315 results about "Space transformation" patented technology

The disclosure discloses a method and an apparatus for combining panoramic image. The method includes: obtaining multiple original images of the same scene, performing folding change and coordinates transformation to the multiple original images, and determining an overlapping area of the multiple original images; establishing a mathematical model of the multiple original images, aligning the overlapping area of the multiple original images, and transforming the multiple original images to a coordinate system of a reference image; obtaining the space transformation relationship among / between the multiple original images according to the coordinate system of the reference image, selecting an appropriate image combining strategy, and completing the combining of the images. The solution can realize obtaining scene picture with large field of view without reducing image resolution.

A color space transformation matrix calculating system is provided that optimizes a color space transformation matrix. The matrix is obtained as a product of a first and a second matrix and transforms colors in a first color space to colors in a second color space. The system comprises first and second optimizers that calculate elements of the first matrix and second matrix by multiple linear regression analysis. The input colors which correspond to color patches and hue corrected colors obtained by using the first matrix and the input colors are set as explanatory variables. First and second goal colors respectively relating to hue and saturation in a second color space, and which correspond to the color patches, are set as criterion variables. The elements of matrices are set as partial regression coefficients.

A technology of determining obstacles around a vehicle through utilizing bird's-eye-view images; wherein, a plurality of image fetching devices disposed in various positions of said vehicle fetch a plurality of images around said vehicle, said images of two adjacent regions contain at least an overlapped region; an image processor transforms said images into said respective independent bird's-eye-view images; and an obstacle detection unit compares said overlapped region in said independent bird's-eye-view images of two adjacent regions, so as to obtain their correlations, and existence of said obstacle is determined based on said correlations. Moreover, a correspondence table is set up containing a set of space transformation information based on vehicle driving condition information. Therefore, a surrounding bird's-eye-view image of an appropriate visual angle can be produced quickly and a position of said obstacle is marked on said surrounding bird's-eye-view image.

Reconstructing a second component signal relating to a second component of a multi-component picture from a spatially corresponding portion of a reconstructed first component signal and a correction signal derived from a data stream for the second component promises increased coding efficiency over a broader range of multi-component picture content. By including the spatially corresponding portion of the reconstructed first component signal into the reconstruction of the second component signal, any remaining inter-component redundancies / correlations present such as still present despite a possibly a priori performed component space transformation, or present because of having been introduced by such a priori performed component space transformation, for example, may readily be removed by way of the inter-component redundancy / correlation reduction of the second component signal.

The invention discloses a method for correcting distortion in real time based on a GPU camera and video camera. The method adopts digital image processing technology to calculate a system of nonlinear equations consisting of two radial distortion coefficients, two centrifugal distortion coefficients and two thin prism distortion coefficients by a least square method in a correction algorithm based on a planar grid model so as to obtain a space transformation equation, and carries out synchronous spatial alternation on pixel spots of an image shot by an imaging system so as to obtain an image after distortion correction. The method solves the problem of optical distortion, and improves the correction precision to sub-pixel grade; through adopting the GPU technology with parallel processing capability, the processing velocity is improved by over 40 times, realizing real-time distortion correction (over 30 frame / second) of an 1920*1080 high-definition image, and reducing the processing cost for lens optical design by 20 to 70 percent.

The invention discloses a method for optimizing ultrasonic probe imaging plane space position calibration and belongs to the field of medical ultrasonic three-dimensional imaging and ultrasonic image fusion. The method is an optimizing method for calibrating the transformation relation of the space position of an ultrasonic probe imaging plane and the space position of a space position sensor, namely the space transformation matrix. Under the assisting of a three-dimensional positioning system, the method includes steps of: 1, constructing a three-dimensional positioning system; 2, manufacturing an N-line model; 3, building a coordinate system set of the positioning system; 4, calibrating a placing space position of a water tank model; 5, collecting ultrasonic images under different positions; extracting N-line marks and calculating the ultrasonic plane position; 7, mapping N-line mark points on the ultrasonic images onto a three-dimensional space plane; 8, performing registering on three-dimensional coordinate points on the imaging plane and an N line on the water tank model; 9, performing optimization calculation; and 10, calculating the optimum transformation between an imaging plane and the position sensor. The method has the advantages of reducing errors in space position calibration of the imaging plane.

The invention provides license plate recognition method, device and electronic equipment, and relates to the technical field of image processing, wherein the license plate recognition method comprisesthe following steps: acquiring a to-be-recognized license plate picture; extracting the preliminary characteristics of the license plate picture by using a pre-constructed convolutional neural network; inputting the preliminary characteristics into a space transformation network to obtain the transformed characteristics; inputting the transformed characteristics into a recurrent neural network togenerate an output vector sequence; and analyzing the output vector sequence to obtain a recognition result of the license plate picture. According to the method, on the basis of extracting characteristics by using the convolutional neural network, a transformation matrix is learned in combination with the space transformation network, the characteristics are further extracted, and the result ispredicted through the recurrent neural network; and thus, the prediction accuracy is greatly increased, and the license plate recognition ability is enhanced.

The invention discloses a cross-modal medical image registration method. The method comprises the following steps: providing a training set comprising a floating image of a first modal and a referenceimage of a second modal; inputting the floating image into an image conversion network, and converting the floating image into a conversion image of a second modal; inputting the floating image and the reference image into a cross-modal flow sub-network to output a first deformation field; inputting the converted image and the reference image into the single-mode flow sub-network to output a second deformation field; inputting the first deformation field and the second deformation field into a deformation field fusion network to output a final deformation field; inputting the floating image and the final deformation field into a space transformation network to obtain a floating image after distortion transformation of the final deformation field; obtaining a first total loss function according to the transformed floating image and the reference image, and performing supervised training on the network by taking minimization of the first total loss function as a target; and inputting ato-be-registered image into the trained network to obtain a registered image. The cross-modal medical image registration method can greatly improve the cross-modal medical image registration effect.

The invention relates to a system for measuring the height of a human body in a video image based on camera calibration and further discloses a method for measuring the height of a human body in a video image based on camera calibration. The system comprises a calibration point selecting module, a camera parameter calculating module, a body height measuring end point selecting module and a human body height calculating module. By the adoption of the system and the method, the height of a person erectly standing on the ground can be measured only by relying on monitored video image data after camera parameters and corresponding space transformation matrixes of the parameters are calculated. The height calculating process is simplified. The camera parameters are calculated according to the correspondence relation of image space coordinates and actual space coordinates of the same position point. The height of the person erectly standing on the ground is calculated by combining with the camera parameters and foot bottom point and head top point measurement end point information and utilizing a height calculating algorithm.

An image processing method for producing and processing image data based on input data which image data are sent to an image forming apparatus for forming an image on a paper form on which apparatus a recording head for ejecting a liquid drop of recording liquid is mounted and which apparatus is capable of performing double face printing is provided, wherein color space transformation processes different between single face printing and double face printing are performed when input data are transformed into color space values for the image forming apparatus. An image processing method for producing and processing image data based on input data which image data are sent to an image forming apparatus for forming an image on a paper form on which apparatus a recording head for ejecting a liquid drop of recording liquid is mounted and which apparatus is capable of performing double face printing is provided, wherein input tone transformation processes and adhering recording liquid quantity limiting processes different between a case where double face printing is performed and a case where single face printing is performed are performed.

The invention provides a low-power consumption portable real-time image target detecting and tracking system and a method thereof, which can realize automatic cell cultivation process control and online observation in space and equivalent harsh environments. The system comprises a PL (Programmable Logic) subsystem and a PS (Processing System) subsystem; an image acquisition unit comprises a camera interface module, an image pre-processing module and an AXI-DMA module; the camera interface module is used for reading camera data and controlling parameters of an image resolution, an acquisition frame rat and a data format; the image pre-processing module is used for carrying out pre-processing operation such as color correction, histogram equalization and space transformation on an acquired image flow; the AXI-DMA module is used for realizing self return from image data to DDR3 RAM via an AXI-HP interface; an image acceleration unit comprises a reconfigurable image accelerator and the AXI-DMA module; the reconfigurable image accelerator is used for dynamically loading an image target detecting or image target tracking accelerator according to a system working stage; and the AXI-DMA module is used for realizing self return from image data to DDR3 RAM via an AXI-ACP interface.

The present invention relates to an electric taxi charging station planning method based on an adaptive quantum genetic algorithm. The method comprises the steps of (1) initializing basic data needed by planning, and estimating the value range of a charging station number, (2) randomly generating a chromosome corresponding formula, and forming a population, (3) carrying out solution space transformation and mapping a randomly generated angular space to a charging station address coordinate space, (4) using a Voronoi diagram to divide the service range of a charging station corresponding to chromosome, (5) calculating the average waiting time in queue of each charging station based on the charging requirement in the service range, (6) planning the expression of total society whole year total cost target function minimization according to electric taxi charging stations, (7) carrying out a quantum rotation gate updating operation and a quantum bit variation operation, (8) returning to the step (3) to carry out loop calculation until a convergence condition is satisfied. The method has the advantages that the optimized layout of charging station site is realized, the influence on the planning by taxi distribution, passenger demand distribution, and a road network structure can be reflected, and the method is effective and practical.

The invention discloses a method for automatically recognizing a target at medium and low altitudes and positioning a carrier with high accuracy and belongs to the field of scene matching guidance. The method comprises the following steps of: extracting a constant algorithm and establishing an H model based on an image scale characteristic to match and recognize targets and provide a high-accuracy navigation information source; coordinating and planning relative parameters according to a matching result which is fed back in a self-adaptive mode to further control matching performance; researching integral image-based average filtering which generates an image multi-scale space instead of Gauss convolutional filtering so as to increase a matching speed; establishing an image space transformation model which adapts to the strong matching capability of the algorithm based on the thought of random sampling consistency according to an image matching result; and eliminating accurate positioning pixels of wrongly matched points based on the model to recognize the target and provide the high-accuracy navigation information source through model mapping, and simultaneously, selecting a digraph of known transformation parameters to verify the performance and the result of an image matching algorithm according to the transformation model.

The invention provides a medical image sequence plaque stability index-based quick calculation method and system. The system comprises an image collection and receiving module, an image processing module, a plaque stability calculation module and a result visualization module. The image collection and receiving module is used for collecting, receiving and transmitting a dynamic two-dimensional vascular image sequence; the image processing module is used for performing local feature or global image registration based on dynamic information of artery real-time deformation in a two-dimensional image to obtain a displacement field function of space transformation; and the plaque stability calculation module performs calculation by utilizing the displacement field function to obtain a time-varying vascular diameter sequence, a lumen central line, a contour deformation parameter and a mechanics index. Based on an existing medical image function, information such as the time-varying vasculardiameter sequence, a lumen contour, area strain and stress, and the like is added to the two-dimensional image directly, so that the in-vivo assessment of plaque stability is simplified and effectively realized.

The invention relates to a SIFT and Otsu matching based colored eyeground image splicing method. The method is realized by that (1) a green channel of a reference eyeground image and a target eyeground image is extracted; (2) the green channel is preprocessed via contract limited adaptive histogram equalization and two-dimensional Gaussian filtering to enhance the contract between vessels and tissue background; (3) SIFT characteristics of the image are extracted; (4) characteristic points of the two images are matched initially via a nearest neighbor and a next nearest neighbor algorithm; (5)an Otsu algorithm is used to screen the initial matching points to obtain correct matching characteristic point pairs; (6) a random sampling consistency algorithm is used to calculate an affine transformation matrix to carry out space transformation o the images and complete registration; and (7) the registering images are fused in a maximum value algorithm, and splicing is completed. According tothe invention, SIFT is combined with the Otsu method, the disadvantage of non-ideal registering effect caused by mismatching points is overcome, and the method has significance in the field of diagnosing diabetes and pathology degree thereof.

The invention discloses a fingerprint detection classification method based on the space transformation convolutional neural network. The fingerprint detection classification method comprises a fingerprint image extraction region of interest preprocessing, image high-frequency region extraction, image space transformation processing and convolution neural network classification training and testing. The fingerprint image extraction region of interest preprocessing removes a blank region through extracting a fingerprint part in an image; the high-frequency region extraction means the high-frequency characteristic of the image is extracted through a gaussian high-pass filter; as for the image space transformation processing, the space transformation neural network is used for carrying out translation, cutting and rotating operation on the input image, so that expansion of image data is achieved; the convolution neural network adopts multi-layer convolution pooling, convolution kernels with different sizes are used for extracting image features, and a good classification detection effect is obtained on the test set. The invention provides a fingerprint detection method which is low incost, high in detection precision and short in time consumption.

The invention discloses a multi-target locating and tracking video monitoring method, wherein the locating and tracking video monitoring method is suitable for the moving target where camera field of view is fixed, and there are small dynamic changes at background. Before the system works, firstly the background is trained by Gaussian mixture model, and then the background subtraction method is used to get the foreground image in the next frame of the video, followed by the accurate foreground object area, that is moving target location, is obtained according to the expansion and median filtering; detecting if there is any moving target,if there is moving object, the target location is determined according to the connected domain search, if there is no moving target, then the search turns to the next video frame; the tone space transformation of the target position region image is carried out, and the NTSC space tone diagram I and the HSV space tone diagram H are weighted to get the tone histogram, and the back projection of the image is further obtained, then Meanshift algorithm is used to precisely locate the target position. For the next video frame, the above calculation is repeated. The monitoring method can be used for moving target trajectory analysis, vehicle detection and traffic violations speeding and pedestrian flow detection.

The invention belongs to the technical field of traffic information processing and discloses a traffic sign recognition method and system based on a deep space network. The method includes: extractingimage information with space invariance by using a space transformation network; by improving the VGG network, fusing the features extracted from each stage of the network layer, using a spatial pyramid method to process the features of the fused image, and obtaining more features of the traffic sign images. A spatial transformation network is added to make up for the deficiency that the CNN cannot effectively extract the image information with spatial invariance. The VGG network is improved, fusion of the features extracted from each stage of network layer is carried out, the spatial pyramid method is utilized to process the fused image features, and more information of traffic signs image features are obtained. The experimental results show that the accuracy of traffic sign recognitionis improved on the basis of enhancing the network features.

The invention relates to a uniformity correcting method for separating brightness and chromaticity of an LED display screen, which comprises the following steps of: detecting the chromaticity parameter of any pixel in each module in the LED display screen as a chromaticity coordinate of the module of the pixel; collecting brightness parameters of all display pixels; determining reference chromaticity coordinates and reference brightness of each reference color of the display screen; separating the parameters of the brightness and the chromaticity; forming a brightness correcting coefficient table aiming at any pixel in each module; forming a module chromaticity space transformation parameter matrix for any one module; obtaining corrected real-time driving display data by using the brightness correcting coefficient table and the module chromaticity space transformation parameter matrix; and driving the display screen by the real-time driving display data. By the method, the uniformity correcting problem of the chromaticity of the LED display screen is effectively solved, and the collecting efficiency of the chromaticity parameters is high.

The invention discloses facial expression synthetic arithmetic based on a hybrid transformation strategy. Transition from motion capture data space to target face model motion space is achieved by building a redirecting model of radial basis function facial expressions based on geodesic distance. On a stage of facial expression animation, the result of space transformation is utilized, and local motion of a peak is calculated through a weighted local transformation method based on adjacent feature points. Simultaneously, the whole situation displacement of the peak is calculated through a whole situation transformation method based on radial basis function interpolation. In the end, the final displacement of the peak is obtained by mixing a local displacement and the whole situation displacement. According to the facial expression synthetic arithmetic based on the hybrid transformation strategy, the same catching order can be applied to different face models, model transformation is easy, and simultaneously, different catching orders can be used on the same target model, reusing of motion capture data is achieved, and simultaneously strong animation third dimension is achieved. The algorithm flow chart of the facial expression synthetic arithmetic based on the hybrid transformation strategy is shown in an attached map 1.

The invention discloses a workpiece attitude adjustment method based on a measuring point and adaptive differential evolution algorithm. The method includes the steps of initial matching and accuratematching. Firstly, reference points are selected on a theoretical workpiece curved surface and an actual curved surface to build corresponding local coordinate systems, and initial matching is realized by rough alignment of the local coordinate systems. In accurate matching, a target function based on a least-square distance is built through a point set after initial matching, an optimal solutionenables a value of the target function to be minimum and is searched by the adaptive differential evolution algorithm, so that an optimal accurate matching matrix is acquired. By the two steps, a final space transformation matrix can be acquired, adjustment amount corresponding to a machine tool can be calculated by combining a structure of the machine tool, the machine tool is adjusted, so that aworkpiece attitude is adjusted, a workpiece is accurately positioned, and allowances are uniformly distributed. According to the method, attitude adjustment and accurate positioning of testing partsmachined by an ISO (international standardization organization) five-axis numerical control machine tool are achieved, and effectiveness of the method is verified.

The invention discloses an orthogonal plane calibrating method and an orthogonal plane calibrating system of relation between a laser distance measuring device and an end of a mechanical arm. The laser distance measuring device is fixedly disposed on the end of the mechanical arm, and the calibrating planes, which are orthogonal to each other, are disposed in the space. Under the driving of the rotating translational motion of the mechanical arm, the posture of the laser distance measuring device can be changed, and the distance d measured by the laser distance measuring device can be used to solve the translational vector r and the unit direction vector v of the laser beam, and the calibrating of the space transformation relation between the laser distance measuring device and the end of the mechanical arm can be completed. The reconstruction of the point x of the object in the three-dimensional object can be realized according to the x=r+dv. The transformation relation between the coordinate system of the end of the mechanical arm and the coordinate system of the laser distance measuring device can be determined clearly. By using the calibrating plane as the orthogonal plane and adopting the non-linear optimization method, the coplanarity of the points can be fully used as the constraint condition to solve the calibrating problem of the space transformation relation, and therefore the constraint condition can be added, the errors can be reduced, the noise interference can be reduced, the robustness and the easy implementation of the method can be improved, and the calibrating precision of the space transformation relation can be improved.

The invention discloses a speaker identifying method and a system. The speaker identifying method comprises the following steps of: establishing a universal background model; establishing a to-be-identified speaker module; and using a training voice signal of the speaker to identify the speaker. Compared with the prior art, the invention has the following advantages that the high-performance speaker identifying system is disclosed by combining model space transformation and characteristic space transformation, and personal pronunciation features of the speaker are comprehensively reflected by the transformation of the two spaces. The transformation of the two spaces is calculated by using a self-adaptive algorithm based on the universal background model, so that good stability is achieved. Compared with the way of identifying the speaker by independently adopting model space transformation in the prior art, the identification rate of the system is greatly improved. Meanwhile, the system is more stable, and not easy to imitate.

The invented method processes remote sensing image through following steps. Based on IHS transform of multiple spectrum images, using statistic characteristics of remote sensing image, lowpass filtering is carried out for I-component of multiple spectrum images. Meanwhile, highpass filtering is carried out for panchromatic remote sensing image with high spatial resolution. Through histogram matching in local window, gray scale mapping is carried out one pixel by one pixel for value of pixel in panchromatic image with high resolution according to position of pixel. Thus, syncretic I-component is obtained. Then, IHS inverse transform is carried out so as to obtain syncretic result. The invention possesses advantages of retaining spectrum information of multiple spectrum images, raising spatial resolution, accurate decrypting features of ground object so that the invention is more suitable to each application and easy of interpretation.

The present invention discloses a rendering method of an ambient occlusion image, a method for generating an effect image and an apparatus. The rendering method includes the following steps that: a rendering model located in a camera space is obtained; each pixel of the rendering model in the camera space is projected to a screen space through perspective transformation; the ambient occlusion value of each pixel in the screen space is calculated; and the ambient occlusion image is generated according to the ambient occlusion value of each pixel. According to the rendering method provided by the technical schemes of the invention, the pixels of the rendering model in a scene are projected to the screen space from the camera space; the ambient occlusion values of the pixels in the screen space are calculated; and therefore, a frequent space transformation process is omitted, the number of the instructions of a GPU during operation can be saved, and the operation efficiency of the GPU canbe improved.

The invention provides a data enhancement method based on a depth image, which is applicable to the field of computer vision and algorithms based on depth image recognition, target detection, behaviorrecognition and the like. The method is mainly composed of a pixel coordinate to three-dimensional point cloud part, a three-dimensional point cloud space conversion part, a three-dimensional point cloud to pixel coordinate part and a minimum value filtering processing part. The step of converting a pixel coordinate system into the three-dimensional point cloud is to convert the plane pixel coordinate points in the depth image into the three-dimensional space point cloud under the world coordinate system through the conversion relationship among the pixel coordinate system, the image coordinate system, the camera coordinate system and the world coordinate system. In three-dimensional point cloud space transformation, the depth image is converted into three-dimensional space point cloud, and random translation transformation and random rotation transformation are carried out on the three-dimensional space point cloud to form new three-dimensional space point cloud. The newly generatedthree-dimensional space point cloud is projected into the depth image through the conversion relationship between the world coordinate system and the pixel coordinate system. A new depth image after data enhancement is obtained after minimum filtering processing. The data enhancement method provides a data expansion method for research based on depth images in the computer vision neighborhood. According to the method, the generalization ability of the network model can be greatly improved.

A system for visually comparing a part to a drawing of said part in computer memory or on a screen is provided. The system includes a camera for capturing either a single image or an image stream of the part, a computing module for receiving the image or image stream, for obtaining, scaling, positioning and displaying the drawing of the part on a display unit and for overlaying the drawing on the image or image stream. The system determines if the part is manufactured within predefined tolerances at least one control section. In use, a user defines at least one control section and the system performs a CAD-space to image-space transformation of each control section. A deviation between the CAD drawing and the part at each control section is calculated, and this deviation is indicated to the user. The deviation is compared to a predetermined tolerance in order to determine if the part at each control section is within the predetermined tolerance; and a pass / fail indication is determined based on the comparison. For more complex shapes, a user can select an entire geometrical shape, and the system automatically determines a plurality of control sections within the geometric shape, in order to provide the user with a shape pass / fail indication.

A scattering information source positioning method based on beam-space transformation in a large-scale MIMO system comprises the following operation steps of (1) signal preprocessing stage, wherein a beam forming vector is used for transforming a received user signal vector into a vector with low dimensionality, namely the beam-space transformation, and a sample covariance matrix of the transformed vector is calculated; (2) estimation stage, wherein a signal subspace and a noise subspace are obtained through the sample covariance matrix, and the user position is estimated by the utilization of the linear relation between parts of the signal subspace and the orthogonal relation between the array manifold and the noise subspace. By the utilization of the characteristic of less information loss of the beam-space transformation of large-scale MIMO signals, the user position information is obtained after the dimensionality of the signal vector is reduced. The method can reduce the estimation complexity on the premise of improving the estimation precision of the user position information.