14 results about "Spectral compression" patented technology

The invention relates to the technical field of remote sensing image processing, in particular to a few-sample hyperspectral remote sensing image classification method and system based on multi-path evolution. The method comprises the steps of generating a spectral attention weight by utilizing spectral compression reconstruction according to acquired multi-scale deep feature representation, and performing three-dimensional convolution evolution on weighted features based on adaptive multi-path feature evolution to obtain information fusion features; performing weighted fusion on the information fusion features of the three paths through a dynamic gating mechanism; and obtaining a hyperspectral image classification result according to the fused features. According to the invention, self-adaptive evolution and cross-path selective fusion of multi-scale features are realized; the deep features subjected to multi-level feature coding are used for final classification reasoning, so that high-precision hyperspectral image classification is realized under the condition of extremely few samples.

Apparatuses and methods are disclosed for encoding and decoding mesh data. Encoding techniques are disclosed including coding a mesh into a bitstream. The coding includes generating a base mesh from the mesh, obtaining connectivity and geometry data of the base mesh. Then, subdividing the base mesh, obtaining connectivity and geometry data of the subdivided mesh. Coding proceeds by generating GFT coefficients, based on a Graph Fourier Transform (GFT), using displacement data, and then, coding into the bitstream the coefficients and connectivity data of the base mesh. Decoding techniques are disclosed including decoding the mesh from the bitstream. The decoding includes decoding from the bitstream connectivity data of the base mesh and subdividing the base mesh, obtaining connectivity data of the subdivided mesh. Decoding proceeds by decoding from the bitstream the GFT coefficients and reconstructing the mesh based on the decoded connectivity data of the subdivided mesh and the decoded coefficients.

This invention discloses a hyperspectral image classification method and related apparatus based on a dual-branch Mamba-Former network. First, the hyperspectral image data to be classified is input. Then, spectral compression and local spatial feature extraction are performed using the SS-ResNet module, and classification tokens are added to generate a token sequence. The token sequence is then input into the parallel encoding branch, MAPE Block, for feature encoding. Further, the classification tokens from the parallel encoding branch are extracted and deeply fused using Fusion-MLP to obtain the final discriminative feature representation. Finally, the fused features are input into a classifier to predict the land cover categories in the hyperspectral image. This invention effectively integrates the local inductive bias of CNNs, the global alignment of the global context branch MHSA, and the linear, efficient long-range modeling capabilities of Mamba, improving the accuracy of deep learning applied to hyperspectral image classification.

The invention discloses an interpolation filter design method based on fuel measurement signal multi-phase decomposition, and belongs to the technical field of filter digital signal processing. According to the method, the cut-off frequency and transition band characteristics of an interpolation filter are optimized through the multi-sampling-rate conversion theory and the multi-phase decomposition technology, and filtering parameters are dynamically adjusted in combination with an adaptive algorithm; the high-precision reconstruction of a fuel measurement signal is realized; the core is that a rear digital low-pass filter is designed, a frequency spectrum compression and expansion technology is used for inhibiting an aliasing effect, and calculation efficiency is improved through a multiphase decomposition structure; the method is especially suitable for an aviation fuel measurement system, can balance data redundancy and signal fidelity under different sampling rates, and significantly improves the measurement precision and the system stability.

The application discloses a large-oblique TOPS mode ground plane BP self-focusing method based on improved spectrum compression, and comprises the following steps: representing a BP image in a range frequency domain, obtaining a spectrum compression function by using a stationary point principle; obtaining a spectrum support region center of a certain point in a scene based on the spectrum compression function, and replacing radar positions of each time of the spectrum support region center of the certain point in the scene with a beam rotation center to obtain a new spectrum support region center; obtaining a final spectrum compression function of a TOPS mode based on a beam rotation center of a frequency domain mode introduced by the new spectrum support region center; converting the BP image to an azimuth frequency domain based on the final spectrum compression function of the TOPS mode to obtain an error function, and obtaining a SAR image of the TOPS mode finally completing self-focusing and sidelobe suppression based on the error function and an azimuth window function. The application greatly reduces the operation complexity of a BP algorithm processing TOPS mode data.

The application discloses a kind of training-free snapshot photometric stereo vision methods based on spectral compression imaging, comprising: (1) setting multiple different wavelength monochromatic light sources to simultaneously irradiate object from different directions, form high-dimensional spectral-angle data cube;(2) the spectral-angle data cube is modulated in space by coded aperture, along the spectrum dimension is sheared translation by using optical dispersion element, finally integrates on single-channel image sensor, generates single two-dimensional gray compressed measurement image;(3) inverse problem model is constructed, and multiple illumination image sequence is decoupled and recovered from single two-dimensional gray compressed measurement image using iterative optimization algorithm based on physical drive;(4) the surface normal map of the measured object is predicted by inputting multiple illumination image sequence into pre-trained photometric stereo vision network module.The application can realize low-cost, high data efficiency and high physical fidelity three-dimensional reconstruction by single snapshot.

This invention provides an optimal scene coordinate system establishment method for rapid temporal imaging using medium- and high-orbit SAR, relating to the field of radar imaging technology. Specifically, this invention first proposes four methods for establishing the scene coordinate system under large squint mode. By analyzing the characteristics of spectral compression performance and spectral tilt distribution under different coordinate systems, a maximizing scene coordinate system suitable for large-scene imaging is derived. This coordinate system maximizes the mapping bandwidth of a single processing operation, thereby improving imaging efficiency. Furthermore, this invention provides a method for selecting the scene coordinate system under different imaging conditions. By determining whether the scene is regular and whether the scene area is smaller than a preset size, a suitable scene coordinate system is selected to meet the imaging requirements of medium- and high-orbit SAR under large squint mode and reduce the computation of redundant data.

The application discloses a GCBP imaging method and system of a non-normal side array multichannel SAR of an ultrahigh-speed diving platform, and the method comprises the following steps: constructing a non-normal side array multichannel SAR geometric model; performing pulse compression and sub-aperture division on echo signal data; constructing a multichannel steering matrix and calculating a weight vector of reconstructed sub-aperture echo signal; constructing a sidelobe suppression window function; traversing all grid points in an imaging area, substituting the weight vector and the sidelobe suppression window function into integral operation of BP imaging, performing BP imaging processing on echo signal in the sub-aperture, and sequentially performing two-stage spectrum compression, azimuth up-sampling and spectrum decompression processing on the sub-aperture multichannel BP image after coherent accumulation; and traversing all sub-apertures, coherently fusing all sub-aperture BP images, and obtaining a non-lattice lobe high-resolution SAR image; and the multichannel steering matrix and the sidelobe suppression window function are fused in the sub-aperture BP integral operator, so that ghosting can be directly eliminated in the BP imaging process, and the anti-fuzzy performance of the SAR image is improved.

The present application relates to the technical field of remote sensing image processing, in particular to a multi-path evolution-based few-sample hyperspectral remote sensing image classification method and system. The method comprises generating spectral attention weights by using spectral compression reconstruction according to the obtained multi-scale deep feature representation, performing three-dimensional convolution evolution on the weighted features based on adaptive multi-path feature evolution to obtain information fusion features; the information fusion features of the three paths are weighted and fused through a dynamic gating mechanism; and the classification result of the hyperspectral image is obtained according to the fused features. The present application realizes adaptive evolution and cross-path selective fusion of multi-scale features; the deep features coded by the above multi-level features are used for final classification reasoning, so that high-precision hyperspectral image classification is realized under the condition of few samples.

The invention discloses a three-dimensional imaging method and device, a program product and a storage medium, and the method comprises the steps: dividing sub-apertures with equivalent phase centers at the same position in a multiple-input-multiple-output (MIMO) array into the same sub-aperture group, and obtaining a plurality of sub-aperture groups, the sub-aperture is obtained by dividing a receiving antenna and a transmitting antenna in the MIMO array in advance, and the equivalent phase center is the center of a center connecting line of each antenna in the sub-aperture; performing frequency spectrum compression on the plurality of sub-aperture groups based on the echo data acquired by the plurality of sub-aperture groups in the transmitting and receiving period, and determining a plurality of target sub-images corresponding to the plurality of sub-aperture groups; and coherently superposing the plurality of target sub-images to obtain a target three-dimensional image. According to the invention, the problem of low imaging efficiency in the prior art is solved, and the effect of efficient imaging is achieved.

To provide a fiber amplifier system with high power and narrow linewidth.SOLUTION: A fiber amplifier system 10 includes an optical source providing an optical seed beam, and an FM electro-optic modulator (EOM) 20 that frequency modulates the seed beam to broaden its spectral linewidth. The system also includes an AM EOM 24 that amplitude-modulates the seed beam to provide an amplitude modulated seed beam that is synchronous with the frequency modulated seed beam. The system also includes a nonlinear fiber amplifier 28 that receives the seed beam that is AM-modulated and FM-modulated, the amplitude modulated seed beam causes self-phase modulation in the fiber amplifier that phase-modulates the seed beam as it is being amplified by the fiber amplifier that acts to cancel the spectral linewidth broadening caused by the frequency modulation.SELECTED DRAWING: Figure 1

This invention discloses a multispectral super-resolution imaging device based on compressed sensing, belonging to the field of optical microscopy. The device comprises a light source and beam expansion system, a polarization modulation and structured illumination microscopy system, an coded spectral compression imaging system, a synchronization control system, and a data processing and reconstruction system. This invention solves the problem of the trade-off between increasing the number of spectral channels and maintaining imaging speed in existing spectral-resolution structured light super-resolution microscopy. By combining compressed sensing with single-snapshot spectral coding and dispersive imaging, structured illumination, and a two-step depth unfolding network reconstruction algorithm based on a physical model, this invention can rapidly and synchronously reconstruct super-resolution images of multiple spectral channels with only a small number of compressed images acquired, significantly improving the spectral resolution of imaging without sacrificing the imaging frame rate.

The invention discloses a lightweight speech enhancement method based on a grouped dual-path LSTM (Long Short Term Memory). The lightweight speech enhancement method comprises the following steps: downloading and preprocessing a VoiceBank + DEMAND data set required by a model; performing short-time Fourier transform on the noisy voice to convert the noisy voice into a frequency domain; compressing the high-frequency spectrum to an equivalent rectangular bandwidth (ERB) sub-band space by using a frequency band compression module; the compressed spectrum features are input into an encoder to extract high-order time-frequency features, in-depth modeling is carried out on contextual information in time and frequency dimensions through a grouping double-path long and short term memory module, and then the features are restored through a decoder; reconstructing an original spectrum resolution by means of a frequency band recovery operation; converting a result from a frequency domain to a time domain through short-time inverse Fourier transform, and reconstructing a voice waveform; constructing a joint loss function; and the result is converted from the frequency domain to the time domain through short-time inverse Fourier transform (iSTFT), the voice waveform is reconstructed, and the performance of the proposed model is evaluated. Through spectrum compression and a grouping parallel modeling strategy, while the voice perception quality, the voice definition and the background noise suppression capability are improved, the model parameter quantity and the calculation complexity are remarkably reduced.

The application discloses a snapshot spectral compression imaging method based on L1 norm and low rank technology, studies data fidelity and regularization in hyperspectral image reconstruction, models noise, and uses L1 norm to measure fidelity of the recovered hyperspectral image; a plug-and-play deep low rank prior is designed based on deep learning as a regularization term, the designed deep low rank prior is expressed as designing two neural networks to generate two matrices representing low rank characteristics of the hyperspectral image, random Gaussian noise is inputted to train the two networks in a self-supervised manner; the fidelity term with L1 norm and the regularization term with the deep low rank prior are combined as a hyperspectral image reconstruction optimization algorithm, and the hyperspectral image reconstruction optimization algorithm is solved based on a multiplier alternating direction method to obtain a target hyperspectral image. In the application of hyperspectral image reconstruction, the application has significant advantages, and higher fidelity and better imaging quality can be realized.