57results about How to "Noise immunity" patented technology

The invention discloses a method for classifying electric energy quality mixed disturbances based on multi-feature quantity of time-frequency domain. Voltage dip, voltage swell, short-term voltage interruption, impulsive transient, oscillatory transient, harmonic waves and flickering electric energy quality disturbances and mixed disturbances of a combination thereof are classified. The method for classifying the electric energy quality mixed disturbances concretely comprises the steps of: firstly, processing a disturbance signal by using an EEMD (Ensemble Empirical Mode Decomposition) and MIST (modified incomplete S-transform), and extracting nine time-frequency domain characteristic values; and then, inputting characteristic quantity to a blocked automatic classifying system to recognize the disturbances. By using the method, the mutual interference among single disturbances is fully considered and is effectively inhibited through the complementary time-frequency domain characteristic values. A simulation result shows that, under conditions of certain noises, the method can be used for effectively classifying the voltage dip, the voltage swell, the short-term voltage interruption, the impulsive transient, the oscillatory transient, the harmonic waves and the flickering electric energy quality disturbances and the mixed disturbances of the combination thereof.

The invention relates to a default performance formation controller structure for multi-mobile robots and a design method. N double-mobile robots containing unknown dynamic states are regarded as followers, the networked system formed by connecting the followers with a leader through a one-way topological graph is regarded as a controlled object, and a time-varying formation controller with default performance is designed by using the active-disturbance-rejection and inverse techniques, so that the following robots track the reference trajectory of the leader and build and keep a desired time-varying rank, and the tracking error is within a preset range. The default performance formation controller structure has the advantages that according to the characteristic that an extended state observer is independent of a precise mobile robot model, the unknown dynamic states are compensated through estimation of the extended state amount in real time, so that the designed time-varying formation controller has a disturbance rejection capacity; the derivative of a complicated nonlinear function is effectively estimated by means of a tracking differentiator; the convergence precision and rate of a formation error are improved by using a default performance function; the problems about the unknown dynamic states, complicated derivation reduction and precision control in a system can be effectively solved.

The invention relates to a multi-single-arm manipulator output consistence active disturbance rejection controller structure with pre-defined performance and a design method thereof, in particular toa multi-single-arm manipulator output consistence active disturbance rejection controller, and belongs to the technical field of industrial process control. The active disturbance rejection technologyand the inversion technique are used for designing the multi-single-arm manipulator output consistence active disturbance rejection controller, the characteristic that extension state observers do not rely on a system model is utilized, and the influence of disturbance can be estimated and compensated in real time, so that the designed controller has the anti-disturbing property; tracking differentiators are designed to simplify the phenomenon of complex signal derivation of "explosion"; a manipulator system is restrained by a driving torque, the input saturation characteristic is considered,and the amplitude limiting problem of the driving torque is solved by an auxiliary system; the convergence speed and accuracy of the output consistence error are improved by using a pre-defined performance function; and according to the designed scheme, the unknown dynamic of the system can be effectively estimated, the derivation is simplified, and the control precision is improved.

The invention provides a feature extraction and matching method in visual navigation. According to the method, a feature region is extracted by adopting a two-dimensional maximum entropy threshold segmentation method; an external square region of the feature region is subjected to image normalization processing; each external square region subjected to the image normalization processing is used as a feature region to be matched; a feature vector of a feature point is obtained; each feature point in each feature region in a first image is matched with feature points in each feature region in a second image, and the number of matching points is obtained; and two feature regions with the most matching points are selected as matching regions, wherein matched feature points are used as matching feature points. According to the scheme, the number of the matching points is obtained through feature point matching; the feature regions with the most matching points are obtained by using the number of the matching points as a screening condition; the matching of the feature regions is completed; the practical matching pair is further obtained; and the matching points of navigation images with fine, dense and single textures under the condition of generating affine transformation such as shear transformation is increased.

The invention relates to a classification and identification method for a communication signal modulating mode based on ART2A-DWNN, belonging to the field of classification and identification of communication signal modulating modes and solving the problem that single neural network has long period and low accuracy for classifying and identifying communication signals. In the method, an ART2A-E algorithm based on an ART2A network is taken as a first layer of a combined neural network, and similar modes is roughly classified by selecting relatively smaller vigilance parameters; a DWNN is directly connected with the output layer with the corresponding type of the ART2A network, Morlet mother wavelet Phi(x) with higher resolution in frequency domain and time domain are adopted, learning is carried out by utilizing error back-propagation algorithm, a synaptic weight can be modified with a conjugate gradient method till output is within the error range, and the number of modes in each typeis reduced after rough ART2-E classification, so that the DWNN can quickly converge. The invention is used for classification and identification of communication signals.

The invention discloses a remote sensing image change detecting method with the combination of multi-resolution NMF (non-negative matrix factorization) and Treelet, aiming at solving the problem that the balance is often hard to make when image details and smooth regions are considered in terms of single resolution, and also aiming at keeping both the detailed information of an image and the information of a smooth region in the image change detection. The implementation process of the method is as follows: inputting two time-phase images, and constructing a differential image and carrying out median filter by using a direct differential value; subsequently extracting images of different resolution by using a NMF algorithm; obtaining thresholds of the filtered differential image and images of different resolution respectively; combining threshold images by using a Treelet algorithm; and dividing the combined image by using a region growing method so as to obtain a final change detection result. With the adoption of the method, the problem that an adjacent region structure of the image is likely to be affected by independent noise points is solved, both the detailed information of the image and the information of the smooth region can be kept, the independent noise can be eliminated, the change detection precision is improved, and so that the method can be used in fields of disaster monitoring, land utilization, agricultural investigation and the like.

The invention provides an active contour model image segmentation method based on local Gaussian distribution fitting and local signature energy driving, and belongs to the technical field of image processing. The active contour model image segmentation method based on local Gaussian distribution fitting and local signature energy driving includes the following steps: S1, inputting an original image I(x); S2, calculating the local entropy of the image so as to obtain a local signature energy item of the image; S3, initialization level function Phi=Phi0(x); S4, initialization coefficients: alpha, beta, lambda1, lambda2, Mu, Nu, epsilon, sigma and deltat; S5, calculating local fitting energy items e1 and e2; S6, update level set function Phi; and S7, determining whether the level set evolution curve can satisfy the convergence criterion, and if not, turning to the step S5 to continue calculation until the termination condition is satisfied. The active contour model image segmentation method based on local Gaussian distribution fitting and local signature energy driving can realize segmentation of the target having nonuniform gray scale, is not sensitive to the shape, the size and the position of the initial contour curve, and has certain noise immunity.

The invention belongs to the cognitive radio communication signal modulation identification technical field, especially a continuous phase frequency shift keying (Continuous Phase Frequency Shift Keying, CPFSK) modulation identification method based on signal linear fitting instantaneous phase. According to the CPFSK modulation identification method, the feature of the CPFSK signal is used sufficiently, the feature is that the instantaneous phase of the CPFSK signal in every symbol are linearly increased or reduced; the new feature extracting method is better in noise immunity through modeling the CPFSK signal, extracting the signal instantaneous phase and combing the linear fitting method; a simulation experiment shows that the algorithm is visual in result, good in performance and relatively low in operation complexity.

The invention discloses a visual inspection method for secondary printing precision of crystalline silicon photovoltaic solar cell electrodes. Devices to implement the method mainly include an array camera, a camera fixture, a rotary positioning platform, a frame, a printing table and a control system; the control system is used for controlling the rotary positioning platform and controlling waferimage acquisition and algorithmic processing. The method herein includes: mounting the rotary positioning platform, calibrating the array camera, using the calibrated array camera to take pictures for a silicon wafer that is secondarily printed, extracting wafer grating edges from the pictures by a machine vision processing algorithm, detecting width of wafer gratings, and recognizing and locating positions of markers; calculating deviations between secondarily printed gratings and primarily printed gratings according to the grating widths and positional information of the markers detected bycameras of the array camera through the control system. The visual inspection method and system herein have the advantages of high recognition speed, good stability and the like.

The invention discloses a polarized SAR image classification method based on ACGAN, and the method comprises the steps: carrying out the Pauli decomposition of a polarized scattering matrix, and building a feature matrix based on pixel points; replacing each element in the feature matrix with an image block of a neighborhood of the element to obtain a feature matrix based on the image block; constructing a training data set by using the feature matrix based on the image blocks, and training the ACGAN network model by using the training data set to obtain a pixel-level classification result; and finally, converting the feature matrix into an RGB pseudo-color image, and dividing the image into K super-pixel areas by using an SLIC super-pixel algorithm. And a final classification result is optimized by combining the pixel-level classification result and the super-pixel block. According to the method, the polarization scattering information and the spatial neighborhood information of the polarization SAR data are fully utilized, and the generative adversarial network with the auxiliary classifier is used for competitive adversarial training, so that the classifier can extract classification features more effectively, and higher classification precision is obtained.

The invention discloses a multiple dual-connecting-rod mechanical arm inclusion controller based on output positions and a design method. N dual-connecting-rod mechanical arms in an unknown dynamic state are regarded as followers, and a networking system which is formed by connecting the N dual-connecting-rod mechanical arms with M leaders through a one-way topological graph serves as a controlledobject. The leaders constitute a static convex hull, and the inclusion controller is designed through the positions output by joints 1 and joints 2 of all the followers, so that output of the followers is converged into the static convex hull. Compared with the prior art, the multiple dual-connecting-rod mechanical arm inclusion controller and the design method have the beneficial effects that extension state observers are designed in all sub controllers through input and output of the followers to reconstitute the system state of the dual-connecting-rod mechanical arms, and through estimation of the extension state quantity, the unknown dynamic state is real-timely compensated, so that the designed output-feedback inclusion controller has disturbance resistance; and the problem of computational complexity explosion is solved, and the unknown dynamic state in the system can be effectively treated, and the complex differential operation is simplified.

The invention discloses a wavelet iterative estimation based finite difference contrast source full waveform inversion method. The method comprises the steps of S1, inputting an initial velocity model, constructing a finite difference operator and calculating a background wave field; S2, calculating an approximate seismic focus ratio, calculating a scattering field thereby, and acquiring a contrast source initial value, a contrast initial value and a total wave field initial value; S3, calculating a gradient value corresponding to the initial contrast source at each shot, and constructing a conjugate gradient direction; S4, solving the update step length of the contrast source at each shot, and performing iterative update on the contrast source; S5, judging the size of an objective function value and an allowable error at the moment so as to decide whether iteration stops or not; and S6, calculating the propagation velocity of waves in a medium according to the current contrast value. The method disclosed by the invention effectively overcomes a problem that a result acquired by inversion seriously deviates from a real model when the difference between applied wavelets and actual wavelets is great, and the calculation amount is significantly reduced.

The invention relates to a generalized S transformation-based Q-value estimation method adopting a spectral correlation coefficient. The method includes the following steps that: step 1, seismic data are inputted; step 2, generalized S transformation is performed on the seismic data, so that post-transformation seismic data can be obtained; step 3, the estimated amplitude spectrum and actual amplitude spectrum of the post-transformation seismic data are obtained; step 4, a spectral correlation coefficient is constructed based on the estimated amplitude spectrum and actual amplitude spectrum; and step 5, a Q value which appears when the spectral correlation coefficient is minimum is solved. With the method of the invention adopted, the problem of frequency band range selection can be avoided, certain noise resistance can be realized, and the influence of a time-frequency analysis method can be reduced.

The invention discloses a traffic video vehicle identification method based on an SRC and support vector machine (SVM) combined classifier. The histogram of oriented gradient (HOG) features of the vehicle samples are trained by a dictionary training (K-SVD) algorithm to obtain a redundant dictionary, so that the sparse representation-based classifier (SRC) is constructed, and meanwhile, the HOG features of the vehicle positive and negative samples and samples to be classified are subjected to sparse reconstruction, and the SVM is trained by the reconstructed features. At the end, the combined classifier is formed by the linear weighting of the SRC and the SVM based on reconstruction to complete the comprehensive decision of vehicle identification. According to the invention, the identification rate and robustness of the whole system in the complex traffic scenes of adhesion, shielding, and target category diversity, and the training time is reduced.

The invention discloses an anti-noise moving object detecting method based on frequency domain processing and belongs to the field of digital image processing. The anti-noise moving object detecting method comprises the steps that firstly, two-dimension Fourier transform projection is conducted on a collected image, so that image frequency domain information is obtained; secondly, through a low-pass filter, the low-frequency portion of a frequency spectrogram is selected and processed; finally, according to the change tendency of the low-frequency frequency spectrum of a front image and a rear image of the image, a moving target is selected out through a frequency spectrum energy function. After frequency domain change and low-frequency selection, noise information at the high frequency is effectively filtered out, so that the moving object detecting method has the anti-noise performance, and the moving object existing in the noise background can be accurately extracted.

The invention discloses a ZPW-2000 orbit shift signal decoding method based on Duffing oscillators, which comprises the steps of: firstly, attenuating an acquired ZPW-2000 orbit shift signal and then adding to a carrier coding 4-oscillator array consisting of 4 Duffing oscillators; decoding according to state change of the oscillators to obtain a carrier frequency, then carrying out spectrum shift on the orbit shit signal by using the carrier frequency and then carrying out bandpass filtering; and finally, adding a filtered signal to a low-frequency coding 18-oscillator array consisting of 18Duffing oscillators, and decoding according to state change of the oscillators to obtain a low frequency. The method disclosed by the invention has the advantages of strong reliability and easiness in hardware realization, can be used for accurately and reliably decoding ZPW-2000 and UM-71 orbit shift signals with low signal to noise ratio, provides a feasible scheme for the detection of the orbit shift signals under a high-noise environment, and has larger application value in high-speed railways and passenger special lines.

The invention discloses an indoor space geometric contour reconstruction method based on an acoustic mirror image model; a sound receiver, a phase transformation generalized cross correlation analysismodule, a sound source positioning module, a plane parameter analysis module and the acoustic mirror image model are provided; the method uses sound signals as sound source signal, and uses a movableM array element star array as the sound receiver; the method comprises the following steps: locating the sound source according to a controllable response power phase transformation and area contraction method; using the acoustic mirror image model to obtain distance estimated values between the positions of a talker making single step movement and each indoor reflection surface before and afterthe single step is moved via the cross correlation analysis of sound transmit-receive signals; finally, combining with indoor space shape priors to convert the space geometric contour reconstruction problem into the sound source positioning problem, solving the mirror image sound source position, and realizing the indoor space geometric contour reconstruction. The method can improve the indoor space geometric contour reconstruction accuracy, and considers the sound source versatility, thus obtaining robust contour reconstruction effects.

The invention discloses an improved laser point cloud registration method based on a covariance matrix, and belongs to the technical field of graphics three-dimensional reconstruction. According to the method, a key point extraction method combining tensor voting and ISS is adopted, so that the method has certain noise immunity, and meanwhile, part of boundary points can be eliminated to reduce the mismatching rate. The construction of the feature descriptor needs to fully contain neighborhood information of the point cloud, is not easily influenced by translation and rotation, and has robustness. The covariance matrix descriptor is a descriptor capable of fully describing neighborhood information, the angle quantity is not influenced by translation and rotation, and the covariance matrix descriptor is an excellent choice for constructing a feature description vector. According to the method, a covariance matrix containing angle quantity and curvature information is constructed in combination with an establishment method of a three-dimensional coordinate system in a feature histogram, a bidirectional nearest distance method is used for feature matching, and the mismatching rate is reduced.

The invention discloses a frequency domain transformation-based MEMS acoustic film surface defect detection method, which comprises the following steps of: calculating a Fourier spectrogram and a frequency spectrum gradient map of an MEMS acoustic film surface defect image through two-dimensional discrete Fourier transformation, and calculating a segmentation threshold Tg between two peak values in a superimposed gradient histogram through a maximum between-class variance method; and establishing a Boolean mask, performing morphological operation on the Boolean mask, eliminating a dominant frequency component corresponding to the periodic structure identified by gradient calculation in the frequency domain according to the Boolean mask, and performing two-dimensional inverse discrete Fourier transform on the residual frequency spectrum to obtain a reconstructed image containing a defect area, and finally, carrying out binarization segmentation by adopting a threshold method to output a binarization image containing defect information. The problem that a traditional defect detection algorithm depends on an imaging light source and a reference image is solved, and the method has certain anti-noise performance and is not interfered by factors such as image translation, rotation and zooming.

The invention discloses a low-resolution image recognition algorithm for compensating edge information. The algorithm comprises: an image recognition module which used for recognizing a low-resolution image; and an edge information generation module which is used for estimating edge information approximate to the high-resolution image according to the low-resolution image. According to the low-resolution image recognition algorithm, the edge information of the low-resolution image is predicted by using the low-resolution image, the edge information contains more lost high-frequency detail information in the low-resolution image, and the identification rate of the low-resolution image can be improved by using the predicted edge information. Meanwhile, an attention mechanism module is added into a residual block of the edge generation model, so that the extracted features can be more beneficial to generating the edge of an approximate high-resolution image, and certain anti-noise performance is achieved.

The invention proposes a particle filter sound source tracking positioning method based on Delaunay triangulation, and the method can achieve the ordered correlation of the random particles in an indoor physical space under the condition that the size of an indoor space is known through the Delaunay triangulation of the space via a smartphone carried by a speaker and a Delaunay triangulation result. Based on the theoretical analysis of a sound decay model, the method determines a particle search range nearby a mobile sound source, and a gathered estimation value of the particle search range issubstituted into a particle filter target tracking frame, and an enhanced particle filter algorithm is used for the tracking and positioning of a speaker in an indoor scene. The positioning method has certain noise resistance, reverberation resistance and robustness. The method provided by the invention is suitable for an indoor sound field environment which is shaped regularly, and is also suitable for the positioning and tracking of the speaker in an indoor sound field environment which is shaped irregularly.