34results about How to "Keep useful information" patented technology

The invention discloses a method and a system for denoising partial discharge signals of power cables and extracting effective signals. A new threshold function is proposed to perform threshold processing on wavelet coefficients after wavelet decomposition. The problems existing in an existing hard threshold method and a soft threshold method for processing signals can be solved. Denoising processing and extraction of the effective signals can be performed on cable operation data laid in the cable trench or directly buried underground. Accurate positioning of cable faults is realized, and fault points are effectively found out in time, thereby guaranteeing safe and effective operation of electric power facilities.

The invention provides a method for utilizing a terahertz time-domain spectroscopy technology to detect amino acid content in grain. The method comprises the following steps: grinding and then performing tableting on a grain sample to be detected, directly utilizing a terahertz time-domain spectroscopy system to detect the sample subject to tableting in a transmission measurement mode and under the nitrogen condition to obtain terahertz time-domain spectroscopy signals as sample signals, under the same conditions, acquiring the terahertz time-domain spectroscopy signals under the nitrogen condition as reference signals, then obtaining an absorptivity spectrum of the grain tableting sample to be detected, at last dividing the absorptivity spectrum of the grain tableting sample to be detected into a calibration set sample absorptivity spectrum and a verification set sample absorptivity spectrum, utilizing a partial least squares regression method to establish a quantitative analysis mode, and obtaining the quantitative detection value of the grain sample to be detected. The method disclosed by the invention is simple in preparation of the sample, is not required to process any pretreatment, and can practically and effectively realize fast and accurate quantitative detection on the amino acid in the grain or food.

The invention discloses a fatigue driving recognition method and system based on multiple features, and the method and system carry out the preprocessing of an image, filter noise, and avoid problemsthat the image quality is poor and the detection precision is low because the image is affected by an external environment factor. The AdaBoost algorithm is adopted to stably, quickly and efficientlydetect the human face, so that the complexity of human face detection is reduced. According to the scale space-based face target tracking algorithm, a self-adaptive high-confidence updating strategy is adopted, and when an error occurs in a target tracking stage, the confidence of target detection is relatively low, and a model is not updated, so that the risk of drifting of the tracking algorithmis effectively reduced, and the tracking precision is improved. The SVM classifier is adopted for eye state recognition, the precision of eye state recognition is improved, and therefore the method is high in recognition precision and high in adaptability to the environment.

The invention relates to an automobile anti-halation method fusing visible and infrared images by means of enhanced HIS-Curvelet transformation in order to solve the automobile halation problem at night. The technical scheme comprises acquiring visible and infrared images of the road condition in front of an automobile at night; filtering and denoising the two kinds of images; taking the infraredimage as a reference image to register the visible image; converting the visible image to an IHS color space to obtain three components including the luminance I, the hue H and the saturation S; carrying out Curvelet decomposition on the luminance component I and the enhanced infrared image to obtain high and low frequency coefficients; employing a designed weight automatic adjusting strategy to fuse the low frequency coefficients; employing a modulus value maximization strategy to fuse the high frequency coefficients; carrying out Curvelet reconstruction on the fused high and low frequency coefficients to obtain a new luminance signal component I'; and carrying out HIS inverse transformation on the I', the original hue H and the saturation S to obtain a final fused image. According to theinvention, high-luminance halation information is rejected, and the image definition is effectively improved.

The invention discloses a face image normalizing method. The method comprises the following steps of 1, normalizing a target face image to acquire an initial face gray image; and 2, adjusting the gray value of the initial face gray image, wherein the gray value of a relatively flat face area is relatively reduced, and the gray value of a relatively un-flat face area is relatively increased. The scaling factor of a Weber face is adaptively adjusted according to a space flatness mask, intrinsic information of different space positions of a face base image under normal lighting condition can be used, and when shadow is caused by illumination in a target image, the shadow part has relatively small scaling factor, so that large response is not caused, and the problem that a shadow edge cannot be effectively processed by the primary Weber face method is effectively solved. Complementary information under different scales can be effectively used by a multi-scale adaptive Weber fusion method provided by the invention, so that more useful information for face identification / authentication is kept.

The invention relates to a word splitting method and device aiming at a URL. The method comprises the steps of 1, dividing a URL address according to an internal layer structure of the URL address toobtain several layer parts; 2, conducting symbol dividing and regular expression filtering on the layer parts in sequence; 3, dividing a character string obtained after processing conducted in step 2to obtain a URL word splitting sequence. The URL address is divided into five layer parts in step 1, namely, a protocol type, a free domain name, a secondary domain name, a primary domain name and a path; a bidirectional maximum matching algorithm and a probability model are utilized to divide the character string in step 3. The layer structure of the URL itself is made full use of, the URL can beefficiently divided, useful information in the URL address is maximally reserved, the obtained URL word splitting sequence can be used for feature analysis in tasks like webpage classification and angling URL detection, and the task accuracy can be effectively improved.

The invention provides a method for detecting amino acid content in grain by use of terahertz frequency-domain spectrum technology. The method comprises the steps of grinding grain samples to be tested and then tableting the grain samples to be tested, directly testing the grain samples to be tested by use of a terahertz time-domain spectrum system in a transmission measurement mode under a nitrogen atmosphere to obtain the terahertz time-domain spectrum signals of the samples, performing Fourier transform on the time-domain spectrums to obtain the frequency-domain spectrums of the samples, performing normalization spectrum preprocessing on all the obtained frequency-domain spectrums, and finally, dividing the preprocessed frequency-domain spectrums of the grain tableted samples to be tested into calibration set frequency-domain spectrums and verification set frequency-domain spectrums, and establishing a quantitative analysis model by use of the partial least squares regression method so as to obtain the quantitative detection values of the various grain samples to be tested. The method for detecting the amino acid content in the grain by use of the terahertz frequency-domain spectrum technology is simple in sample preparation, simple to operate, and capable of truly and effectively realizing fast and accurate quantitative detection on the amino acids in the grain.

The invention provides a pulse wave signal denoising method based on DTCWT-Spline. The method includes the steps that firstly, original noise-containing pulse wave signals are subjected to double-tree complex wavelet decomposition, wavelet coefficients on all layers are denoised through a Bayes maximum posteriori estimation threshold value, then double-tree complex wavelet inverse transformation is carried out, and pulse wave signals obtained after high-frequency noise is filtered out are obtained; the obtained pulse wave signals obtained after the high-frequency noise is filtered out are detected through a sliding window method, wave trough points in the pulse wave signals obtained after the high-frequency noise is removed are recognized, a wave trough curve is fit through a cubic spline interpolation method to serve as the estimated baseline drifting distance, finally subtracting the estimated baseline drifting distance from the pulse wave signals obtained after the high-frequency noise is removed is carried out, and the pulse wave signals are denoised. The high-frequency noise and baseline drifting can be effectively removed, overall characteristic information of the original pulse wave signals is well kept, the method is simple, low in calculated quantity and low in occupied memory, and a technical foundation is provided for research and development of miniaturized and mobile noninvasive continuous blood pressure detection equipment based on pulse waves.

The invention relates to a video abstract generation method and device, electronic equipment and a storage medium, and belongs to the technical field of the Internet. The method comprises the steps ofobtaining multiple frames of images in a target video, inputting the multiple frames of images into an image clustering model, dividing the multiple frames of images into multiple categories based onthe image clustering model, selecting the target image from each category, and splicing the selected multiple target images to obtain a video abstract of the target video. The method is based on theimage clustering model, can quickly cluster multiple frames of images, does not need to analyze the motion trail of the object in each frame of image, shortens the processing time, and improves the efficiency of generating the video abstract. Moreover, the target image is selected from each category, the target image can represent the content of the multiple frames of images in the category, the content of the target video can be accurately summarized according to the video abstract generated by the multiple target images, and the accuracy of the generated video abstract is improved.

The present invention discloses a three-dimensional CBCT (cone-beam computed tomography) image denoising method on the basis of improved nonlocal means. The three-dimensional CBCT image denoising method on the basis of the improved nonlocal means comprises: obtaining projection data of three-dimensional CBCT images having different angles, projection data of a three-dimensional CBCT image having an angle corresponding to a set of projection data of the three-dimensional CBCT images, calculating edge information of the three-dimensional CBCT images, and dividing background subblocks and texture subblocks of the three-dimensional CBCT images; respectively calculating a noise standard deviation of a background region and a mean of the average gradient values of the texture subblocks in the three-dimensional CBCT images; respectively calculating filtering intensity values of the projection data of the three-dimensional CBCT images having different angles according to the edge information, the noise standard deviation of the background region and the mean of the average gradient values of the texture subblocks in the three-dimensional CBCT images; and searching for other pixel points in the three-dimensional CBCT images similar to filtered pixel points, and calculating the similarity among other pixel points similar to the filtered pixel points to achieve the three-dimensional CBCT image denoising according to the filtering intensity values of the projection data of the three-dimensional CBCT images having different angles.

The embodiment of the invention provides a hierarchical nearest neighbor undersampling method based on clustering, which comprises the following steps: elbow diagrams of a plurality of class samples are obtained by using a Kmeans clustering algorithm, and an optimal cluster number k of clusters is selected according to the relationship between the sum of the cluster number and the distortion degree of each cluster; Kmeans clustering algorithm is used to cluster most of the samples into k clusters so as to obtain the number of center points and sample points in each cluster. According to the number of sample points in each cluster, stratified sampling is carried out, and the nearest neighbor of the center point of each cluster is combined with a small number of samples as the sampling result. The technical proposal provided by the embodiment of the invention fully utilizes the distribution characteristics of the majority class samples, better retains the useful information of the majority class samples, and can effectively improve the classification effect of the subsequent classification algorithm.

The invention provides an audio processing method and device based on artificial intelligence, electronic equipment and a computer readable storage medium, and relates to a cloud technology and an artificial intelligence technology; the method comprises the following steps: acquiring an audio clip of an audio scene , and the audio clip comprises noise; performing audio scene classification processing based on the audio clip to obtain an audio scene type corresponding to noise in the audio clip; and determining a target audio processing mode matched with the audio scene type, and applying the target audio processing mode to the audio clip of the audio scene. According to the invention, the audio quality can be improved based on the targeted audio processing of the audio scene.

The invention discloses a probability pairwise constraint-based self adaptive semi-supervised dimensionality reduction method comprising the following the steps: a projection matrix is calculated, a sparse representation graph is constructed, and weight of pairwise constraints can be calculated; according to an amount of information contained in each constraint, differentiating treatment can be performed in a self adaptive manner, a sparse representation method is used for capturing an inner structure of data, and a quadratic programming algorithm is adopted for assessing probabilities accounted for by information of all constraints; calculation of the projection matrix, construction of a adjacent graphs and updating of constraint weight are combined organically so as to form an integral body; the projection matrix is finally obtained, and low dimension representation of data can be obtained after the projection matrix is applied to original data. The method disclosed in the inventionis simple in required supervision information, high in calculation speed and large in popularization value; pertinent dimensionality reduction can be realized.

The invention provides a special-shaped radar combined decision target detecting method. A high threshold and a low threshold are utilized for performing CFAR detection on original video data of eachradar. After performing high-threshold detection and low-threshold detection on each data, a distance condensation point trace is extracted from the data. Normalization is performed on the point traceamplitude according to distance point trace local signal clutter / noise ratio and the noise substrate of each radar. Combined detection and decision is performed on the low-threshold detection condensation point traces of all radars according to a radar distance measurement error. Furthermore the high-threshold detection condensation point traces of multiple radars are combined, thereby finishingoriginal-video-grade multiple-radar combined target detection processing. According to the method of the invention, the high threshold and the low threshold are utilized, thereby improving target detection probability. For each radar, the distance point trace is extracted and then combined determining is performed, thereby preventing a problem of direct fusion processing incapability caused by different multiple-radar sampling rates, different resolutions, different noise substrates and the like.

A handheld molecular imaging navigation system comprises a multi-spectral light source module configured to provide light in a plurality of different spectrum bands in a time division control manner according to a control signal sequence to irradiate an inspected object; a time division control module configured to generate the control signal sequence; an optical signal acquisition module configured to acquire a near-infrared fluorescence image and a visible light image of the inspected object in a time division control manner according to the control signal sequence provided by the time division control module; and a processing module configured to perform image processing on the acquired near-infrared fluorescence image and visible light image according to the control signal sequence to fuse the visible light image and the fluorescence image and output the fused image, and output a feedback signal according to the acquired near-infrared fluorescence image and visible light image.

The invention discloses a pain detecting and positioning method and system based on brain waves and neural network, and the method comprises the steps: removing related noise of original electroencephalogram signals through independent component analysis, carrying out the pain grading of the electroencephalogram signals, carrying out the processing through a signal time division window, integrating the pain level data and then establishing a pain level data set; respectively generating spectral topographic maps of time windows of Theta, Alpha and Beta frequency bands related to pain through Fourier transform, azimuth isometric projection and a CloughTocher interpolation algorithm, and combining the spectral topographic maps into a multi-channel electroencephalogram sequence; obtaining a time-space feature vector of a brain wave sequence related to the pain degree and the pain position through a CNN-LSTM-AM neural network; and inputting the brain wave pain features learned by the CNN-LSTM-AM neural network into the pain classifier model to evaluate the pain level and the pain position. According to the invention, the pain degree change and position change characteristics of the brain waves can be accurately and efficiently extracted and processed, and the pain level and the pain position can be automatically identified.

The invention belongs to the technical field of privacy protection, and particularly relates to a general data release privacy protection method based on a random response technology. The method is based on the random response technology, and is characterized in that the calculation complexity of the reconstructed unbiased estimation result is reduced from exponential growth to linear growth by utilizing a matrix decomposition method and the Kronecker property, the error minimization of the unbiased estimation result is realized, the privacy protection parameters are allowed to be independently set for each attribute, and the calculation efficiency, the data availability and the application flexibility are effectively improved. The method supports the data release of a single sensitive attribute and multiple sensitive attributes, wherein the sensitive attributes comprise a Boolean attribute, a classified attribute and a numerical attribute; an efficient and flexible data publishing privacy protection mechanism can be provided for the scientific research and management in the fields of medical treatment, finance, biological information, transportation and the like, the data privacyis protected in the data publishing process, the useful information of published data is reserved, and the sharing of the data is promoted.

The invention discloses an earth surface deformation prediction method and system based on an InSAR technology and behavior characteristic deep learning, is applied to the technical field of earth surface deformation monitoring of the InSAR technology, and is used for obtaining image data, obtaining a prediction model based on an earth surface time sequence deformation prediction method combining the InSAR technology and a 3D-CNN network, and carrying out earth surface deformation prediction by combining noise reduction processing. According to the method, the ground surface settlement is monitored in a large range and with high precision through the InSAR technology, the economic investment of ground surface time sequence settlement monitoring can be effectively reduced, then the 3D-CNN network is utilized to learn ground surface time sequence deformation characteristics, the future ground surface deformation trend is predicted, and auxiliary decision making and technical support are provided for preventing ground surface settlement disasters; the method does not need to select a stable region to estimate deviation on a regional scale, carries out statistical analysis operation on the image after removing null values and abnormal values, can globally and automatically correct strip noise, and can retain useful information of a landslide deformation detection result to the greatest extent.

The invention provides a partial discharge signal denoising method for an oil-immersed transformer, and the method comprises the steps: inputting a to-be-denoised signal, and decomposing the to-be-denoised signal into K components through a VMD algorithm; determining a final decomposition component number K according to the energy loss coefficient; screening components containing more noises according to the warpage and permutation entropy of each component; using a wavelet packet denoising algorithm to denoise the component with more noise; and reconstructing the processed component into a partial discharge signal. According to the method, two types of noise contained in the partial discharge signal can be effectively removed, and useful information of the partial discharge signal is reserved to a greater extent.

The invention discloses a rapid image repairing method based on a rough set. The rapid image repairing method comprises the following steps: establishing a rough set model for an acquired digital image, and abstracting a knowledge representation system; then performing brightness grade division according to an equivalence relation in the knowledge representation system; solving illumination brightness layers of a discourse domain (an image) through upper approximation and lower approximation according to the division; then introducing approximate classification accuracy and system parameter importance, and calculating the variation trend of a value along with the division layer number; finally, feeding back optimal division according to the convergence of the approximate classification accuracy and the system parameter importance so as to self-adapt to the brightness layers, dividing all the brightness layers into an excessive region, a normal region and an over-dark region, and restoring the brightness layer by layer by using the normal region as the reference. A clear image with high visibility and outstanding details is obtained finally. The rapid image repairing method is high in adaptability and high in processing speed, has certain online timeliness, is of great realistic significance, and has a great practical value; various severe illumination environment conditions can be overcome.

The invention provides a pulse wave signal denoising method based on DTCWT-Spline. The method includes the steps that firstly, original noise-containing pulse wave signals are subjected to double-tree complex wavelet decomposition, wavelet coefficients on all layers are denoised through a Bayes maximum posteriori estimation threshold value, then double-tree complex wavelet inverse transformation is carried out, and pulse wave signals obtained after high-frequency noise is filtered out are obtained; the obtained pulse wave signals obtained after the high-frequency noise is filtered out are detected through a sliding window method, wave trough points in the pulse wave signals obtained after the high-frequency noise is removed are recognized, a wave trough curve is fit through a cubic spline interpolation method to serve as the estimated baseline drifting distance, finally subtracting the estimated baseline drifting distance from the pulse wave signals obtained after the high-frequency noise is removed is carried out, and the pulse wave signals are denoised. The high-frequency noise and baseline drifting can be effectively removed, overall characteristic information of the original pulse wave signals is well kept, the method is simple, low in calculated quantity and low in occupied memory, and a technical foundation is provided for research and development of miniaturized and mobile noninvasive continuous blood pressure detection equipment based on pulse waves.

The invention relates to an image feature recognition method, in particular to an image feature recognition method for a soil drought region, which comprises the following steps of: acquiring an image, storing, processing the image, and transmitting a processing result to an edge server module of a cloud server; the cloud server module is in wireless communication connection with the edge server module and used for storing soil basic information and providing computing power, and the user side module is in wireless communication connection with the cloud server module and used for controlling and checking the edge server module, judging an area needing irrigation and taking irrigation measures. According to the technical scheme for monitoring the crop area and identifying the drought area, the defect that the agricultural production cost is greatly increased due to the fact that a traditional crop irrigation technology is backward, can only be manually carried out, cannot intelligently monitor the crop condition and cannot carry out fixed-point and fixed-area precise irrigation can be effectively overcome.

The invention provides a method and a device for retrieving a three-dimensional object, wherein, the retrieval method comprises the following steps: obtaining the characteristic view of a three-dimensional object, and extracting the first Shock diagram skeleton descriptor and the first Fourier descriptor of the characteristic view; extracting the second Shock diagram skeleton descriptor and the second Fourier descriptor of an input picture; calculating the first similarity metric of the first and the second diagram skeleton descriptors; calculating the second similarity metric of the first and the second Fourier descriptors; and obtaining a mixed similarity metric according to the first similarity metric and the second similarity metric, confirming that the characteristic view corresponding to the maximum mixed similarity metric is the view most similar to the input picture, and further confirming that a three-dimensional object corresponding to the characteristic view is the three-dimensional object ready for retrieving. The invention effectively combines the Shock diagram skeleton descriptors and the Fourier descriptors, and realizes excellent 3D-object retrieval performance.

The present invention provides a method for detecting the content of amino acids in grains using terahertz frequency-domain spectroscopy technology. By grinding the grain samples to be tested and pressing them into tablets, a terahertz time-domain spectroscopy system is used to directly measure the content of amino acids in a transmission measurement mode under a nitrogen atmosphere. Test, obtain the terahertz time-domain spectrum signal of the sample, and obtain the frequency-domain spectrum of the sample after the time-domain spectrum is Fourier transformed, perform normalized spectral preprocessing on each obtained frequency-domain spectrum, and finally the food to be tested The preprocessed frequency domain spectrum of the tableted sample is divided into a calibration set sample frequency domain spectrum and a verification set sample frequency domain spectrum, and a quantitative analysis model is established with a partial least squares regression method to obtain the quantitative detection of each of the grain samples to be tested. value. The method of the invention has simple sample preparation and simple operation, and can truly and effectively realize rapid and accurate quantitative detection of amino acids in grains.

The present disclosure relates to a method, device, electronic equipment and storage medium for generating video summaries, and belongs to the technical field of the Internet. The method includes: acquiring multiple frames of images in the target video, inputting the multiple frames of images into an image clustering model, dividing the multiple frames of images into multiple categories based on the image clustering model, and selecting a target image from each category , splicing the selected target images to obtain a video summary of the target video. Based on the image clustering model, the method can quickly cluster multiple frames of images without analyzing the motion trajectory of objects in each frame, shortening the processing time and improving the efficiency of generating video summaries. Moreover, the target image is selected from each category, and the target image can represent the content of multiple frame images in this category, and the video summary generated according to multiple target images can accurately summarize the content of the target video, which improves the generated video summary. the accuracy rate.