12 results about "Curvelet" patented technology

This invention provides a curvewave domain-based method for removing strong axes based on reflection coefficient attribute volumes, relating to the field of seismic data processing technology. The method includes: performing spectral inversion processing on raw seismic data to generate a reflection coefficient volume; based on the reflection coefficient volume, performing multi-scale, multi-directional layer-by-layer analysis in the curvewave domain, i.e., extracting reflection coefficient curves for each analysis window, determining inflection point locations by calculating the zero point of the second derivative of the curves to identify abrupt changes in the reflection coefficient; simultaneously, constructing local discrete surfaces based on the reflection coefficient volume, identifying local geometric anomaly regions of the reflection coefficient volume by fitting surface patches and calculating Gaussian curvature; and extracting strong reflection coefficient components by combining the determined inflection point locations and the identified local geometric anomaly regions. This invention can achieve precise stripping of strong reflection energy and faithful protection of effective weak signals, improving the reliability of seismic data in identifying and analyzing subsurface lithology and geological structures.

The invention discloses a satellite remote sensing image fusion method based on Wrapping-Curvelet transformation, and the method comprises the steps: carrying out the spatial registration of a multispectral image and a panchromatic image, carrying out the resampling of the multispectral image according to bilinear interpolation, obtaining a multispectral image with the same pixel size as the panchromatic image, and carrying out the fusion of the multispectral image and the panchromatic image. The method comprises the following steps of: respectively performing histogram matching on a panchromatic image and three wave bands of a multispectral image to be fused to obtain a new panchromatic image, taking the three wave bands of the multispectral image and the corresponding panchromatic image as examples, respectively performing Wrapping-Curvelet transformation on the three wave bands of the multispectral image and the corresponding panchromatic image to obtain respective Curvelet coefficients, fusing the coefficients of each layer, and then performing inverse Wrapping-Curvelet transformation on the fused layers to obtain a new panchromatic image; and obtaining a fused image. The blocking effect caused by blocking is avoided, the number of parameters of the implementation method is small, redundancy is reduced, and the algorithm is simple, convenient and rapid.

The present application relates to a reinforced concrete defect positioning method and system based on ground penetrating radar feature fusion, belonging to the technical field of structure nondestructive testing and ground penetrating radar signal processing. The method comprises: obtaining the original ground penetrating radar B-scan image of the reinforced concrete structure, removing the direct wave by Radon transform, and combining Curvelet multi-scale directional decomposition, adaptive threshold and region of interest constraint to suppress the steel clutter, to obtain a clutter suppression image; calculating the pixel-by-pixel absolute difference value of the original gray image and the clutter suppression image to obtain a clutter image; constructing the original gray image, the clutter suppression image and the clutter image into a pseudo-color three-channel fusion image and inputting the target detection network, and outputting the defect category, the boundary box and the actual physical coordinates. The present application can improve the detection and positioning performance of the hollow, honeycomb and delamination defects in the reinforced concrete under complex steel environment.

The invention provides an ltmnet contrast enhancement method based on a YUV color domain, and the method comprises the steps: S1, collecting and sorting low-contrast-high-contrast image sample pairs, and constructing a training set and a test set; s2, mapping RGB image data of the sample pair to a YUV color space, and carrying out 2: 1 sampling in the horizontal direction and 2: 1 sampling in the vertical direction on U and V channels; s3 is carried out on low-contrast YUV, and S6 is carried out on high-contrast YUV; s3, converting the sizes of the sampled Y, U and V channels into (512, 512, 3) as the input and output sizes of network training; s4, constructing an ltmnet adjustable network model with enhanced contrast in the YUV domain; a low-contrast-high-contrast data set is cut and processed, and a model structure is adjusted, so that YUV domain contrast enhancement is adapted; curve form transformation operation is transferred to a post-processing part from the inside of the network, so that mapping contrast enhancement of curves of different sizes is realized; s5, performing loss calculation on different curve mapping results and the corresponding high-contrast image YUV; and updating the network parameters.

The invention belongs to the technical field of air compressor efficiency evaluation, and provides an air compressor efficiency evaluation method and system based on decline trend prejudgment, and the method comprises the steps: comprehensively judging whether the high-frequency oscillation of an efficiency signal is caused by a rotating stall cluster or not through the time domain analysis and judgment of high-frequency periodic oscillation and the frequency domain judgment of a characteristic peak; if yes, performing dual-band division of a stall mass oscillation characteristic band and an efficiency decline low-frequency characteristic band; the method comprises the following steps: acquiring a compressor efficiency oscillation and decline composite signal, performing wavelet packet decomposition according to divided dual bands, removing decomposed oscillation characteristic frequency band signals, only retaining efficiency decline characteristic frequency band signals, and reconstructing an oscillation-interference-free pure efficiency decline trend curve through spline interpolation and moving average; and on the basis of the pure efficiency decline trend curve, fitting calculation analysis of the linear / nonlinear decline rate is conducted, and judgment of the efficiency decline trend of the gas compressor is achieved in combination with pure efficiency decline trend curve morphological analysis.

The application is suitable for the technical field of disease detection, and provides an automatic detection method for corn leaf disease targets, comprising the following steps: acquiring an original corn leaf image; extracting parallel features of the image, and capturing global structure features and local texture features respectively; fusing and encoding the global structure features and the local texture features; encoding by using a Transformer encoder; introducing target queries and decoding prediction; matching the prediction result with real labels, optimizing the allocation of each query, retaining the detection result with the highest confidence, and outputting the boundary box of all disease spot areas in the image and the corresponding classification. The application combines Curvelet transformation and a Transformer architecture, introduces a multi-scale and multi-direction structure information modeling mechanism, and fuses and models the local texture features extracted by a CNN in the Transformer, so that the perception ability for disease areas is improved.

The invention relates to the field of deep learning data processing based on coal rock CT scanning, in particular to a multi-modal deep learning coal rock stress-strain curve prediction method oriented to multiple loading rates. Comprising image feature coding; carrying out gray scale statistics and strain prior feature coding; loading a rate code; carrying out multi-stage cross-modal fusion; carrying out sequence modeling and curve decoding; physical quantity outline recovery and total stress-strain curve generation; and training and verifying. Wherein the loading rate is used as input, a loading rate and strain prior mapping mechanism is introduced, a multi-loading-rate working condition is uniformly supported in a single model, the robustness and generalization ability of loading rate change are remarkably improved, and an independent model is prevented from being trained for each loading rate. A double ResNet18 convolutional backbone network is adopted to decouple a crack image sequence and construct a component image sequence, gray statistics and a strain prior sequence are combined, and through a hierarchical fusion strategy, the morphology consistency of a complete stress-strain curve and the prediction stability of a peak segment and a residual segment are remarkably improved.

The invention discloses a method for detecting the wake direction of a surface ship from a synthetic aperture radar image. The method comprises the following steps of: identifying a surface ship; acquiring a rectangular frame surrounding the ship and a wake thereof; acquiring a Curvelet coefficient of a radar image corresponding to the rectangular frame; determining a scale serial number of the Curvelet coefficient used for detection; determining the direction corresponding to the maximum coefficient matrix average value under the selected scale, determining the normal direction of the ship wake, excluding 180-degree direction blur, and determining the direction of the ship wake. The method is used for automatically detecting the direction of the ship wake on the water surface from the synthetic aperture radar image.

The application discloses a kind of construction curve generation method, including the following main steps: S1, the first arch of standard cycloid is selected as the basis arc segment of construction curve half section;S2, the second arch of standard cycloid is copied after mirror image with straight line y=2r as the axis of symmetry, and the basic unit of construction curve is obtained;S3, the basic unit obtained in S2 is copied 2 (n-1) along the positive direction of x axis, and then copied 4 (n-1) r along the positive direction of y axis, that is, the nth basic unit (n=2, 3, 4, 5,...) of construction curve can be obtained, and the curve formed by the connection of the n basic units forms a complete construction curve.The application also discloses a flow channel structure and a printed circuit board heat exchanger.The application can improve the average rate of medium flow, reduce the flow dead space in the transition area of wave crest and wave trough, enhance the disturbance of fluid medium in the channel, thereby destroying the flow heat transfer boundary layer to achieve the purpose of heat transfer enhancement.

This invention discloses a method for classifying and recognizing ornamental plants based on graph neural networks, comprising the following steps: acquiring sample images of ornamental plants, performing preprocessing and organ region segmentation; calling the GSG network to construct a curvelet scattering module and performing multi-scale directional decomposition; mapping directional feature representations to edge space to construct a topological representation, preserving node connectivity and directional propagation relationships; performing Hodge decomposition on the topological representation, dividing it into three branches; constructing GSG scattering pyramids in each of the three branches, progressively expanding the depth and scale; performing feature alignment and weighted fusion to generate plant-level feature vectors; inputting the plant-level feature vectors into a classification module, outputting classification probabilities and category labels, and generating an organ region contribution heatmap. This invention enhances multi-scale feature extraction and orientation-sensitive modeling capabilities, improving the accuracy and stability of plant classification and recognition.

The invention discloses a multiple suppression method, electronic equipment, a storage medium and a device. The method comprises the following steps: acquiring seismic data of a target area; predicting a multiple model based on the seismic data; preprocessing the multiple wave model; curvelet domain decomposition is conducted on the preprocessed multiple wave model, and multiple pieces of multiple wave model component data are obtained; optimizing the component data of each multiple model based on the seismic data and the threshold value corresponding to each component data; converting and combining the optimized multiple wave model component data to obtain a time-space domain multiple wave model; and suppressing the multiples based on a time-space domain multiple model. According to the method, curvelet domain decomposition is carried out on the multiple wave model, multiple wave energy of different frequencies and inclination angles is optimized through the seismic data and the set corresponding threshold values, then multiple waves are suppressed, fine suppression of the multiple waves is achieved, and the suppression effect of the multiple waves is improved.

The application provides a method for removing strong axis based on a reflection coefficient attribute body in a curved wave domain, and relates to the technical field of seismic data processing. The method comprises the following steps: performing spectral inversion processing on original seismic data to generate a reflection coefficient body; based on the reflection coefficient body, performing multi-scale and multi-directional layer analysis in the curved wave domain, that is, for each analysis time window, extracting a reflection coefficient curve, determining the inflection point position by calculating the zero value point of the second derivative of the curve, and using the inflection point position to identify the mutation characteristics of the reflection coefficient; at the same time, constructing a local discrete curved surface based on the reflection coefficient body, fitting the curved surface sheet, and calculating the Gaussian curvature to identify the local geometric abnormal area of the reflection coefficient body; and comprehensively determining the inflection point position and the identified local geometric abnormal area to extract a strong reflection coefficient component. The application can realize accurate stripping of strong reflection energy and faithful protection of effective weak signals, and improves the reliability of identification and analysis of underground stratum lithology and geological structure of seismic data.