258results about How to "Stable extraction" patented technology

The finding of the present invention is that the melody extraction or automatic transcription may be implemented clearly more stable and if applicable even less expensive when the assumption is considered sufficiently that the main melody is the portion of a piece of music which man perceives the loudest and the most precise. Regarding this, according to the present invention the time / spectral representation or the spectrogram of an interesting audio signal is scaled using the curves of equal volume reflecting human volume perception in order to determine the melody of the audio signal on the basis of the resulting perception-related time / spectral representation.

In order to implement the melody extraction or the automatic clearly more stable or to improve the transcription result, respectively, at the resulting segments or trajectories, respectively, of a melody line gained from a spectrogram of the audio signal a harmony mapping is performed such that a follower segment directly neighboring a reference segment in time direction is virtually shifted in the frequency direction by stages of octave, fifth and / or third in order to examine whether among the resulting lines of the octave, fifth and / or third there is one that fulfills a predetermined condition, like e.g. that the time / spectral representation along this line comprises a minimum which is larger by a certain factor than a minimum that it comprises along the reference segment line, and, if such a line exists, selects the same and actually performs the shifting of the follower segment. This way, errors in melody line determination may be corrected again. By a suitable selection of predetermined conditions for the lines of the octave, fifth and / or third, the direct neighborhood and further possible conditions, a faulty shifting may be prevented.

The invention provides a method for evaluating beautifulness of human faces based on video. The method is achieved by following steps: (1) capturing the face information by video; (2) effectively detecting the face according to the captured information by a cascade classifier based on haar-like characteristics; locating the eyes and the lips on the basis of the face detection, including detecting eyes by adopting a Sobel edge detection operator and a region growing method and determining the boundary of the lips according to the projection of red pixel dots in the face image; (4) extracting the face contour by adopting Snake model; and (5) comparing the extracted information with a face beautifulness standard library to determine the face beautifulness evaluation result. The method can achieve the real-time face beautifulness evaluation by capturing the face video information by a computer, and training to obtain the beautifulness classification standard according to a large amount of face samples.

Signals are separated by introducing a function having a monotonously increasing characteristic like an exponential type function as a cost function, and applying an adaptive algorithm that minimizes that cost function in terms of a signal separation matrix. That is, there is provided a signal processing apparatus that separates and outputs an original signal from the observed signal x(t), in which multiple multidimensional signals are mixed, wherein the nonlinear function 21 is operated on an input observed signal x(t) and an estimated separation matrix W(t-1) estimated at a previous cycle. Then, an error signal e(t) is calculated 22 based on y(t) formed by this nonlinear function 21, the estimated separation matrix W(t-1) estimated at the previous cycle, and the observed signal x(t) at that time. Then, based on the calculated error signal e(t), the update of the separation matrix W(t) at that time is performed 23 such that consideration weight is increased when estimation errors are large using the cost function having a monotonously increasing characteristic.

The invention discloses a method for visually tracking a plane abut-jointed weld beam by a linear laser aiming at the complex condition of a welding site. The method comprises the steps of: building a weld beam tracking system, establishing a coordinate system model of each part of the weld beam tracking system, calibrating a system structure by adopting a checkerboard target and a point laser; passively scanning a weld beam by a linear laser, stable and accurately positioning a laser weld beam center by adopting a weld seam center detection technology; and flexibly changing the detection distance between a weld point and an image through setting delay time by a user. According to the method, the influence on the quality of the image by arc light and spattering can be effectively avoided, small breakage of a laser bar is effectively resisted, and automatic tracking of the plane abut-jointed weld beam is stably and accurately realized on line. The method has an important significance on realizing the welding automation.

The invention discloses an infrared image and visible image registration method based on a visual attention mechanism. The infrared image and visible image registration method based on the visual attention mechanism comprises the following steps of inputting an original infrared image and an original visible image, and obtaining a visual salience region of the infrared image and a visual salience region of the visible image through a visual attention mechanism model; carrying out rough matching on the visual salience region of the infrared image and the visual salience region of the visible image through invariant moment calculation of the Hu moment; searching for the center of mass in the visual salience region of the infrared image and the center of mass in the visual salience region of the visible image after rough matching is accomplished, using the centers of mass as visual salience points, and using the visual salience points to represent the visual salience regions; carrying out fine matching on the visual salience points on the basis of the de-mean normalization cross-correlation principle; purifying accurately-matched images through an RANSAC algorithm, calculating a registration estimation function, carrying out image transformation through the registration estimation function to achieve registration and output of the images. According to the infrared image and visible image registration method based on the visual attention mechanism, due to the fact that the visual attention mechanism of human eyes is simulated, serves as an image characteristic detection method, and is used for extracting stable characteristics, after matching is carried out, registration between the infrared image and the visible image is accurately achieved.

A plug connector is for connecting to a receptacle connector. The receptacle connector comprises two resilient plates. The plug connector has an insulating housing and a metal shell. The insulating housing includes a tongue portion and a base portion for accommodating and connecting the terminals. The metal shell has a top plate, bottom plate, and two side plates. Multiple recesses, protrusions or elongated recesses are selectively formed on at least one of the top plate, the bottom plate for connecting the resilient plates. These can avoid electromagnetic interference, dusts or any external substances from attaching to the insulating house and also ensure a high quality of the insertion force and the extraction force after frequent uses.

The invention discloses a binocular vision obstacle detection system and a binocular vision obstacle detection method based on a convolutional neural network. The binocular vision obstacle detection system consists of an image acquisition module and an obstacle detection module. The image acquisition module acquires a binocular image and transmits the binocular image to the obstacle detection module to perform corresponding data processing on acquired image data, so as to obtain a precise obstacle region. The binocular vision obstacle detection method comprises the steps of: firstly, subjecting the acquired original image to median filtering processing; then correcting the binocular image according to camera parameters; designing a new convolution kernel to be used in a convolutional neural network structure for generating a precise disparity map; and finally, detecting the precise obstacle region in the image by adopting an improved V-parallax method. The binocular vision obstacle detection system and the binocular vision obstacle detection method have the advantage of being high in obstacle detection precision under conditions such as complicated light and small obstacles, and showing good robustness.

The invention discloses a data warehouse based on data fusion and data mining and an application method of the data warehouse. The data warehouse mainly solves the problems that in the prior art, a data warehouse cannot meet requirements of users for reliability, consistency and sharing performance of storage and extraction of large amounts of data information. The data warehouse comprises a data extraction layer, a data storage layer and a data access layer, wherein the data extraction layer is used for extracting data in online transaction processing system, an external data source and off-line data storage media and leading the extracted data into the data storage layer; the data storage layer comprises an ODS used for storing the data which are subject-oriented, integrated and current or close to the current and constantly change, an EDW storing enterprise-level data and a data mart; the data access layer is used for accessing data in the data storage layer in a statement, image or data analysis mode and carrying out analysis prediction. According to the data warehouse based on data fusion and data mining and the application method, the aim of storing and accessing data safely, reliably and conveniently is achieved, and the data warehouse has high practical value and popularization value.

The present invention is a method for extracting electromagnetic parameters of artificial electromagnetic material based on support vector machine (SVM). The invention relates to a new method for researching electromagnetic parameter measurement, capable of testing shield-hand material and artificial electromagnetic material having a periodic structure, and the testing result precision is high andthe production of testing samples is simple. The support vector machine (SVM) method is based on a VC-dimension theory of a statistical learning theory and a structure risk minimum principle, seeks an optimum compromise between the complexity and the learning capacity of a model based on limited sample information so as to obtain best popularization capability, and is widely applied to statistical classification and regression analysis. According to the invention, transmission and reflection coefficients of material to be tested are calculated by numerical computation methods FEM and FDTD ofelectromagnetism, and the corresponding computed result is used as training sequences to train the the support vector machine. When the support vector machine is trained fully, it is capable of calculating equivalent dielectric constant and equivalent magnetic conductance of the material to be tested by inputting testing values of the transmission and reflection coefficients.

In a code modulating method and a code modulating apparatus, a run length has an encoding rate of ⅔ which is equal to that of (1, 7) modulation, and indicates the number of “0” bits between adjacent ones of “1” bits in the channel bit train. A data bit train is converted into the channel bit train so that the run length has a minimum value 1 and a maximum value 10. Further, upon converting any data bit train, the channel bit train does not include a pattern “1010101010101” in which the run length 1 is continuously repeated six times or more. The channel bit train has a DSV (Digital Sum Value) control bit which selects the “0” bit or “1” bit in accordance with a DSV. The channel bit train obtained by using random data for the data bit train is NRZI converted into a signal. A frequency component of the signal is reduced from a maximum value of the frequency component by 20 dB or less as a power density at a frequency of {fraction (1 / 10,000)} or less of a channel clock frequency.

The invention discloses a hydraulic pump fault diagnosing method based on local mean decomposition (LMD), singular value decomposition (SVD) and information-geometric support vector machine (IG-SVM) so as to improve fault diagnosing accuracy in the case of small samples. The LMD, serving as a self-adaptive signal processing method, can decompose original vibration signals of a hydraulic pump into a limited number of signal components in a self-adaptive mode; then, the signal components are processed through the SVD, the data size of the signal components is compressed, and simpler and more stable fault feature vectors are extracted; finally, the fault states of the hydraulic pump are classified through the IG-SVM, and the fault diagnosing accuracy in the case of the small samples is improved. By the adoption of the fault diagnosing method based on the LMD-SVD-IG-SVM, a complete and effective fault diagnosing scheme in the case of the small samples is provided for the hydraulic pump, and the method has good actual engineering application value.

An image processing system including: a CCD which outputs an image signal; an edge extraction unit which extracts an edge signal from the image signal; a false edge estimation unit which estimates an edge signal that arises due to noise components based upon the image signal; an edge correction unit which corrects the edge signal by performing coring processing based upon the edge signal that arises due to noise components; and an edge enhancement unit which performs enhancement processing on the image signal based upon the edge signal thus corrected.

The invention discloses an automatic identification method and system for a vehicle logo. The automatic identification system for the vehicle logo comprises an offline training subsystem and an online identification subsystem. Features of the vehicle logo are extracted by adopting dense scale invariant feature transform (dense-SIFT), and abundant and stable features which are identifiable and discriminative can be extracted from a low-resolution vehicle logo picture; clustering analysis is performed on the dense-SIFT features, and a visual word bank and a K-dimension (KD) tree index are established so as to accelerate computation speed; according to the relevance between the dense-SIFT and visual words, the dense-SIFT is mapped into all visual words for representation to enhance feature descriptiveness; when a visual word histogram is established, a rectangular-ambulatory-plane spatial pyramid is proposed to increase the spatial position information of the features ,and thus, discriminating effect is promoted; and by adopting a support vector machine (SVM) training vehicle logo classifier, vehicle logo identification is realized. The automatic identification system for the vehicle logo disclosed by the invention still has high identification rate on the low-resolution picture, has high robustness on blocking, illumination, weather and shooting angle changes and the like, is short in computation time and has real-time performance.

The invention discloses a cartoon animation producing method based on video extraction and reuse. The invention comprises the following steps of: firstly dividing a cartoon video into cartoon frames; next, extracting cartoon roles via the foreground dividing technique based on the boundary; then, calculating the distance between any two frame roles and obtaining a distance matrix via using the role similarity on the boundary; then constructing an undirected graph via using the manifold learning arithmetic according to the obtained distance matrix; finally, appointing any two points in the undirected graph and establishing the shortest path between the two points which comprises the frames to form a smooth cartoon frame sequence. The method has the advantages of: effectively extracting the cartoon roles from the complex background via designing a foreground extraction technique based on boundary; making the similarity between roles more accurately by calculating the movement attribute of the roles in every frame via optical flow tracking technique; and constructing a cartoon data reuse platform to facilitate a user to produce cartoon animations conveniently and rapidly via manifold learning.

The invention relates to a mixed brain white matter nerve fiber automatic cluster and a marking method. The marking method is characterized in that: firstly, using a high dimensional elastic registration method to match a standard brain atlas to a single brain space, thereby being capable of dividing a single brain into 104 anatomical regions with structural information; secondly, using differentnerve fiber bundles to extract and mark 15 nerve fiber bundles according to the characteristics of different brain anatomical regions; thirdly, to other 4 nerve fiber bundles which can not be extracted by the above method due to strong winding between the 4 nerve fiber bundles and the nerve fiber bundles at the periphery thereof, using a method combing Kernel-PCA and Fuzzyc-mean and based on similarity measurement to automatically carry out clustering analysis; and at last, using an identification method based on characteristics to mark the result of clustering analysis. By the above method, 19 main white matter nerve fiber bundles of the brain can be extracted and marked automatically.

The invention discloses a method for enhancing and extracting high stars and other sliding fixed stars in the view field of a telescope. Aiming at the problems that high background stars and other sliding fixed stars are low in detection rate and not high in positioning precision in the celestial positioning process, the astronomy and an image processing technique are combined, the sliding tracks of the fixed stars are calculated by using astronomical information, and reliable enhancement and high-precision extraction of the high stars and other sliding fixed stars are realized by using a background equalization method, a quasi-randomized orbital determination method, a noise normalization method, a double-threshold orbital determination correlation method, a two-step centroid extraction method and other image processing techniques, so that technical support is provided for the celestial positioning. The method has the benefits that the false detection rate and the misdetection rate are extremely low; the influence from interference factors among fixed stars, of other celestial bodies and background noise and the like can be better overcome; weak high stars and other fixed stars are extracted; information about the number, the size, the position, the precision and the like of the fixed stars are synchronously obtained; besides, fewer parameters are set and the set parameters are simple, and real-time processing can be realized.

The invention relates to a method for measuring nicotine content in gum base type chewing tobacco. The method comprises the following steps: (1) extracting a gum base type chewing tobacco sample by using alkali liquor and a mixture which cannot be dissolved with the alkali liquor, thereby obtaining coarse extracting solution containing nicotine; (2) standing and layering the coarse extracting solution containing nicotine, and collecting an organic phase layer containing nicotine; (3) filtering the obtained organic phase layer containing nicotine by using an organic phase filter, thus obtaining nicotine-containing extract liquor; and (4) detecting the nicotine-containing extracting solution by using a gas chromatograph or a gas chromatograph-mass spectrometer, and obtaining the nicotine content in the gum base type chewing tobacco by adopting an internal standard method. According to the measuring method disclosed by the invention, the nicotine content is accurately measured, and the method is simple, convenient to operate and has a high popularization value.

The invention discloses an automatic fast mine monitoring image stitching method comprising the following steps: first, a Harris algorithm is adopted to initially obtain feature points at a single scale, an SIFT algorithm is adopted to precisely locate the feature points, and SIFT descriptor information is utilized on the basis so as to stably and quickly extract image feature points; second, a locality sensitive Hashing (LSH) algorithm is adopted to carry out feature matching search and solve the problem that the time efficiency of a Kd tree search algorithm in a high-dimensional data set is not high; third, an improved RANSAC algorithm is adopted to screen matching points and calculate a transform matrix so as to improve the accuracy of matching point screening; and finally, stitching is carried out by an image fusion algorithm. Experimental results show the method can be applied to automatic coal mine underground video monitoring image stitching, and has scientific reference value to image stitching in other complex environments.

The invention discloses a supervised multimodal brain image fusion method comprising the following steps: S1, calculating the characteristic of each modal; S2, matrixing and normalizing the characteristics of the modals; S3, reducing the dimension of the modal characteristics using a singular value decomposition algorithm; S4, based on the dimension-reduced modal characteristics obtained in S3, maximizing the sum of squares of correlation between the typical variables of the modals and between the typical variables and prior information, and performing an iterative cycle until convergence; and S5, connecting the modal components obtained in S4 in series, and calculating out an independent component and a mixed matrix of each modal significantly associated with the prior information using a joint independent component analysis algorithm, thus realizing supervised multimodal brain image fusion. According to the embodiment of the invention, the method has good robustness and can reveal the physiological and pathological mechanism of complex brain disease cognitive impairment.

Disclosed is a method for visible serial segmentation of human body slice images based on skeleton angular points. The method comprises the steps of loading a serial colored slice data set and a color similarity threshold value; developing a seed point set, selecting multiple seed points manually, and storing the seed points in the seed point set; calculating the color similarity between each seed point and a neighboring pixel, storing a mark if the color similarity is larger than the similarity threshold value, discarding the mark otherwise, and taking the obtained seed point set as the target binary image of the current image; ending the process if the current image is the last one; conducting skeleton extraction on the target binary image to obtain a target skeleton image, and emptying the seed point set; acquiring multiple angular points of the skeleton image with the angular point detection method to serve as the seed point of the next image, and storing the seed points in the seed point set; traversing the image sequence to achieve continuous segmentation until all the images to be segmented are segmented. By means of the method, a target human organ can be obtained from a three-dimensional colored slice image data set continuously and automatically through segmentation.

It is possible to compatibly set multiple “dropout” color ranges and “non-dropout” color ranges and uniquely determine a dropout boundary. An object of the present invention is to greatly conserve maintenance cost of adding a new dropout form after apparatus operations. A conventional technology aims at assuring relation to a predetermined color region, determining the presence or absence of contention, or uniquely settling a dropout boundary. The present invention provides a means for supplying levels to a “dropout” color range and a “non-dropout” color range. A registered color range histogram can be quasi-three-dimensionally visualized so that an operator can make adjustment by viewing a contention determination result and an image.

The invention discloses a nucleic acid protective liquid and an application thereof for long-term storage and transportation of a tissue sample at constant temperature. The nucleic acid protective liquid comprises the following components: D-glucose, 2-[4-(2-hydroxyethyl)-1-piperazinyl]ethanesulfonic acid, inorganic salt, amino acids, choline chloride, folic acid, nicotinamide, inositol, vitamin A, vitamin H, vitamin E, Holo-human transferrin, bovine serum albumin, N-(2)-L-alanyl-L-glutamine, insulin, adrenalone, linoleic acid, linolenic acid, progesterone, penicillin, streptomycin, amphotericin B and water. The nucleic acid protective liquid is low in cost. By using the nucleic acid protective liquid, nucleic acid in the tissue sample is long in storage time at room temperature, and the obtained nucleic acid is high in concentration and great in total amount. A preparation method for the nucleic acid protective liquid is simple and fast.

Upon data recording, a data recording unit inserts revise bytes as a predetermined specific code train into at least two or more portions including the head and last portions of data and records the data onto a medium. Upon data reproduction, a data reproducing unit separates a head reproduced signal by using clocks and, thereafter, executes a clock extraction and an amplitude correction by using a signal corresponding to the revise bytes as a specific code train. In principle, an RLL code for the clock extraction and gain tracking is eliminated and, in place of the RLL code, the revise bytes comprising the specific code train are inserted into the data and the data is recorded onto the medium.

The invention relates to a wrist point and arm point extraction method based on a depth camera, applied to the fields of virtual reality and augmented reality. Segmenting hand areas according to color images and depth maps collected by a Kinect depth camera, and palm point three-dimensional positions provided by an OpenNI / NITE (open type natural interactive API); extracting accurate hand areas by a Bayesian skin color model, and extracting wrist points and arm points in a vision method; and designing a plurality of man-machine interaction gestures based on the extracted wrist points and arm points and palm points provided by an NITE bank. Since the Kinect depth camera comes out, many researchers have utilized the Kinect depth camera to do research on gesture identification, however, researches based on wrists and arms are very few, and methods of extracting the wrist points and arm points based on a vision method are rare. The wrist point and arm point extraction method of the invention has a small calculated quantity, is simple to operate, and can timely, stably and accurately extract the wrist points and arm points.

In order to provide an optical modulator capable of suppressing generation of mode mismatching light in the Y-multiplexer in the MZ-type waveguide or mixing of the mode mismatching light with the radiation-mode light or the output light, and separately extracting output light and radiation-mode light, there is provided an optical modulator having a Mach-Zehnder type waveguide on a surface of a dielectric substrate, wherein a waveguide after multiplexing in a Y-multiplexer in an output side of the Mach-Zehnder type waveguide is a multiple mode waveguide 2, a subsidiary output waveguide as a high-order mode waveguide 2 is connected to a portion where the multiple mode waveguide is changed to a main output waveguide 3 as a single mode waveguide, and the multiple mode waveguide 2 has a length equal to or longer than 150 μm.

The invention provides an image classification identifying and judging method. The method includes the steps of firstly, acquiring an image, preprocessing the image, denoising the preprocessed image by a top-hat transformation method and enhancing the denoised image by means of histogram equalization; secondly, subjecting the enhanced image to feature extraction to obtain a joint histogram of an N-dimensional vector, and subjecting the N-dimensional vector to dimensionality reduction, wherein the N is a positive integer; thirdly, training the N-dimensional vector subjected to dimensionality reduction by a training set to summarize a classification set, and extracting a feature image by means of classifying or screening.

The invention discloses an SAR image river channel extracting method based on a minimum circumscribed rectangle window river channel segmentation model. The method comprises the steps that firstly, a histogram equalization method is adopted in an original SAR image to stretch the gray scale contrast ratio; then, a river channel is separated into scattered and even linear stripe-shaped areas, a minimum circumscribed rectangle window is adopted to tightly wrap each segmentation area of the river channel, and approximation of a complex river channel outline type is achieved through the rectangle windows. Threshold functions which are generated by converting a river channel model and used for judging zone boundary confidence coefficients in a graph model are used for optimizing the segmentation result so as to enable the segmentation result to be more suitable for river channel extracting tasks on this basis. Finally, according to the prior experience of the river channel outline, according to the river channel area identification method, the length-width ratio of the rectangle windows is considered, the communication between the rectangle windows is also considered, and the effective identification on the river channel area is achieved. The method can be stably and reliably used for extracting an SAR image river channel, and is particularly suitable for river channel extracting under the complex background or the extremely complex river channel outline.

The invention discloses an extracting method of sisal hemp pectin. The extracting method comprises the following steps: 1, dividing sisal hemp residues into two parts, adding water into the lesser part of the sisal hemp residues, boiling to obtain a first mixed solution, and adding water into the other part of the sisal hemp residues to obtain a second mixed solution; 2, adding galactose into the first mixed solution, inoculating a beer yeast, fermenting to obtain fermentation liquor, adding the fermentation liquor into the second mixed solution, fermenting to obtain secondary fermentation liquor, supplementing galactose during fermentation, cooling the secondary fermentation liquor to obtain a mixed solution, filtering, collecting filtrate, and concentrating the filtrate to obtain a concentrated solution; 3, adding ethanol into the concentrated solution, then performing centrifugal separation, and collecting a precipitate; and 4, air-drying the precipitate, then drying, and crushing to obtain the sisal hemp pectin. According to the extracting method disclosed by the invention, pectin is extracted by adopting a microbiological fermentation method, the extracted pectin is relatively high in molecular mass and high in extraction rate, does not need heat and acid treatments, and is low in consumption and pollution and higher in galacturonic acid content.