63results about How to "Effective denoising" patented technology

The invention discloses a dispersion tensor magnetic resonance image tensor domain non-local mean denoising method, and belongs to the technical field of digital image processing and applied mathematics interdisciplines. The problem that a dispersion tensor magnetic resonance image is easily affected by noise is solved. The method comprises the steps of firstly, sequentially conducting traversal on voxels of the dispersion tensor magnetic resonance image, setting a corresponding search region by using each voxel obtained by traversal as the center, then, conducting tensor matrix similarity comparison between all the voxels inside the search region and the center voxel, finally giving different weights to the voxels inside the search region according to the degree of the tensor matrix similarity, calculating a weighted mean tensor matrix, and obtaining the denoising result of the center voxel. The problem that the dispersion tensor magnetic resonance image is easily affected by the noise is solved.

The invention relates to a power quality signal denoising method based on an improved wavelet threshold function. Firstly, an original power quality signal is collected; the signal is preprocessed, aset of wavelet coefficients are obtained by selecting db3 wavelet basis function to decompose the one-dimensional contaminated power quality analog signal into three wavelet layers; then, the improvedwavelet threshold function and the threshold calculated by the unified threshold method are used to quantize the threshold, and the estimated wavelet coefficients are obtained; finally, the high-frequency wavelet coefficients of the first layer to the third layer and the low-frequency wavelet coefficients of the third layer after the threshold quantization process are used to carry out the wavelet inverse transform, and the signal is reconstructed to obtain the reconstructed signal. The improved wavelet threshold function solves the contradiction between denoising and retaining the local characteristic information of the original power quality signal.

The invention provides a time sequence signal efficient denoising and high-precision reconstruction modeling method and system. The method comprises: carrying out data preprocessing on original pulsewave signals; selecting a preset signal duration, and dividing the pulse wave signals after data preprocessing into a prediction set, a training set and a test set; selecting a convolutional neural network as a basic model of the deep convolutional noise reduction auto-encoder, and obtaining a deep convolutional noise reduction auto-encoder model according to a signal denoising requirement; inputting the training set into a deep convolution noise reduction auto-encoder model for training, and optimizing and selecting parameters of the deep convolution noise reduction auto-encoder model by using the regularization parameters and the test set to obtain an optimal deep learning model; and inputting the noisy pulse wave signal prediction set into the optimal deep learning model to obtain deepstructure features, performing signal reconstruction and denoising processing, and evaluating model performance. According to the method, denoising and reconstruction of the pulse wave signals are effectively carried out, and a new thought is provided for filtering same-frequency noise interference in the pulse wave signals.

The invention discloses a leakage detection method and a leakage detection device based on a combination of a wavelet and EMD reconstruction. The pipeline leakage detection device takes an STM32 single-chip microcomputer as a core, and mainly comprises a sensor, an amplifier, a filter, an A/D conversion module and a transmission module. After the pipeline leakage detection device collects data, the data are transmitted to a computer through a wireless communication module; the computer uses Matlab software to carry out denoising processing on the collected data by combining the wavelet universal threshold denoising and the EMD reconstruction of a multi-azimuth neighborhood mode, positioning is carried out through correlation analysis, so that the position of a pipeline leakage point is obtained. According to the leakage detection method and the leakage detection device, noise can be effectively removed, useful signals can be extracted, and therefore the accuracy of positioning the pipeline leakage point is remarkably improved.

The invention provides an electric vehicle wireless charging device automatic alignment system and method. The system comprises a wireless charging starting button, a wireless charging receiving device, a wireless charging transmitting device, an image recognition automatic locating device and a moving control driving device, wherein the image recognition automatic locating device obtains the distance between the wireless charging transmitting device and the wireless charging receiving device according to collected information, and controls the moving control driving device and the wireless charging receiving device to move in the horizontal direction and the longitudinal direction. According to the electric vehicle wireless charging device automatic alignment system and method, a mobile wireless receiving device is adopted to achieve automatic alignment, compared with a method in which a driver uses a park assist system, the complicated parking procedures are reduced, the installation is simple, and the cost is lowered; the relative distance of a wireless transceiving device is obtained through a binocular vision system, through a horizontal control driving device, a longitudinal control driving device and a vertical lifting control driving device, the purpose that the receiving device moves in the vertical direction is achieved, and the problem that charging efficiency of vehicles is low when underpans differ in height is solved.

The invention provides a de-noising method for an ultra-high-voltage direct-current corona current signal. The de-noising method mainly comprises a step of discrete wavelet transform of the signal and a minimum description length criterion based on an information theory. According to the main idea of the method, a best signal model is searched to achieve best expression of an original signal, and then the module is encoded, thereby achieving the purpose of de-noising; and by virtue of the combination of the discrete wavelet transform of the signal and the minimum description length criterion,a better de-noising effect is achieved. The method has the characteristics of being free of a predetermined threshold, and self-adaptive to data.

The invention discloses an intelligent counting method of interference fringes. The intelligent counting method comprises the following specific steps of: filtering random impulse noise from the data acquired by a linear array charge coupled device (CCD) sensor by adopting a self-adapting median diffusion filtering algorithm; reducing Gaussian noise by using a non-linear diffusion filtering algorithm; determining the reference range of effective data of the interference fringes, wherein the maximum gray value point Max, the minimum gray value point Min and the threshold value T are calculated, and the threshold value T refers to the point of which the gray value changes most in the determined reference range of the effective data; and performing second order derivation on a curve which is determined by an effective data range, and judging the changing state of the interference fringes according to second order derivative numerical value changes so as to realize automatic counting of the interference rings. By using the method, a diffusion median filtering algorithm is used for performing filtering denoising on the acquired data, the changes of the interference rings are further judged by judging the second order derivative numerical value changes in the effective reference range in real time, and the counting stability is promoted.

The invention relates to a ground penetrating radar B-scan image denoising method, which comprises the following steps: step 1, enabling a ground penetrating radar GPR to detect a single background medium area in which a pipeline target is pre-buried on the ground surface, and obtaining Z GPR B-scan images to form a noise-free GPR label data set; 2, training a multi-scale convolution auto-encoder by using each noise-containing GPR data set and the corresponding noise-free GPR label data set, wherein the multi-scale convolution self-encoder comprises an encoder E and a decoder D; and 3, inputting the GPR image to be denoised into the trained multi-scale convolution auto-encoder, outputting the denoised GPR image through encoding and decoding, calculating the signal-to-noise ratio of the denoised GPR image, and verifying the denoising effect of the multi-scale convolution auto-encoder. According to the invention, the noise-containing GPR image under the condition of low signal-to-noise ratio can be effectively denoised.

The invention relates to a circular self-adaptation template based image weighted mean filtering method. The technical key points of the method are that a circular template is adopted to carry out the expansion of a self-adaptation template, and noise elimination is carried out by combining and calculating the distance weighting and the normalization weight of signal points relating to calculation. The method has the advantages that firstly, the circular template is used, and contains more valuable information, compared with a traditional square template, so that noise-eliminated images can better approximate to primary images; secondly, the circular template can be adaptively expanded according to the number of signal points in the template, with regard to images polluted by salt and pepper noises with various intensity (1-90 percent), which all obtain a relatively very good filtering effect; and lastly, mean calculation utilizes distance information of image space, and the filtering effect is further improved.

The invention discloses a method for detecting image changing by combining deep convolutional neural network with morphology. The method comprises the following steps: segmenting registered remote sensing images of different time phases; rotating and mirroring the segmented images, and combining the remote sensing images at the corresponding locations of different time phases into a 8-channel image; inputting the obtained 8-channel image into a SegNet network model to train, and outputting a 2-channel image; adopting and operating the image so as to perform hole filling on the image, and thenremoving noise information by adopting a corrosion operation so as to obtain an image processing model; segmenting the to-be-detected remote sensing images and then inputting into the model of the previous step to process, and outputting the images; combining the output images into the size of the original to-be-detected remote sensing image, thereby accomplishing the image change detection. By adopting the method of combining the deep convolutional neural network with the morphology, the detection precision is high, the noise is effectively removed, the method is simple, the detection on thebuilding change has high accuracy and robustness.

The invention discloses a method for denoising based on a pulse type ToF depth image. The method comprises the following steps: directly starting from an imaging model of a pulse type ToF depth camera, designing a regularization device of original measurement data of a pulse type ToF sensor, carrying out depth denoising by adopting a variation-based semi-blind deconvolution algorithm, effectivelydenoising a pulse type ToF depth image, and keeping edge detail information of the depth image.

The invention discloses a life signal enhancement method based on segmented and classified enhancement processing, which is used for removing background clutters of the ultra wide band radar echo signals; selecting a target signal registration block by utilizing simulation scene simulation data; and judging whether the actual measurement window is matched with the registration block or not by using a sliding window, performing classified and segmented denoising and enhancement on a radar signal matrix, performing multi-frame weighting processing on a processing result, and finally enhancing aremote signal by using a local normalization method to obtain a finally enhanced signal matrix for positioning and detecting a life entity. According to the invention, enhancement of life signals is realized; different types of signal matrix blocks are processed differently, denoising and enhancement are effectively carried out for the types of signals, correction is carried out by utilizing adjacent frames, and the amplitude of remote signals is enhanced, so that the position of a target signal is more obvious and accurate in both a time domain and a frequency domain, and a guarantee is provided for subsequent detection.

The invention belongs to the technical field of digital image processing and relates to a method for adaptively quantizing optical flow features on complex video monitoring scenes. According to the method, the local statistical features are calculated after probability denoising is performed on video space based on the optical flow features, then, the video space position is adaptively quantized, and the video space is divided into a plurality of micro-block areas; finally, each micro-block area is filtered through a motion complexity threshold value, the quantization number is judged, a visual dictionary is generated, and adaptive quantization is achieved. According to the method, the effectiveness and the diversity of motion on the video monitoring scenes are described based on the local statistical features of optical flow. The effective pixel ratio and the motion complexity features are fused, the liveliness of local motion is described, and then the optical flow feature position can be adaptively quantized. On the basis of the motion complexity features, the diversity of the local motion is described, and then the optical flow feature direction can be adaptively quantized. Better discriminability can be played through adaptive quantization of the optical flow features in the next scene analysis based on a word bag model.

The invention discloses an EMD and energy thereof based electrocardiosignal denoising algorithm and equipment, and a storage medium. The algorithm includes the following steps: performing signal averaging processing and determining a tolerance value; performing EMD on a signal to obtain IMF at all levels; according to the energy of the IMF at all levels after EMD, calculating, except for boundarypoints, the order where the first maximum point and the first minimum point are located of an energy curve, combining the tolerance value, judging an IMF order change point that needs to be denoised,and presetting the IMF order change point if the first maximum point and the first minimum point of the energy curve do not exist; performing threshold denoising on the IMF at all levels before the IMF order change point; and reconstructing the IMF at all levels after threshold denoising, the IMF without performing threshold denoising and residual, and generating a denoised electrocardiosignal. Through the scheme, the information greater than the threshold part can be completely reserved, additional shock and jump points cannot be generated, the smoothness of the original signals can be well guaranteed, and denoising quality can be enhanced.

The invention relates to an improved self-adaptive wavelet terahertz image denoising method. The method comprises the following steps of inputting a to-be-denoised original terahertz image; performingpixel distinguishing on the original terahertz image; performing fuzzy C mean method clustering on pixels, and dividing the original terahertz image into five parts; calculating average texture values of the five parts, and comparing the average texture values of the parts with a set threshold, thereby dividing the five parts into a smooth region and a non-smooth region; selecting proper waveletbasis and decomposition layer numbers for the smooth region and the non-smooth region, and decomposing the layers in sequence by utilizing a decomposition algorithm to obtain wavelet coefficients belonging to the layers; processing the obtained wavelet coefficients, determining a preset threshold according to a threshold selection principle of wavelet transformation, and performing filtering through a self-adaptive threshold function to remove noises; performing image reconstruction by using wavelet inverse transformation; and outputting the whole image. According to the method, the terahertzspectral image can be quickly and effectively denoised.

The invention provides a method and a device for processing a calligraphy and painting image. The method and the device can conduct the binarization processing for a captured calligraphy and painting image to obtain an image processed by binarization; extracting the outline of the image after the binarization processing to obtain at least a closed outline curve; smoothening the closed outline curve, including conducting Fourier transformation for the outline curve, that is, eliminating radio-frequency components in Fourier coefficients and then conducting the Inverse Fourier transformation, so as to the smoothened outline curve. The method and the device of the invention can not only effectively de-noise the calligraphy and painting image, but correctly restore the image.

The embodiment of the invention provides a voice processing method and device and a vehicle. The method is applied to the vehicle, a plurality of voice areas are arranged in the vehicle, and each voice area is provided with at least one piece of audio acquisition equipment and at least one piece of audio playing equipment. The method comprises the steps of determining a target voice area where a target speaker is located in the vehicle; determining target audio data played by the audio playing equipment arranged in the target voice area; and based on the target audio data, performing noise reduction on the voice data acquired by the audio acquisition equipment, and extracting the voice data of the target speaker. According to the embodiment of the invention, in the scene that the vehicle is provided with a plurality of voice areas, when the audio playing equipment in a certain voice area plays an audio, the voice data acquired by the audio acquisition equipment in the voice area is subjected to effective noise reduction, and the user voice data with high cleanliness is extracted.

The invention relates to an image processing technology, discloses a Monte Carlo rendering image denoising method based on a generative adversarial network, and solves the problem of low image denoising efficiency caused by long network reasoning time in the prior art, and a denoising result can better recover low-frequency content and high-frequency details of a noise rendering image, so that the denoising efficiency is improved. Therefore, a more real de-noising result in vision can be obtained. According to the method, accurate and efficient denoising of the Monte Carlo rendered image is realized based on the constructed Monte Carlo rendered image denoising model, and the Monte Carlo rendered image denoising model is trained by the generative adversarial network; the architecture of the generative adversarial network comprises a de-noising network and an identification network, and the de-noising network is mainly composed of a noise feature encoder and an auxiliary feature encoder; the identification network is mainly composed of an identifier; the denoising network inputs the noise rendering graph and the auxiliary cache graph and outputs a denoising rendering graph; and the identification network is used for identifying true and false images of the input de-noised rendering image and the target rendering image corresponding to the noise rendering image.

The invention discloses a pavement drainage monitoring and evaluating method and system based on image processing. The method comprises the steps: enabling a double-telecentric lens of a camera to collect image data information in the pavement drainage process; carrying out edge detection preprocessing on the image data information by adopting shear wave transformation; adopting a particle size determination method to extract and calculate bulges of the pavement in the drainage process; constructing evaluation indexes to evaluate the pavement drainage quality to obtain a pavement drainage quality evaluation result; adopting a Bayesian expert system classifier to classify the pavement drainage quality evaluation results, and classifying pavements; and verifying the evaluation performance ofthe Bayesian expert system classifier by adopting a chaotic matrix to obtain the evaluation performance of the pavement drainage process. According to the method, denoising can be effectively carriedout, the edge can be completely detected, and the evaluation efficiency of pavement drainage performance detection is improved.

The invention provides a low-frequency magnetotelluric data denoising method based on an over-complete dictionary and a compressed sensing reconstruction algorithm. The method comprises the followingsteps of firstly, extracting a rough low-frequency effective signal from a noisy magnetotelluric time sequence by using mathematical morphological filtering; then, using complementary set empirical mode decomposition to smooth the rough low-frequency effective signal so as to obtain an accurate low-frequency effective signal, and acquiring a noisy high frequency signal by subtracting the extractedlow frequency effective signal from the noisy magnetotelluric time sequence; and finally, through designing a suitable over-complete dictionary, using the compressed sensing reconstruction algorithmto carry out signal-noise separation on the noisy high frequency signal, and acquiring a de-noise high-frequency effective signal; and acquiring a full spectrum band magnetotelluric effective signal through adding the low-frequency effective signal and the high-frequency effective signal. In the invention, under the condition that the magnetotelluric effective signal is well reserved, a strong human noise in low-frequency magnetotelluric data is removed, a signal-to-noise ratio of the magnetotelluric data is significantly increased, and an apparent resistivity and a phase curve are improved.

The invention belongs to the technical field of hand drawing animation, and provides a bitmap-based hand drawing material generation method, which comprises the steps of enhancing a bitmap image, highlighting a pixel value change trend of the bitmap image, generating an image contour through a connection pixel value change trend, denoising to generate a denoised image contour, and then embedding the bitmap image into the de-noised image contour to generate a hand drawing material, so that the bitmap has a contour path and becomes a displayable hand drawing material. The bitmap image can be vividly and efficiently restored by drawing along the contour path of the bitmap when a user performs hand drawing, thereby greatly reducing the drawing loss, accelerating the rapid generation of the image. and improving the user experience. In addition, the threshold and the cost for the user to prepare the hand drawing animation material are reduced, and the hand drawing animation effect is enhanced.