325 results about "Rotational invariance" patented technology

The invention discloses a method for estimating target arrival angle of a multiple input multiple output (MIMO) radar, which mainly solves the problem of large signal processing capacity in the target positioning process of MIMI radar. The method comprises the following steps of: 1) obtaining a virtual array of echo of each receiving antenna by a matched filter bank; 2) constructing a data conversion matrix and a dimension reduction array according to position of transmitting and receiving array; (3) reducing dimension of the virtual array data by the dimension reduction array to obtain an effective array after dimension reduction; 4) constructing two sub-arrays by rotational variance of effective array, and deriving covariance matrix of data; 5) decomposing eigenvalue of covariance matrix to obtain signal sub-spaces corresponding to two sub-arrays; 6) deriving rotational invariant relationship matrix by least square method to obtain arrival angle of target. The dimension reduction matrix form constructed by the method has versatility; the computation quantity is reduced by the dimension reduction of data and the ESPRIT (estimating signal parameter via rotational invariance techniques) algorithm; the computation speed of MIMO radar is increased; and the real-time signal processing of the MIMO radar is made easier.

The invention belongs to the technical field of substation equipment detection and recognition. The invention discloses a MaskRCNN-based detection and identification method for a disconnecting switchof a transformer substation. The method comprises the following steps: preprocessing a monitoring image, calibrating the state and the contour of an isolating switch in the preprocessed image by utilizing a calibration tool, inputting calibrated image data into a Mask RCNN network model, training the network model, and continuously adjusting parameters at the same time, so that the network model autonomously learns a mask of the state of the isolating switch; Performing detection positioning and state identification on the isolating switch in the monitoring image by using the trained model, and outputting a predicted detection frame of the isolating switch and a foreground of a target; the trained network model has a classification capability while detecting the position of the isolation switch, i.e., can identify the "on", "off" and "intermediate" states of the isolation switch. According to the method provided by the invention, while the target pixel level identification is realized,each target is segmented, the foreground and the background of the target are well distinguished, and the robustness and the rotation invariance are extremely high.

The invention relates to a system and method for detecting the quality of a metal cap based on machine vision, used for detecting the quality defect of the metal cap and rejecting the inferior quality product. The detection system comprises an optical imaging device, an image processing device, a rejecting device and a conveying belt, wherein the optical imaging device comprises a planar array industrial camera, a first optical detection sensor and a light source; the image processing device is used for processing the image; and the rejecting device comprises a control circuit board, a second optical detection sensor, a solenoid valve and an injection tube. The method for detecting the quality of the metal cap based on machine vision comprises the following steps of: acquiring an interested maximum outline scale of a template cap image; establishing a rotational invariance characteristic template matrix for the template cap image; acquiring a round outline of the cap to be detected according to the interested maximum outline scale; calculating the rotational invariance characteristic matrix in the coverage of the round outline of the cap to be detected; and matching the rotational invariance characteristic matrix with the rotational invariance characteristic template matrix, so as to judge whether the cap to be detected is a qualified product or an inferior product.

The invention provides a method for detecting an R wave crest of an electrocardiosignal. The method takes point to point difference vector as a base characteristic. The base characteristic has translational and rotational invariance and can overcome influences of baseline drift of the electrocardiosignal. Meanwhile, logarithm polar coordinate conversion is carried out on the difference vector to measure similarity of waveforms. The measurement is sensitive to morphological characteristics of the adjacent waveforms, can capture the whole contour information of the waveforms at the same time, and has robustness on waveform swing. In addition, the method can effectively eliminate influences of interference signals by setting appropriate thresholds. The method for converting the point to point similarity measurement into measurement of waveforms of the points accurately identifies and detects the R wave crest of the electrocardiosignal. The method is applied to related electrocardiogram analytical instruments, can accurately identify the R wave crest in the electrocardiosignal, and is favorable for improving detection and analysis capabilities of electrocardiogram analytical equipment.

The invention discloses a virtual array DOA estimation method based on an L type array. The method comprises the steps that 1, based on the shift invariant property, a subarray Zx and a subarray Zy of the L type array horizontally shift to obtain a virtual array Zx' and a virtual array Zy', rotation invariance of two sub signals is formed due to the shift invariant property of the subarrays, and the virtual subarray signals are equal to the L type subarray Zx and L type subarray Zy input signals multiplied by a twiddle factor respectively; 2, the output of the four subarrays are combined to form a virtual array output signal Z (t); 3, the signal subspace and the noise subspace can be described by decomposing the features of covariance matrixes output by the array, mutual correlation processing is carried out on the array output signal Z(t) to obtain Rzz, and eigenvalue decomposition is carried out to obtain signal subspaces; 4, the twiddle factor is solved through linear operation, and the signal wave arrival direction can be obtained through the diagonal element of the twiddle factor. According to the virtual array DOA estimation method, no spectral function needs to be calculated, the phenomenon that the wave arrival direction is indirectly calculated without searching for the peak value is avoided, the complexity is lowered, the equipment complexity and cost are reduced, and the positioning precision is high.

The invention discloses a target recognition method based on an image contour characteristic. The method comprises the steps of preprocessing a template image and a to-be-identified object image for generating a binary image; establishing a characteristic database of an object template contour; extracting a whole contour of an object template binary image, equidistantly acquiring a certain number of characteristic points on the contour, and describing the object contour by means of a context characteristic of the characteristic points; performing target recognition on the to-be-identified object image; extracting the contour edge of the to-be-identified binary image; selecting a certain number of characteristic points; converting the shape direction of the to-be-identified object to the direction of the template; describing the contour of the to-be-identified image after direction changing by means of the context characteristic of the selected points; and determining similarity between the to-be-identified object and the template object through matching cost. Compared with prior art, the target recognition method has advantages of settling a problem of rotational invariance in a contour matching process, realizing invariable rotation in the contour matching process, and realizing effective application in target recognition in the image.

The invention relates to a low-sidelobe emission directional diagram design method improving DOA estimated performance of a MIMO radar. The method comprises the following steps: setting an interesting area; ensuring column vectors of a beam space weighting matrix to satisfy pairing characteristics; ensuring signals at a receiving end to satisfy rotation invariance; constructing an optimized model of a beam space weighting matrix with directional diagram coupling characteristics, rotation invariance of the signals and equality of various array element emission powers as constraint conditions; introducing an auxiliary variable and relaxing a rank-1 constraint to be a positive semidefinite constraint through utilization of the semidefinite relaxing technology; and through utilization of an interior point method, obtaining an optimal solution of the relaxing problem; through a Gaussian randomization method, the beam space weighting matrix is solved; and DOA estimation is carried out on a target at the receiving end through the ESPRIT algorithm. The MIMO radar waveform design technology which is good in performance in the invention can raise the signal to noise ratio at the receiving end, provide an important theoretical basis and a specific realization method for raising the precision of MIMO radar angle estimation.

The invention discloses a uniform wearing identification method for the power construction site staff. The method comprises the following steps: detecting whether a person enters a site or not, detecting the head and the body of the detected person and the direction of the detected person relative to a camera, and detecting whether the person wears a safety helmet or not; if the person wears the safety helmet and the face or the back of the person is opposite to the camera, carrying out safety hamlet character detection in a head area; when one side of the person is opposite to the camera, carrying out armband detection in a joint area of the head and the body; according to safety hamlet and armband detection results, judging whether the person is a staff member or not; and if the detection results show that the person is not a staff member, sending an alarm prompt. The method provided by the invention has the beneficial effects that HOG features are combined with SVM classifiers; and the LBP operator is adopted for texture analysis of safety hamlet characters and an armband, so that rotation invariance, gray invariance and other outstanding advantages are achieved, and the illumination robustness is high.

The invention discloses a high-resolution remote sensing image plane extraction method based on skeleton characteristics, comprising the following steps: selecting a remote sensing image edge detection algorithm based on embedded confidence coefficient for edge detection, and realizing the remote sensing image edge detection algorithm based on embedded confidence coefficient; vectorizing a groundfeature target edge; extracting a ground feature skeleton base line from the vector edge of a ground feature based on a constraint Delaunay triangulation network algorithm; carrying out the target main skeleton extraction algorithm based on a binary tree structure; carrying out feature analysis on the target main skeleton of the plane; and realizing the automatic identification method of a plane target. By means of the invention, the plane target can be automatically identified and extracted and better identification extraction effect is obtained. The plane target skeleton has the excellent characteristics of rotation invariance and high discrimination index with other ground features, the vector edge of the ground feature target can be efficiently and precisely extracted from a remote sensing image with high spatial resolution, and the improved target skeleton can be extracted.

The invention discloses a two-dimensional direction of arrival estimation method for a nested array based L-shaped antenna array, and belongs to the field of array signal processing. The implementation method includes the following steps: constructing a nested array based L-shaped antenna array, calculating a cross-covariance matrix by using signals received by different sub arrays of the constructed L-shaped nested array; correcting the obtained cross-covariance matrix, performing vectorization on the corrected cross-covariance matrix to generate a virtual array, constructing a plurality of equivalent covariance matrices by using the virtual array, calculating the signal azimuth angle Theta and the pitch angle Phi by using the rotation invariance between different equivalent covariance matrices, and achieving multi-target and high-precision two-dimensional direction of arrival estimation with few snapshots at a low signal-to-noise ratio. The technical problem to be solved by the invention is to realize multi-target and high-precision two-dimensional direction of arrival estimation by using fewer array elements at a low signal-to-noise ratio and with few snapshots, and to solve related engineering technical problems by using two-dimensional direction of arrival estimation results.

The invention provides an electrocardiosignal R peak detection method based on waveform characteristic matching. The method utilizes waveform characteristic matching to identify R peak of electrocardiosignal, the characteristic matching method takes difference vector between points as basic characteristic, the basic characteristic has translational invariance and rotational invariance and can overcome influence of baseline drift of electrocardiosignal signal; meanwhile, logarithm polar coordinate transformation is carried out on the difference vector and partition is carried out to measure similarity of wave form, the measurement is sensitive to the adjacent morphological characteristic, can capture global outline information of wave form and has robustness to wave form ripple; besides, influence of interference signal can be eliminated by setting proper threshold, and further accurate identification and detection on R peak of electrocardiosignal are realized. The invention is applied to related electrocardiogram analyser, accurate identification on R peak of electrocardiosignal can be realized, thus being beneficial to improving detection and analysis capability of electrocardiogram analysis equipment.

The invention discloses an adaptive ORB object tracking method. The adaptive ORB object tracking method adds FAST characteristic direction information in order to solve the problem that the FAST characteristic has no direction; a BRIEF characteristic descriptor based on pixel point binary digit comparison is adopted in a characteristic point description portion; the defects that the BRIEF descriptor is sensitive to the image noise and has no rotation invariance are improved; and compared with an SIFT algorithm, the ORB algorithm has the rotation invariance and noise resistance, and the operation speed of the ORB algorithm is higher than that of the SIFT algorithm by two orders of magnitude. The adaptive ORB object tracking method fully utilizes the characteristic of the ORB algorithm, performs further improvement on the PRB algorithm, brings forward the ORB algorithm on the basis of that the improvement, further improves the operation speed and the tracking accuracy, and better meets the demand for the real-time tracking.

The invention discloses a nuclear coordinated expression-based hyperspectral image classification method. The method comprises the following steps: carrying out feature selection by adopting a waveband selection strategy with strong operability; carrying out local binary-pattern space feature extraction on the basis of selected different feature groups; carrying out nuclear coordinated expression classification; and finally fusing classification results corresponding to the groups of features and a residual-level fusion strategy and obtaining the final high-precision classification result. According to the method, the textural features of data extracted by a LBP operator are combined, the LBP has the remarkable advantages of rotation invariance and grey level invariance, and the calculation is simple, so that the robustness of the features to be classified is further increased. Finally, a nuclear coordinated expression classifier is used for classification, the calculation efficiency is better than that of the traditional sparse manner and the nonlinear space data can be classified, so that the application range is wider and the application performance is more excellent.

The invention relates to palm print identifying technology, and provides a palm print identifying method and a palm print identifying device aiming at the defects that the preprocessing requirement for the acquired palm print images is high in the prior palm print identifying technology, and the like. The palm print identifying method comprises the step of extracting the palm print characteristics of an image to be extracted, and the extraction specifically comprises the following steps: rotating and correcting the image to be extracted to generate a corrected image; carrying out convolution operation for the corrected image according to the preset Gabor wavelet, carrying out the operation of taking an amplitude value or a phase from the corrected image after the convolution operation, and generating a response image; and carrying out local binary mode characteristic extraction for the response image to acquire a local binary mode code of the image to be extracted so as to generate the palm print characteristics. The invention also provides the palm print identifying device. The technical proposal of the invention can provide rotational invariance and has high processing speed, so the technical proposal can achieve real-time palm print identification.

The invention provides a classification method and a classification system for cancer digital pathological cell images. According to the classification method and the classification system, a suspected lesion region of interest is subjected to block processing, the suspected lesion region after block processing is subjected to feature extraction by utilizing partial matching pattern textural features, and the extracted features are classified and identified by adopting an extreme learning machine training method, so as to determine benign and malignant tumors and differentiate levels. The classification method and the classification system for the cancer digital pathological cell images utilize the partial matching pattern textural features for conducting feature extraction, analyze the textural features of cells from macroscopic and microscopic aspects, have the advantage of rotation invariance, effectively overcome the problems of diversity, irregularity and the like of cell morphology, provide reliable textural feature information for classification, apply an extreme learning machine to the classification of breast cancer cells, shorten the training time, accelerate the speed of classification and identification, and improve the accuracy of recognition.

The invention provides a license plate character identification method. A descriptor adopted in the method directly takes the geometric center point of a character image as a characteristic point, and a main direction of characters in the license plate character image is sought by a principal component analysis (PCA) algorithm so that the descriptor of the character characteristic point is generated; compared with the conventional scale invariant feature transform (SIFT) descriptor, the calculation quantity of the main direction of the characters in the license plate character image is greatly simplified, the operation efficiency of a computer is improved, the real-time requirement on a license plate identification system can be better met, and the generated descriptor simultaneously ensures the rotation invariance of the license plate character image and capabilities of resisting noise and illumination influence and has good robustness; meanwhile, the combination of a support vector machine (SVM) classification algorithm and the SIFT descriptor is adopted for character identification of the license plate character image, so that compared with other classification algorithms adopting K nearest neighbors (KNN) and the like, the method has higher identification rate on the premise that the operation complexity of the classification algorithm is not increased.

The invention relates to a scattered image based crop fruit three-dimensional point cloud extracting method. According to the invention, two-dimensional pictures of a target crop fruit in different angles generated through random photographing by a user are imported; data and characteristic extraction on the two-dimensional images are performed based on SIFT operators having scale and rotation invariant characteristics and matching is performed on data of different dimensions by utilizing an FLANN algorithm; generation of sparse point cloud is realized by utilizing information such as key points in the acquired two-dimensional images, camera parameters and the like; a complex point cloud is generated further through techniques of initial characteristic matching, sparse patch generation, condense patch acquisition through spreading and error patch filtering and the like; in order to ensure accuracy of the fruit point cloud, outliers of the fruit point cloud are removed through combinedinteractive selection and based on a method employing a filter. According to the invention, generation and denoising of a three-dimensional point cloud model of a crop fruit are realized through two-dimensional images and an objectification point cloud model is provided for processing including three-dimensional reconstruction and the like.

The invention discloses an image super-resolution reconstruction method of an improved convolution network based on data enhancement. The method comprises steps of data enhancement and network structure improvement. On one hand, variety of samples are increased by adopting the mode of multi-angle rotation and rotation of a sample set, so features of multi-angle backgrounds can be acquired and rotation invariance of the features is achieved. Sufficient feature information is beneficial to improvement of reconstruction precision of an image. On the other hand, in the network model provided by the method, features are extracted by use of a deep convolution neural network, so multiple layers of convolution layers are beneficial to more advanced extraction and more complete features. De-convolution layers are used as reconstruction layers for carrying out feature mapping on output of the convolution layers so as to recover the image resolution, so the super-resolution image is obtained. The convolution layer is lack of the rotation invariance feature, so due to the variety of the samples in the provided method, an objective of increasing parameters is achieved, so the network is well fitted, improvement of the reconstruction precision is finally achieved and an effect of accelerating convergence speed of a network training is achieved.

The invention provides a detection method for a fatigue state of a driver. Confidence coefficients are acquired through calculation by sending eye SIFT features into a classifying model; the opening-closing states of the eyes are determined according to the confidence coefficients without being affected by a human face angle; and the detection method has a rotation invariable property, high accuracy on eye detection, simple calculation process, and can detect the fatigue state of the driver in time and can satisfy real-time detection requirement. In addition, the invention also provides a detection system for the fatigue state of the driver.

The invention discloses an estimation algorithm for two-dimensional direction of arrival (DOA) of an L-shaped array by adopting time-frequency analysis. The estimation algorithm disclosed by the invention mainly aims at solving the problem of two-dimensional DOA estimation of time-frequency spectrum aliasing and spatial neighborhood information sources and is suitable for low signal to noise ratio and underdetermined conditions. The estimation algorithm is realized by the following steps: firstly, forming the L-shaped array on an xz plane by using two uniform linear arrays to construct total receiving data; secondly, transforming the total receiving data into a time-frequency domain by adopting STFT (Short Time Fourier Transform); thirdly, selecting single information source time-frequency points of all the information sources in a two-dimensional direction, and establishing a time-frequency receiving data matrix of single information sources; fourthly, constructing a time-frequency cross correlation matrix; fifthly, calculating a novel receiving matrix subjected to aperture expansion; sixthly, based on a propagation operator principle, defining a propagation operator subjected to the aperture expansion; seventhly, constructing an angle selection matrix, and calculating the two-dimensional DOA according to rotational invariance among sub-matrixes. According to the estimation algorithm disclosed by the invention, estimation precision of the time-frequency spectrum aliasing and spatial neighborhood information sources and the success rate are improved, high robustness of noises is realized, and required number of array elements can be reduced.

The invention discloses a free-hand sketch offline identification and reshaping method. The method comprises the following steps of: preprocessing an input image; converting a discrete and disordered point set in each connected domain into an ordered point sequence for compression; fitting a plurality of straight lines of the point sequence by using a dynamic programming algorithm, and determining the number of optimum fitted straight lines to obtain the strokes, represented by the straight lines, of each connected domain; analyzing a stroke result obtained by fitting the plurality of straight lines; if the number of the fitted straight lines is greater than the maximum edge number of a shape identified by a system, performing order reduction on the strokes, classifying the strokes, and calculating distance between the strokes; and selecting the nearer strokes for combination and analysis, and performing verification according to the geometrical characteristic to determine the shape formed by combination of the input strokes. The method is higher in identification rate, and the shape to be identified has scale invariance and rotation invariance; and the algorithm supports a multi-stroke form for identifying limited shapes, and the problems in the identification completely based on the geometrical characteristic are solved.

The invention discloses a grating shearing phase contrast CT imaging data acquisition and reestablishment method. The method comprises steps of utilizing a shearing technique of an analysis grating relative to a phase grating to obtain a displacement curve through measuring the displacement change of light intensity along with the analysis grating shearing, when the analysis grating is at the median point position of the displacement curve and the rotating shaft of the sample is parallel with the grid, acquiring CT projection data of a set of samples rotating by 360 DEG; selecting projection light ray data in opposite directions from the CT projection data of 360 DEG to perform a matching combination, separating the absorption projection data and refraction projection data of samples rotating by 180 DEG, reestablishing to form a parallel beam absorption projection data and refraction projection data of the samples rotating by 180 DEG; weighting the refraction projection data according to the rotational invariance to obtain multiple sets of 180 DEG parallel beam projection data satisfying the requirement of rotational invariance, and according to the Radon inverse transformation, reestablishing a three dimensional distribution of absorption coefficient, refractive index, first derivative of the refractive index and the second derivative of the refractive index.

The invention discloses a shape matching and target recognition method based on a PCA-SC algorithm. The method comprises the steps of carrying out preprocessing on a target image, filtering part of noises in the target image, extracting the edge of the target image, extracting information of boundary contour points, working out the rectangular coordinate parameters of the contour points, converting the contour points from rectangular coordinates into polar coordinates, obtaining a corresponding logarithmic polar histogram of each point to forming a local feature descriptor, forming a covariance matrix, extracting a corresponding feature vector of a larger characteristic value of the matrix, adopting a linear transformation method to drop the matrix from high dimension to low dimension, forming a new characteristic matrix, wherein the new characteristic matrix is used for the shape matching and the target recognition, calculating matching degree, and obtaining a matching degree value between the target image and each template image. According to the shape matching and target recognition method based on the PCA-SC algorithm, characteristic extracting and effective representation for the image can be achieved, scale invariance, rotation invariance and translation invariance are achieved, accuracy rate and efficiency are improved, and interference of the noise is effectively restrained.

The present invention provides a positioning method based on scene recognition. The positioning method comprises: carrying out data conversion, wherein three-dimensional laser data is converted into atwo-dimensional single-channel image; establishing a feature map, wherein the image is marked to obtain a label corresponding to the image, training is performed with a neural network training modelby using the image and the label as input to obtain feature vectors, and the feature map is established according to the feature vectors; and carrying out global positioning, wherein the current laserdata is matched with the existing feature map, and the nearest result of the matching result is selected as the current global positioning. The invention further relates to an electronic device, a storage medium, and a system. According to the present invention, the matching between the laser data and the map is completed based on the rapid searching of the k-d tree, and the two-dimensional poseof the robot is globally estimated by using the scene recognition as the observation model and fusing the mileage information and the particle filter algorithm; no change is generated during the rotation, and the orientation of the robot at this time cannot affect the result; and the method can be adapted to the gradual change of the outdoor environment, and is suitable for the long-term operationof the robot.

The invention discloses a visible light-infrared image registration method based on salient region features and edge degree, and is to solve the technical problem of poor registration precision due to low similarity of image features extracted through an existing visible light-infrared image registration method. The technical scheme is characterized by extracting a salient region from an image pair to be registered through a significant region detection method based on adaptive local contrast; describing shape features of the salient region by utilizing a Zernike rotation invariant moment to ensure translation, gray level, scale and rotation invariance of the utilized features; through RANSAC, obtaining a homonymy point pair and carrying out estimation to obtain initial transform parameters; and finally, carrying out optimization on the initial transform parameters by utilizing a three-dimensional joint histogram constructed based on edge degree to obtain a final result. The method improves image registration precision through salient region and edge degree features, the similarity of which are higher, between heterogeneous source images, and invariance of the Zernike moment.

The invention discloses a method, namely an SLS-ESPRIT (structure least square-estimating signal parameters via rotational invariance techniques) algorithm, for joint estimating angle of departure (DOD) to angle of arrival (DOA) of a target in a MIMO (multiple input multiple output) radar system on the basis of an SLS method. Estimation accuracy of signal subspace is improved by solving rotational invariance equation by the aid of the SLS and iterating estimation error of minimized signal subspace, and angle estimation performance is enhanced. In addition, effectiveness of the estimation method is testified through simulation results.

The invention discloses a user identity authentication method for an intelligent mobile terminal, by designing a special movement operation manner of the mobile terminal, active movement of the mobile terminal and screen touching operation are organically combined; by using various sensor behavior data and screen touching behavior data generated during authentication operation of a user, a characteristic vector sequence of a mapping relation between a finger screen touching track and a space pose position of the mobile terminal is established based on a screen touching geodesic distance of the user, so as to judge legality of user identity. The user identity authentication method has the advantages: the operation is simple, and no additional extra equipment need to be increased; behavior characteristics with space translation and rotation invariance are put forward, and influence of change of poses such as rotation and translation of the mobile terminal on an authentication result is effectively avoided.

The invention discloses a picture correction method and system based on an ORB algorithm. The ORB algorithm and perspective transformation are combined in use. The method comprises the specific operation steps of S1, feature point detection; S2, feature point description; S3, feature point matching; S4, preliminary screening of matching point pairs; S5, acquisition of first N optimal point pairs;S6, calculation of a homographic matrix H; and S7, image registration and correction. Compared with existing methods, the method has the remarkable advantages that operation speed is 100 times that ofa SIFT algorithm and 10 times that of a SURF algorithm, correction precision is higher than that of the SIFT algorithm and slightly lower than that of the SURF algorithm, the method has rotation invariance, and the noise sensitivity problem is solved, so that the final correction effect is certainly ideal; and during calculation of the homographic matrix H, a random sampling consensus (RANSAC) algorithm is adopted, so that the homographic matrix H is solved more robustly, and mistaken matching points are eliminated.