195 results about "Multiple-scale analysis" patented technology

A method for estimating 3D posture of a human body combining densely connecting attention pyramid residual network and equidistance restriction is composed of a discriminative human body 2D posture estimation part and a generative human body 3D posture estimation part. Firstly, a 2D human posture estimation model is constructed. The 2D human posture estimation model includes attention pyramid residual blocks and a funnel sub-network composed of several attention pyramid residuals. The attention pyramid residuals are used for multi-scale image feature extraction, and the funnel sub-network is used to generate human joint thermodynamic map. In order to solve the problem that the environmental context information is not fully utilized, the attention mechanism and multi-scale analysis are combined to capture the environmental context characteristics. In order to solve the problem of gradient disappearance/gradient explosion, the dense connection network is combined with the above attentionmechanism to improve the discrimination of a feature map. Then the loss function is constructed and the equidistant constraint term is introduced to fit the 3D posture of a human body by minimizing the loss function. The method of the invention has obvious advantages in the human body 3D posture estimation task.

The invention discloses a multi-scale and multi-hierarchy method for fusing a visible light full-color image and an infrared remote sensing image, which is used for fusing the visible light full-color image and the infrared remote sensing image acquired from an aerospace aviation sensor platform. According to the method, a pseudo-color image of an infrared image is obtained by temperature retrieval of the infrared image, and then the detail information of a visible image is fused into an infrared retrieval image by utilizing a multi-scale and multi-hierarchy image fusion method. The method comprises two parts, firstly, temperature retrieval is carried out on the infrared remote sensing image to obtain the infrared pseudo-color image; and secondly, the infrared pseudo-color image and the visible image are fused by means of multi-scale analysis of the image. By using the method provided by the invention, the fused image can not only reflect the temperature information of an infrared image scene, but also reflect the temperature information of a specific target.

The invention relates to a digital image reinforcing method based on wavelet reconstruction and decomposition and a device thereof, which belongs to the image data processing field. The digital image reinforcing method comprises image pretreatment, decomposition process, transformation process and reconstruction process, and is characterized in that the image decomposition process comprises multi-group integrals and down-sampling process; the transformation process comprises the statistics, analysis and modification processes of mid data information; the reconstruction process comprises multi-group up-sampling, integrals and adding processes; in the transformation process, adaptive histogram equalization technology and noise restraining are adopted. The invention, based on multi-scale analysis, realizes to reinforce the contrast of digital image and prevents noise simultaneously through the wavelet decomposition, transformation and reconstruction and adopting the adaptive histogram equalization technology and noise restraining in transform domain and can be widely used in such fields related to the improvement of digital image quality as medical information systems, industrial nondestructive testing systems and other similar image systems.

A high-speed automatic multi-object tracking method with kernelized correlation filters is provided. The method includes obtaining an image frame from a plurality of image frames in a video, extracting a foreground object sequence from the obtained image frame, and determining similarity between each foreground object of the extracted foreground object sequence and a tracked object. The method also includes calculating HOG features of the foreground objects with a lower similarity, obtaining training samples for each of the foreground objects with the lower similarity using a circular matrix, obtaining a classifier via a kernel method accelerated by FFTW, and obtaining tracking points using a sparse optical flow. Further, the method includes detecting object matching responses using a detection response function, performing multi-scale analysis for the object based on an optical flow method, and processing a next image frame of the plurality of image frames in the video until the video ends.

The invention provides a signal identification and classification method. The method comprises the followings steps of: carrying out noise reduction on initial data containing higher noise by utilizing a wavelet transform method, decomposing signals into high-frequency information and low-frequency information in data analysis, carrying out noise cancelling on the signals by adopting a soft thresholding method and then carrying out signal reconstruction; carrying out further decomposition on the high-frequency part which is not detailedly classified by multiscale analysis while inheriting allthe favorable time-frequency localization advantages of the wavelet transform; analyzing the signals within different frequency bands after multi-layered decomposition by utilizing the wavelet packettransform to extract out characteristic information reflecting a system state; transforming the characteristic vectors of input signals into a high-dimensional characteristic space through non-lineartransform and then solving for an optimal linear classification plane in the high-dimensional characteristic space. The invention overcomes the defects of difficult determination of a network structure, low convergence rate, requirement on large quantities of data samples during training, and the like in neural network learning and enables the neural network learning to be with the characteristics of high precision and strong real time in the aspect of practical application of engineering.

The invention discloses a heterology remote sensing image registration method. According to the core idea, multi-scale matching is taken as a basis; straight line intersection points are used as elements; the point matching method of a joint Voronoi map and a spectrogram is used; iteration feature extraction and a matching policy are integrated; and the problems of heavy dependence on feature extraction, poor reliability, low accuracy and the like of the existing method are overcome. The method comprises the steps that multi-scale analysis is carried out on original images; straight line extraction and intersection point acquiring are carried out on the coarsest scale; the point matching method of the joint Voronoi map and the spectrogram is carried out on intersection point sets to acquire a homonymous point pair; whether a matching result is qualified is detected: if the matching result is qualified, going to the next step is carried out, otherwise self-adaptive parameter adjustment is carried out and straight line extraction and point set matching are carried out again; original transformation is carried out on the images to be registered, and straight line features are respectively extracted; homonymous straight line segments are searched, and a candidate homonymous point pair is acquired; a KNN map is used to acquire an accurate matched point pair; and a transformation parameter is solved. The method provided by the invention is mainly used for the registration of visible light, infrared, synthetic aperture radar (SAR) and other heterology remote sensing images.

The invention discloses an aquaculture water quality short-time combination forecast method on the basis of multi-scale analysis. The method includes the steps that water quality time sequence data are acquired online and repaired; through empirical mode decomposition, the selected water quality time sequence sample set data are decomposed into IMF components and residual rn components, wherein the IMF components and the residual rn components are different in frequency scale; the IMF components and the rn components are classified, a manual bee colony optimization least square support vector regression machine, a BP neural network and an autoregressive sliding average model are respectively selected for forecast according to classifying features, and finally, all results are weighed and summed to obtain a water quality time sequence forecast result. According to the method, the original water quality time sequence data are decomposed into the components different in time frequency through the empirical mode decomposition, and change conditions in original water quality sequences can be mastered more accurately; advantages of the manual bee colony optimization least square support vector regression machine, advantages of the BP neural network and advantages of the autoregressive sliding average model are complemented and combined, and thus performance of a combined forecast model is effectively improved.

The invention discloses a high-resolution remote sensing image segmentation method based on inter-scale mapping. The high-resolution remote sensing image segmentation method has the advantages that aiming at geographical object extraction involved in objet-level change detection, the high-resolution remote sensing image multi-scale segmentation method based on wavelet transform and an improved JSEG (joint systems engineering group) algorithm is provided; aiming at a key problem that a conventional JSEG algorithm affects segmentation accuracy in high-resolution remote sensing image segmentation, a corresponding improvement strategy is adopted to achieve a good effect; wavelet transform is introduced to serve as a multi-scale analysis tool, excessive rough color quantization of the conventional JSEG algorithm is abandoned, and thus, detail information in original images is kept to the utmost; a novel inter-scale segmentation result mapping mechanism is set up, image segmentation of a current scale is realized on the basis of a segmentation result of a previous scale, and correction of the segmentation result of the previous scale is also realized, and thus, accumulation of inter-scale segmentation errors is decreased effectively; finally, an improved multi-scale segmentation strategy and an improved region merging strategy are provided, and over-segmentation and mistaken merging are effectively reduced.

The invention relates to a flexible DC power transmission overhead line fault rapid determining method without setting a threshold. The method is characterized in that single-end travelling wave signals which are rapidly spread in a power transmission line during a fault are employed as original data for fault determination, the rapidness and the economic benefits are excellent; and mold maximum change conditions after single-end transient voltage travelling wave signals and single-end transient current travelling wave signals are subjected to wavelet multiple dimensioned analysis are taken as a criterion for determining whether a system generates a fault and whether the fault is a zone inside fault or a zone outside fault, and at the same time, amplitude of forward traveling waves and backward travelling waves of the single-end transient voltage travelling wave signals is taken as a fault direction criterion. The method is simple in principle, is effective, is easy to implement, does not require setting and arranging a threshold value, avoids determining errors and protection maloperation which are easily caused by manual arrangement and setting of the threshold through reliance on engineering experience in a conventional method, and the reliability is higher compared to other methods. Finally, through combination with wavelet feature entropy identification fault types, wrong determination is effectively prevented, and the fault identification accuracy is effectively improved.

The invention discloses a partitioning algorithm for choroidal neovascularization in an OCT image. The algorithm includes steps of S01, designing a structure prior learning method for a training image and constructing a structure prior matrix used for distinguishing a choroidal neovascularization zone and a background zone; S02, converting the OCT original image to a saliency enhancing image used for enhancing the saliency of the choroidal neovascularization zone based on the structure prior matrix; S03, adopting multi-scale analysis on the saliency enhancing image and dividing the saliency enhancing image into m scales; S04, acquiring m trained convolutionneural network model based on each scale training; S05, processing a testing image by utilizing the step S01, S02, S03 and performing testing by utilizing the trained convolutionneural network model in the step S04, outputting m portioning results and fusing the m partitioning results into the final partitioning result. By adopting the algorithm provided by the invention, the precision of partitioning of the choroidal neovascularization in the OCT image can be improved distinctively.

The invention discloses an extraction method of visual salient regions based on multiscale relative entropy. The extraction method of the visual salient regions based on the multiscale relative entropy comprises extracting different global color feature characteristic patterns and direction characteristic patterns from input images, carrying out gaussian pyramid decomposition and multiscale normalization on the global color feature characteristic patterns and the direction characteristic patterns, respectively computing color partial relative entropy and direction partial relative entropy of respective feature space, respectively normalizing results of the different feature space, carrying out linearity superposition of the results of the different feature space, linearly adding color saliency maps and direction saliency maps together, carrying out two-dimensional gaussian smoothing, and therefore extracting visual salient regions. Compared with a traditional method, the extraction method of the visual salient regions based on the multiscale relative entropy takes a full account of color features and direction features reflecting global significance and relative entropy reflecting partial novelty, adopts a method of multiscale analysis, and has the advantages of being effective and reliable. The extraction method of the visual salient regions based on the multiscale relative entropy is capable of being used for self-adaption compressed sampling and target detection of natural images, remote sensing images and the like, and is wide in application prospect of lowcost imaging devices.

The invention discloses a distributed optical fiber fence intrusion detection and location method based on a phi-OTDR (Optical Time Domain Reflectometer), and the method comprises the following steps of: firstly, subtracting an acquired ith signal by an acquired (i-1)th signal to obtain a preprocessing monitoring signal; secondly, carrying out at least four-layer wavelet decomposition on the preprocessing monitoring signal by adopting a plurality of wavelet functions; thirdly, judging whether actual intrusion exists according to the consistency of distribution positions of maximum-wavelet mode maximums in all scale signals; fourthly, judging that the actual intrusion exists if the distribution positions of the maximum-wavelet mode maximums in at least three scale signals are consistent; fifthly, corresponding multiplying the scale signals in which the intrusion exists and the distribution positions of the maximum-wavelet mode maximums are consistent to obtain a composite signal of which pseudo maximums are inhibited and mode maximums at a catastrophe point are enhanced, using a position corresponding to the maximum-mode maximum in the composite signal as a position of the determined intrusion point, determining the intrusion generation and the intrusion position by combining various composite signal judgment results of wavelet multi-scale analysis according to the majority voting criterion.

The invention discloses a texture force touch sensing method based on single image fractional order processing. The method includes the steps that a fractional differential algorithm is utilized, the order gamma of the fractional differential algorithm is adjusted dynamically, an enhanced texture image is obtained, original image texture information is extracted from the enhanced texture image, and multi-scale analysis of the detail texture and the edge contour of different rates of gray level changes is achieved from the point of image airspace gray level change features; force touch rendering is carried out on the virtual surface texture, and tangential force frictional damping coefficients of a force touch sensing model are controlled on the basis of gradient vector distribution of the enhanced texture image; obtained resultant force is calculated and fed back to an operator through a hand controller. By the utilization of sensibility of fractional differential to detail information, the sense of reality of texture force touch sensing is enhanced by manually selecting orders, extracting texture information in a two-dimensional image signal from different scales and converting the texture information into frictional damping control.

The invention provides a micro-motion target characteristic extraction method and system based on wavelet multi-scale analysis. The micro-motion target characteristic extraction method based on wavelet multi-scale analysis comprises the steps that a micro Doppler model of a precessional motion target and a micro Doppler model of a rolling target are established, narrow-band radar scattering cross section data of the precessional motion target and narrow-band radar scattering cross section data of the rolling target are obtained based on ballistic trajectory simulation, the high-frequency component of the narrow-band radar scattering cross section data of the precessional motion target and the high-frequency component of the narrow-band radar scattering cross section data of the rolling target are extracted by means of wavelet multi-scale analysis, and column diagram characteristics are extracted aiming at the high-frequency component of the precessional motion target and the high-frequency component of the rolling target. According to the micro-motion target characteristic extraction method based on wavelet multi-scale analysis, the micro Doppler high-frequency components of the targets in different micro-motion forms are separated out by means of wavelet multi-scale analysis, new characteristics are extracted for recognizing the truth of targets, and the recognition accuracy of precessional motion targets and rolling targets is improved.

The invention discloses an image fusion quality evaluation method. The image fusion quality evaluation method comprises the steps of obtaining source images, carrying out image preprocessing, carrying out image fusion and carrying out fusion quality evaluation. According to the step of obtaining the source images, the source images need to be synchronous in time and cover the same area in space. According to the step of carrying out image preprocessing, the quadratic polynomial registration method is adopted and the nearest neighbor interpolation method is adopted for resampling. According to the step of carrying out image fusion, the fusion method combining the multiscale analysis method and the component substitution method is adopted for the source image to obtain fusion images through different fusion methods. According to the step of carrying out fusion quality evaluation, cross entropy and structural similarity of the fusion images and the source images are calculated, a weighting function model of the cross entropy and the structural similarity is built, and then a total fusion quality evaluation value under a preset weight is calculated. According to the image fusion quality evaluation method, the problems that evaluation indexes are multiple and related and subjective evaluation varies from person to person in a traditional fusion quality evaluation method are solved. The image fusion quality evaluation method is simple and universal.

The invention discloses a grain size nondestructive evaluation method based on haar wavelet. The method comprises the following steps: data acquisition is conducted on a reference test block, time-size distribution is obtained by utilizing haar wavelet transform, furthermore, the average multi-scale attenuation coefficient of all reference blocks is calculated, an ultrasonic multi-scale attenuation evaluation model for the average grain size is built according to the preset size assembly and the preset normalized weight, and finally grain size evaluation of the test block of which the grain size is unknown is realized through the built ultrasonic multi-scale attenuation evaluation model of the average grain size. The method reduces system errors in grain size measurement; for test blocks of which the average grain size is 103.5 [mu]m measured by adopting metallography, the evaluation result is 101.7 [mu]m, and the error is controlled to be +/-2%. Therefore, through multi-scale analysis of original ultrasonic A signals, the method provided by the invention can discover more rich grain size information in the original ultrasonic A signals, and further improves the nondestructive evaluation precision of the grain size.

The invention provides a multi-dimensional predicting method for strength of a braided ceramic base composite material. Firstly a parameterization method is used for establishing a braided ceramic base composite material single-cell model which comprises yarns and a matrix, and then period grid dividing is performed. Afterwards a mesomechanics method is used for simulating the mechanics behavior of the yarn unit. Furthermore a strain conversion matrix is used for converting a yarn element stiffness matrix from a local coordinate system to an integral coordinate system. Finally through applyingan increment type periodic boundary condition, a micro-stress and a strain field of a single cell are calculated. A volume averaging method is used for obtaining a microscopic stress and strain of the braided ceramic base composite material. If a failure unit band which penetrates through the single cell is formed, failure of the braided ceramic base composite material is determined, and the single-cell average stress is the strength of the braided ceramic base composite material. According to the multi-dimensional predicting method, the multidimensional analysis method in which macrography and mesography are combined is utilized. The strength of the braided ceramic base composite material can be accurately predicted without dependence to a large number of experiments with time consumption and high cost.

The invention discloses a dynamic supervising method for gas-solid fluidized bed and the system. The method is: using wavelet analysis technology to analyze the pressure vibration signal, judging its flow state, displaying the flow state with capacitance chromatography imaging technology, measuring the aperture rate and its vibration of the fluidized bed, extracts the character value and carries on flow state judgment, using multi-sensor information melting technology to carry on flow state judgment, upgrading the accuracy of the judgment. The supervising system has a fluidized bed main body, under which there has a gas distribution board, gas chamber, on the external wall in the middle of the main body there arranges a capacitance chromatography imaging sensor, pressure sensor, the capacitance chromatography imaging sensor is connected to the tiny capacitance measuring model, data collecting system, and computer, the pressure sensor is connected to A/D converting card, and computer.

The present invention relates to a wind turbine converter fault diagnosis method based on wavelet transformation and a DBN and belongs to the field of wind turbine converter fault diagnosis based on wavelet transformation and DBNs and provides a wind turbine converter fault diagnosis method based on wavelet transformation and a DBN. As the installed capacity of wind turbines increases, the wind turbines occupy an increasingly high portion in a power grid, and thus, fast identification of the locations of wind turbine faults is pivotal to the stable operation of the wind turbines and the powergrid. According to the method, multi-scale analysis is performed on signals through using wavelet transformation; fault signal feature vectors are extracted; and the DBN performs supervised learning on the feature vectors, so that a fault identification model can be obtained. The method can excellently identify converter faults.