35 results about "Spectral variation" patented technology

The invention discloses a poplar degenerated forest stand accurate restoration method, and relates to the technical field of poplar degenerated forest stand restoration, and the method comprises the steps: obtaining a multi-temporal satellite remote sensing image and an unmanned plane multi-spectral image of a target area; in combination with soil moisture content, rhizosphere respiration rate and micrometeorological data collected by ground Internet of Things sensing nodes arranged in a forest stand, geographic coordinate registration and time sequence alignment are performed on the data to form a multi-source observation data set; the method comprises the following steps: extracting vegetation indexes, short-wave infrared reflectivity, soil moisture and rhizosphere physiological parameters based on a multi-source observation data set, constructing a degradation diagnosis index fusing spectral change, moisture stress and root activity, calculating the degradation degree of each pixel, and generating a degradation level spatial distribution diagram; the area with the degradation degree reaching a set threshold value is determined as a to-be-repaired area; and selecting representative poplar individuals in the to-be-repaired area, measuring the photosynthetic ability of the individuals, and calculating the individual competition intensity in combination with the three-dimensional point cloud data.

The invention relates to the technical field of image classification, and discloses a hyperspectral image domain generalization classification method, system and device and a medium. The method comprises the following steps: acquiring a plurality of hyperspectral images of different ground feature types; obtaining a mean value and a variance of each hyperspectral image in a channel dimension, performing random disruption, and determining a spectral variation parameter of the hyperspectral image according to the mean value and the variance before and after random disruption so as to generate a corresponding spectral variation image; separating the center and the background of each hyperspectral image to obtain respective center image and background image, randomly disorganizing the center image and the background image, and generating a corresponding spatial variation image by combining the disorganized center image and background image of each hyperspectral image; a hyperspectral image classification model is obtained by training an enhanced hyperspectral image obtained by fusing a plurality of corresponding spectral variation images and spatial variation images, so that ground feature classification is performed on a to-be-classified hyperspectral image, and the domain generalization performance of the hyperspectral image classification model is improved.

The invention discloses a remote sensing image semantic change detection method, which belongs to the field of remote sensing image semantic change detection, and comprises the following steps: carrying out double-temporal feature extraction by using a residual network, sending the double-temporal features into a constructed difference enhancement module to model difference information, and obtaining a change weight of the double-temporal features; the change weight and the double-temporal feature are sent to a feature interaction module, interaction among the three branches is completed, the interacted double-temporal feature and a differential feature rich in semantic information are obtained, a binary change graph is obtained from a segmentation head to a semantic segmentation graph and through a change detection head, the semantic segmentation graph is superposed on the obtained binary change graph, and the binary change graph is obtained. And finally generating a semantic change graph. Compared with other existing semantic change detection methods, the method provided by the invention is more excellent, can reduce leak detection conditions under different spectrums similar in semantics and error detection conditions under the same spectrums change semantics, and also has a good effect in detail and edge detection.

A method, system, device and medium for detecting hyperspectral changes by integrating lightweight spatial-spectral features across time periods, the method comprising: preprocessing hyperspectral image data and dividing them into training sets and test sets; constructing a feature extraction network; constructing a classification network; using the training set, training the feature extraction network and the classification network with a gradient descent algorithm, calculating the accuracy of the training set in each iteration, and using the weight of the first generation network model with the highest accuracy on the training set as the final detection model weight to obtain a trained model; inputting the test set into the trained model for testing, obtaining a detection result, and outputting a predicted label map of the hyperspectral image data; the system, device and medium are used to implement the method; the present invention designs a model based on spatial-spectral feature extraction and cross-temporal feature fusion from the two aspects of refined feature extraction and cross-temporal feature fusion, and simplifies the model using a lightweight method, thereby realizing a lightweight hyperspectral image change detection method with good performance.

The invention provides a laboratory digital management method and system, electronic equipment and a storage medium, and relates to the technical field of laboratory automation and spectral analysis. The method comprises the following steps: acquiring a multi-wavelength spectral image sequence in an experimental reaction container in a non-contact manner, associating the multi-wavelength spectral image sequence with a final performance index, and performing differential enhancement to generate a spectral change feature map; performing differential amplification on the spectral electric signal to suppress noise and enhance an effective signal, dividing a region of interest according to a container structure, calculating average spectral intensity, and generating a spectral response curve; the method comprises the following steps: extracting first-order and second-order derivative curves, identifying an intensity change inflection point and a slope abrupt change interval, screening out a key optical change mode in combination with a final performance index, finally establishing a mapping relation between the key optical change mode and the performance index, and constructing a digital model for monitoring a reaction state and predicting final performance. The reaction process can be monitored and the experiment result can be predicted through the spectrum change characteristics, so that the intelligence and accuracy of laboratory management are improved.

The invention belongs to the technical field of optical measurement, and discloses an implementation device and method of a single-photon spectrograph, and the device comprises a spectral wavelength scanning module which is used for collecting a wide-spectrum optical signal, carrying out the wave division processing, and calculating the photon counting change rate between adjacent wavelengths; wavelength stepping is dynamically adjusted according to the photon counting change rate, self-adaptive scanning of spectrum change is achieved, and a single-spectrum optical signal is output; the self-focusing detection module is used for converting the single-spectrum optical signal into a photon event through a preset array channel; in a preset time window, calculating the variance and the Shannon entropy of each photon event, and constructing a local uncertainty index; performing spectral domain self-focusing and resampling according to the local uncertainty index to obtain photon counting data; high-precision, high-stability and high-efficiency spectrum light signal detection and visual analysis are realized.

The invention discloses a position-aware hyperspectral image classification method based on a subgraph convolutional network. The method comprises the following steps: carrying out data preprocessing on hyperspectral image data; establishing a hyperspectral image classification model by using the subgraph convolutional neural network and fusing pixel position codes; training a hyperspectral image classification model based on the preprocessed hyperspectral image data to obtain the hyperspectral image classification model; and classifying the pixel categories of the whole hyperspectral image by using the hyperspectral image classification model. The invention provides a sub-graph convolutional neural network method for efficiently fusing Laplacian-based symbol-invariant position coding, so that the diversity of pixel features can be enhanced, the classification deviation caused by spectral variation is reduced, and the classification precision is remarkably improved.

The application provides a fault identification method and system based on multi-element spectral mineral anomaly analysis, relates to the fields of geology, spectroscopy, petrology and engineering geology, and a specific scheme comprises the following steps: in-situ testing of tunnel surrounding rock is carried out by using a visible light-near infrared, mid-infrared and laser Raman spectrometer, and mixed spectral variation characteristics are extracted; based on the mixed spectral variation characteristics, the mineral composition of the tunnel surrounding rock is identified according to the spectral identification standard of the abnormal mineral combination of the fault zone, quantitative inversion of the minerals is carried out, and the mineral content is obtained; according to the mineral anomaly mode of the fault zone, mineral anomaly analysis is carried out on the mineral composition, and the mineral anomaly type is obtained; based on the differential change of the mineral content and the mineral anomaly type, mineralogical anomaly characteristic analysis is carried out, and the fault identification result is output; the application replaces the experience knowledge of traditional geologists with quantitative information of multi-element spectral mineral anomaly, and further improves the accuracy and efficiency of fault identification in tunnels and underground engineering.

This invention discloses a method for establishing a spectral characteristic database of trees affected by wood nematode, involving the interdisciplinary fields of forestry disease monitoring, spectral analysis, and database technology. By capturing subtle spectral changes in pine needles in the red and near-infrared bands, this invention can identify signs of infection before obvious symptoms appear on the tree's surface. Through a multi-step, multi-index comprehensive analysis method, it effectively eliminates interference from other ground features such as forest land and bare soil. The established dedicated spectral model and indices are highly targeted, achieving higher classification and identification accuracy than using a single general index, significantly reducing the false positive rate. The use of UAV hyperspectral remote sensing technology overcomes the shortcomings of manual inspection, such as limited coverage, numerous blind spots, and poor timeliness, enabling rapid and comprehensive monitoring of large-scale pine forest resources. It can not only determine whether pine trees are infected but also quantitatively assess the severity of the disease through chlorophyll content estimation models and spectral characteristic differences.

The invention provides a hyperspectral and multispectral image fusion method and system. The method comprises the following steps: acquiring hyperspectral and multispectral data of the same spatial region; performing up-sampling on the low-resolution hyperspectral image by adopting bilinear interpolation; one-dimensional wavelet transform is utilized to extract reflection peak and spectrum change characteristics in a spectrum dimension, two-dimensional wavelet transform is utilized to extract texture and edge information in a space dimension, and pixel-by-pixel wavelet basis adaptive selection is realized through Softmax gating; multi-granularity feature modeling is realized for the spectral branches by adopting grouped multi-scale convolution, and a self-adaptive weighting mechanism is formed for the spatial branches through global average pooling, a convolution layer and Sigmoid activation so as to enhance spatial details and compensate spatial feature loss by utilizing MSI; a cross-scale two-way attention mechanism is introduced, and long-range dependence modeling and information interaction of spectrum and spatial characteristics are achieved; and a high-resolution hyperspectral image is generated through a feature aggregation and reconstruction module.

The invention relates to a hyperspectral calculation imaging technology, and aims to solve the problems that in an actual remote sensing observation environment, a single imaging mode is limited by a physical sensor, high space resolution and hyperspectral resolution are difficult to obtain at the same time, and geometry and spectrum complementarity exists among multi-source images. The invention provides a hyperspectral fusion imaging method based on topology perception and a differential homeomorphic mechanism. According to the method, differential homeomorphic transformation is introduced to construct spatial degradation priori, a dual learning framework is adopted to jointly optimize high-resolution generation and bidirectional reconstruction tasks, meanwhile, a topology sensing Transform module is utilized to explicitly model a spectrum-space double-domain manifold structure, complementary correlation among multi-source data is fully mined, and the spectrum-space double-domain manifold structure is constructed. The spatial structure consistency and the spectral fidelity of the fused image are effectively improved, the problem of fusion quality reduction caused by geometric deformation and spectral variation in a complex ground feature scene is significantly improved, and a more reliable data support is provided for applications such as remote sensing monitoring and environmental perception.

The invention belongs to the technical field of cultural heritage protection, and discloses an organic residue identification method based on small sample class incremental learning, which comprises the following steps: preprocessing archaeological organic residue infrared spectrum data to eliminate baseline drift and peak position dislocation; constructing a protoGate-based small sample classification module, introducing chemical prior constraint to optimize feature stability, and designing dynamic weighted prototype measurement to enhance classification capability; building a lightweight incremental learning module, and combining a lightweight memory bank and cross-stage knowledge distillation to realize knowledge migration and forgetting resistance; and designing a data enhancement strategy based on an aging mechanism to generate a simulation spectrum, and finally realizing accurate identification of the organic residues in small-sample and multi-category scenes. The problems that archaeological organic residue samples are scarce, spectral variation is complex and new categories are difficult to access are solved, and efficient technical support is provided for cultural relic research and protection.

The application discloses a kind of position-aware subgraph convolution network-based hyperspectral image classification methods, comprising: data preprocessing is carried out to hyperspectral image data;Utilize subgraph convolution neural network and fuse pixel position coding to establish hyperspectral image classification model;Based on the hyperspectral image data after pre-processing, the hyperspectral image classification model is trained, and the hyperspectral image classification model is obtained;The overall hyperspectral image pixel class is classified using the hyperspectral image classification model.The application proposes a subgraph convolution neural network method that efficiently fuses laplace-based sign-invariant position coding, which can enhance pixel feature diversity, reduce classification bias caused by spectral variation, and significantly improve classification accuracy.

The invention discloses a hyperspectral anomaly detection method based on double-domain multi-scale feature reconstruction, which comprises the following steps: firstly, in order to relieve the influence of abnormal pixels and spectral variation on background recovery, reducing the abnormal pixels through a space-spectral mask strategy, thereby avoiding interference on subsequent feature extraction; then, due to the complexity of the background, the invention designs a multi-scale double-domain feature interaction module which is used for extracting multi-scale features of the background in different domains. Specifically, multi-scale spatial features are extracted in a spatial domain through a multi-scale convolution block, and frequency domain features are extracted in a frequency domain through multi-scale wavelet convolution. In addition, the features of different domains are fully fused through feature interaction, and the background modeling capability of the network is improved. And finally, performing anomaly detection on the residual error of the reconstructed background and the input image through the mahalanobis distance. According to the method, features can be effectively extracted for targets with different shapes and multi-scale targets, so that a relatively good detection effect is achieved.

The invention discloses a planetary surface typical target high-precision positioning identification system and method, and the method comprises the steps: a data enhancement module receives hyperspectral data, carries out the data enhancement of the hyperspectral data, and builds a contrast learning frame based on an enhanced sample; extracting a background sample and a prior sample in the hyperspectral data, enhancing the sample and extracting a space-spectrum combined feature, combining the space-spectrum combined feature with a contrast learning framework to calculate a contrast loss, and performing sample reconstruction on the background sample; the weight reconstruction module calculates a reconstruction error of an original hyperspectrum and a prior sample, calculates a deviation vector according to the reconstruction error, converts the deviation vector into a reconstruction weight, and dynamically adjusts the comparison loss by using the reconstruction weight; the system trained by the modules performs feature extraction and calculates feature similarity to generate a target detection image; according to the method, high-precision and robust typical target positioning and recognition can be realized in a planetary surface environment which is lack of labeled samples and has spectral change.

The invention relates to the technical field of intelligent environment monitoring, and discloses a flood season water quality inversion monitoring method based on an unmanned aerial vehicle and passive sampling, and the method comprises the steps: uniformly distributing water sampling points in a research area before the flood season, recording the position information of each sampling point, and after the water sampling of the sampling points is completed, carrying out the inversion monitoring of the water quality; the method comprises the following steps: sampling a water sample in a research area, sending the water sample to a laboratory to complete chemical analysis of water quality parameters, acquiring a hyperspectral remote sensing image of the research area by using a hyperspectral imager carried by an unmanned aerial vehicle while collecting the water sample at each sampling point, preprocessing the hyperspectral remote sensing image, extracting the waveband reflectivity of the position in the image according to the latitude and longitude of the manual sampling point, different forms of spectral changes are carried out on the original spectral reflectivity to form a water quality inversion data set, water quality parameters of a river are simulated through a water quality parameter inversion model, spatial distribution characteristics of water quality parameter concentration are displayed, and a high water quality parameter concentration point is selected to arrange a passive monitoring device.

The invention provides a spectral data processing method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the initial spectral data of a spectrometer, determining the target spectral values of a plurality of detection pixels according to the initial spectral data, determining the spectral change rate between a plurality of adjacent detection pixels according to each target spectral value, and determining the spectral change rate of the detection pixels according to the spectral change rate. The spectral interpolation of each to-be-interpolated position is determined based on two target spectral values corresponding to each adjacent detection pixel and a plurality of spectral change rates corresponding to each target pixel, the to-be-interpolated position is used for representing the position between the adjacent detection pixels, and the target pixels comprise a plurality of detection pixels corresponding to the to-be-interpolated position before and after; combining the spectral interpolation of each to-be-interpolated position with the target spectral value of the corresponding adjacent detection pixel to obtain target spectral data; according to the method, the spectral resolution is improved, the difficulty of interpolation calculation is reduced, and the real-time performance of the interpolation calculation is also improved.

The invention relates to the technical field of coal detection and intelligent control, in particular to a coal moisture prediction method based on time sequence hyperspectral change, which comprises the following steps of: acquiring a coal sample surface image, identifying particle edge gray change, marking a particle distribution area, judging a particle size distribution trend according to center-of-gravity and center-line offset, and dividing a spectrum observation window. The method comprises the following steps: acquiring a spectral reflection sequence, establishing a time number, carrying out cross additional recording on channel data, marking a moisture absorption interval, extracting an emissivity change curve, identifying a sudden change section, extending the sudden change section to reconstruct an absorption peak track, extracting a reflectivity decline curve section, dividing a moisture release trend section and generating a moisture release curve group. According to the invention, through coal sample image analysis and spectral data processing, construction of a high-precision moisture prediction model, extraction of particle distribution and spectral change trend, and combination of absorption peak reconstruction and moisture release analysis, accurate monitoring is realized, and through combination of multi-region spectral observation and a reflectivity sequence, environmental interference is reduced, and accurate prediction of coal moisture is realized.

The present invention relates to the technical field of crude oil detection, and in particular to a method and system for tracing the origin of crude oil based on big data. The method provided by the present invention comprises obtaining spectral variation characteristics, calculating spectral variation characterization parameters for the crude oil to be tested, determining the spectral variation category of the crude oil to be tested, determining a beam diameter and a sample scanning time when performing spectral analysis on the crude oil to be tested based on the spectral variation characterization parameters, determining spectral data acquired within a predetermined time, dividing the spectral data in a time domain dimension, comparing the obtained spectral data segments, determining whether the spectral analysis meets standards based on similarity between the spectral data segments, adjusting the scanning speed of a spectrometer, acquiring spectral data again and storing it, or acquiring spectral data while maintaining initial operating parameters of the spectrometer and storing it, tracing the origin of the crude oil to be tested based on a similarity comparison result between the stored spectral data and spectral data of corresponding samples from each origin, thereby improving the accuracy of tracing.

The application relates to the technical field of data analysis, in particular to an inland river basin ecological environment quality monitoring method based on remote sensing data analysis; different partitions are obtained according to the spectral difference characteristics and distribution distance characteristics between pixels in the ground reflectivity image; the characteristic indexes of the pixels under different preset remote sensing indexes are acquired; the spectral response sensitivity of the preset remote sensing index is obtained according to the discrete characteristics of the characteristic indexes of different pixels in the partition and the discrete characteristics of the band reflectivity of the preset remote sensing index; the anti-interference stability of the preset remote sensing index is obtained according to the interference characteristics of the spectral change of the pixels in the partition on the characteristic indexes; and the adaptation degree of the preset remote sensing index in the partition is obtained according to the spectral response sensitivity and the anti-interference stability. The final remote sensing index of the partition is selected according to the different adaptation degrees, and the ecological environment quality is monitored, so that the accuracy of the ecological environment quality monitoring is improved.

The present invention belongs to the field of image processing technology. In order to solve the problem of inaccurate detection in existing hyperspectral detection, a cervical cancer detection system, method, equipment and medium based on hyperspectral imaging are proposed. The long-range dependence of hyperspectral cervical images in the spatial dimension is captured through spatial state modeling, so as to capture the contextual features of the lesion structure and improve the modeling ability of the integrity of the lesion structure. In addition, the spectral band response of the lesion area may span multiple adjacent bands. GCN mapping can model the nonlinear connection between the bands and enhance the perception of weak spectral variations. The importance of spatial and spectral features to anomaly detection varies in different samples. The gated fusion mechanism is used to dynamically balance the spatial spectrum contribution, realize automatic adjustment in different scenarios, improve robustness, and reduce overfitting of specific feature channels.

This invention discloses a Hapke hyperspectral autoencoder unmixing method for spectral variations. Addressing the problem of spectral variations in hyperspectral images, this method combines the Hapke physical model of spectral imaging with a deep network to construct a unmixing network architecture for the Hapke model. This effectively reduces the impact of spectral variations on endmember extraction, resulting in more accurate endmembers and abundances. The core of this method is to construct a deep network to learn the spectral variation parameters in the Hapke model. Based on the autoencoder unmixing network architecture, this invention extracts abundance features through convolutional autoencoder layers, while the weights of the decoding layer are endmember features. To reduce the impact of spectral variations on endmembers and abundances, the Hapke physical model is fused with the network, and a parameter estimation network is designed to learn the spectral variation parameters, which are then used as illumination-topography attention applied to abundances. Simultaneously, these parameters are substituted into the Hapke model, transforming endmember features into single-scattering albedo. Combined with reflectance and albedo spatial losses, iterative optimization of endmember abundances is achieved. Numerical experiments show that the proposed unmixing method has good unmixing performance and significantly improves the accuracy of endmember estimation.

The invention relates to the technical field of spectral detection, in particular to a multi-spectral detection method for bacteria in the early stage of surgical wound infection, which comprises the following steps: scanning a target wound at a preset time interval by using a multi-spectral imaging technology to obtain spectral data, and then acquiring a multi-spectral detection result based on continuously monitored data; the spectrum signal availability is analyzed, and the comprehensive possibility that the pixels of the image target area belong to the early bacterial infection state is obtained by combining the characteristics of wound secretion and the spectrum change trend caused by the change of the bacterial chemical component proportion along with time, so that the early identification and evaluation of the infection state are realized, the false positive and false negative are reduced, and the detection accuracy is improved. The sensitivity and the specificity of early bacteria detection are remarkably improved, and accurate support is provided for clinical early intervention and debridement.

The application belongs to the technical field of optical measurement, and discloses an implementation device and method of a single-photon spectrometer, which comprises a spectral wavelength scanning module, collects a wide-spectrum light signal, carries out wave separation processing, and calculates a photon counting rate between adjacent wavelengths; wavelength steps are dynamically adjusted according to the photon counting rate, adaptive scanning of spectral changes is realized, and a single-spectrum light signal is output; a self-focusing detection module converts the single-spectrum light signal into a photon event through a preset array channel; in a preset time window, the variance and Shannon entropy of each photon event are calculated, and a local uncertainty index is constructed; spectral domain self-focusing and resampling are carried out according to the local uncertainty index, and photon counting data are obtained; high-precision, high-stability and high-efficiency spectral light signal detection and visual analysis are realized.

The invention discloses a hyperspectral and laser radar data open set identification method based on a biequiangular constraint hybrid expert model, and belongs to the technical field of remote sensing image processing. Aiming at the problems of loose decision boundary and weak unknown class identification capability caused by intra-class spectral variation and multi-modal difference in fused data in the existing method, the invention provides a DEC-MoE framework. The framework adopts a hybrid expert structure, and dynamically routes and fuses features to a special classifier through a gating network to adapt to intra-class changes; the method comprises the following steps of: firstly, selecting an expert classifier, innovatively introducing biequiangular constraint, and cooperatively optimizing the weight of the expert classifier and a known class prototype, so that a compact simplex structure with a maximum angular interval is formed in a feature space, thereby maximizing the inter-class separation degree and tightening a known class decision boundary. According to the method, the open set recognition performance of the multi-source remote sensing data can be effectively improved, known surface features are reliably classified, and unknown categories are detected.

The application discloses a terahertz spectrum molecular identification method and system based on physical constraint adversarial training and a storage medium. The method resamples the terahertz frequency domain spectrum of a substance to be identified into a one-dimensional sequence, constructs a selective state space classifier without using absolute position coding, and defines four differentiable perturbation operators, namely spectral line broadening, frequency calibration offset, baseline drift and additive noise, in the order of physical degradation chain that the intrinsic spectral variation of a sample is in the front and the measurement artifact of an instrument is in the back, the parameter range of which is constrained by the specification of the terahertz instrument, to constitute a six-dimensional physical perturbation space. During training, the perturbation parameters that increase the classification loss are searched in the normalized parameters of the space by projecting the sign gradient ascent, and the classifier is updated by using the degraded spectrum, so that the robustness and computational efficiency of long sequence terahertz spectrum molecular identification under the condition of instrument physical degradation are improved.

This application relates to the fields of remote sensing image processing and environmental monitoring technology, and discloses a remote sensing monitoring method for soil salinization in semi-arid regions based on small samples. The method extracts regional baseline endmembers and performs fully constrained linear spectral unmixing to obtain the baseline reconstruction error. This error is used to screen high-confidence unlabeled pixels to generate pseudo-labels, constructing an enhanced training sample set. An ensemble regression model capable of outputting predicted mean and variance is trained. For pixels with high predicted variance, an endmember elastic perturbation mechanism is triggered, iteratively optimizing under the condition of satisfying physical reconstruction error constraints to obtain the endmember combination and inversion results that minimize the predicted variance. Finally, the results are integrated to generate a salinity distribution map. This invention solves the problem of small-sample training through a physical-statistical coupling strategy and uses adaptive perturbation to correct spectral variations in mixed pixels, effectively improving the inversion accuracy and reliability of results in complex surface environments.

The invention relates to the field of hyperspectral image classification, and discloses a multi-hop differential Transform and pyramid gating network for Mars hyperspectral image classification, the multi-hop differential Transform and pyramid gating network comprises a multi-hop differential Transform module, the multi-hop differential Transform module is composed of neighborhood feature mapping and a Transform encoder based on multi-hop differential attention, a PoNF is used for aggregating neighborhood spectral features, and the PoNF is used for aggregating the neighborhood spectral features; the TE with multi-hop differential attention further extracts global spectral features and captures fine local spectral changes; the spatial pyramid gating enhancement module is composed of an SPGB and an MLP, a double-threshold gating strategy is further designed, and extracted spatial features are finely selected; the spectral space adaptive fusion module is used for realizing deep interaction and fusion of spectral features and spatial features through an adaptive weighting mechanism; the network of the invention further extracts global spectral features and captures local fine spectral changes through a multi-hop differential attention mechanism, thereby enhancing the ability of the model to extract discriminative spectral features from highly similar spectrums.

The present invention relates to the field of satellite-borne hyperspectral image processing and analysis technology, solving the technical problem that existing methods fail to fully consider the huge spectral and spatial resolution differences between low-resolution hyperspectral data and high-resolution multispectral data. In particular, it relates to a deep and shallow dual-branch super-resolution method for forest hyperspectral satellite imagery, comprising constructing a dual-branch network architecture, performing feature fusion reconstruction on the features output, and obtaining high-resolution hyperspectral data with low-spectral variation characteristics that characterize forest vegetation in satellite-borne hyperspectral imagery data. The present invention effectively solves the modal difference problem between low-resolution hyperspectral and high-resolution multispectral data by learning at different feature levels and scales, and through multiple feature attention mechanisms, enables the model to pay more attention to the low-spectral variation characteristics of different forest vegetation in satellite imagery, which can meet the requirements of subsequent various forest resource precision monitoring tasks.