsule classifier is fine-tuned to output the authenticity identification result of the target domain sample.

This invention belongs to the field of magnetic resonance imaging technology, and particularly relates to a multi-voxel spectral reconstruction method and system with ROI adaptive capability. The method includes acquiring a structural image of the scanned object; performing multi-voxel spectral scanning on the scanned object to acquire K-space spectral data; based on the location information of the region of interest (ROI) and interference regions, combined with the point spread function of the spectral scanning system; performing target reconstruction processing on the K-space spectral data according to the optimal voxel localization parameters to generate reconstructed spectral data; and extracting spectral information corresponding to the ROI from the reconstructed spectral data. This invention, through a strategy of multiple adaptive reconstructions in a single scan, can independently calculate the optimal localization parameters and perform targeted reconstruction for each ROI within the field of view based on the same set of K-space data. This allows different anatomical structures to obtain the maximum volume proportion in their dedicated reconstructed voxels without compromising the localization effect between different regions.

The present application relates to the technical field of atmospheric pollution monitoring, and discloses a plant unorganized emission monitoring and early warning platform based on infrared spectrum technology. In order to solve the problems of analysis distortion caused by spectral baseline distortion in traditional monitoring and lack of adaptive tracing ability of fixed early warning threshold. The platform edge server cascades a spectrum reconstruction deconvolution module, a three-dimensional inversion risk module and an adaptive early warning module; a one-dimensional convolution network is used to jointly reconstruct high-fidelity spectrum and separate multi-component gas by using an adaptive least square method; a three-dimensional leakage source strength is inverted based on a spatial sparse prior Markov chain Monte Carlo engine; and early warning is dynamically determined in combination with a double cumulative sum algorithm. The present application is used for gas monitoring in complex industrial areas, and can accurately locate and adaptively prevent false alarms.

This invention discloses a hyperspectral image unmixing method based on a three-dimensional deformation collaborative unmixing network. The method includes: constructing a hyperspectral image unmixing network comprising a three-dimensional joint feature encoding module, a physically constrained abundance mapping module, and a linear spectral reconstruction module; the encoding end rearranges the input hyperspectral image into a five-dimensional tensor with the spectral dimension as the depth dimension, and performs multi-scale spatial-spectral feature extraction through multiple three-dimensional deformation collaborative feature blocks; after deep feature extraction, the physically constrained abundance mapping module generates an abundance map satisfying the abundance non-negativity constraint and the abundance sum of one constraint; and the linear spectral reconstruction module reconstructs the hyperspectral image and outputs the endmember spectrum. This invention employs an end-to-end unsupervised training method, which can improve the endmember extraction accuracy and abundance estimation accuracy of hyperspectral images in complex scenes, low signal-to-noise ratio scenes, and irregular target regions.