95 results about "Nonlinear deformation" patented technology

The invention discloses a deformable convolutional neural network-based infrared image object identification method. The method comprises the steps of constructing a training set and a test set; establishing a convolutional neural network architecture; adding a softmax classifier to the last layer, and setting an objective function; performing sampling by adopting a convolution kernel of linear ornonlinear deformation; performing pooling operation in a pooling layer by adopting a rule block sampling-based ROI pooling method which is the best in the industry at present; setting learning rate parameters according to experience; and easily performing standard back propagation end-to-end training, thereby obtaining a deformable convolutional neural network. An experiment proves that the spatial geometric deformation learning capability is introduced in the convolutional neural network, so that an identification task of an image with spatial deformation is better finished; the geometric transformation modeling capability of the convolutional neural network and the effectiveness of target detection and visual task identification are improved; and dense geometric deformation in space issuccessfully learnt.

Method and system for the detection of interval change in medical images. Three dimensional images, such as previous and current section images in CT scans, are obtained. An anatomic feature, such as the lungs, is used to select sections containing lung by a gray-level thresholding technique. The section correspondence between the current and previous scans is determined automatically. The initial registration of the corresponding sections in the two scans is achieved by a rotation correction and a cross-correlation technique. A more accurate registration between the corresponding current and previous section images is achieved by local matching. A nonlinear warping process which is also based on the cross-correlation technique is applied to the previous image to yield a warped image after the matching. The final subtracted section images were derived by subtracting of the previous section images from the corresponding current section images. Interval changes such as a change in tumor size and a newly developed pleural effusion are enhanced significantly.

The invention provides a method for correcting geometric distortion of a text image. The method comprises the following steps of: extracting a text row, in a horizontal direction, of the text image, fitting the text row by using a B-spline curve to obtain a text row curve; estimating a vanishing point of a cylindrical surface straight bus, a text row tangent convergence line and a focal length parameter of a camera by using the tangent convergence symmetry of the text row curve; constructing latitude lines of an isometric grid for the fit text row curve by adopting a crossratio interpolation value; estimating a directrix of the cylindrical surface by using similar perspective projection transformation of the text row curve and constructing longitude lines of the isometric grid; estimatinga length-to-width ratio of the isometric grid by using the text row curves at the top end and the bottom end of the image; and mapping meshes of each isometric grid into a square grid through a deformation technology based on the grid to finish the correction of the geometric distortion of the text image. By the method, the problem that perspective distortion, nonlinear deformation distortion andcompound distortion of the perspective distortion and the nonlinear deformation distortion cannot be corrected at the same time in the prior art can be solved; therefore, total correction of the geometric distortion of the text image shot by the camera can be realized.

The invention provides a method, a device and a system for measuring the skew deformation amount of a tower of a wind generating set. The method comprises the following steps of: acquiring a skew angle which is measured by a skew sensor and an acceleration which is measured by an acceleration sensor on the tower; according to the skew angle and the acceleration, which are measured, acquiring the static deformation amount, quasi-static deformation amount and dynamic deformation amount of the tower; according to the static deformation amount, quasi-static deformation amount and dynamic deformation amount of the tower, determining a skew deformation equation of the tower, wherein the skew deformation equation comprises a static and quasi-static deformation equation and a dynamic deformation equation; and according to the skew deformation equation of the tower, determining a skew deformation curve of the tower. By adoption of the technical scheme, a nonlinear deformation curve of the tower can be obtained, and accuracy in measurement of the skew deformation amount of the tower is improved.

The present invention relates to a method and a system for matching MR image feature points before and after the nonlinear deformation of a biological tissue. According to the technical scheme of the invention, a feature point automatic detection method based on a depth-cascaded convolutional neural network is provided. According to the method, firstly, a general region of feature points is obtained through the first layer of the depth convolutional network. Secondly, the position of a target feature point is approximated step by step in the second and third layers of the cascade convolutional network, so that the detection rate of feature points is further improved. The method aims to solve the problem in the prior art that the feature point distinguishing ability is reduced due to the image nonlinear deformation of existing feature point descriptors. In this way, a Riemannian manifold is combined with the kernel method to construct a nonlinear deformation feature point descriptor for robustness. The three-dimensional feature points of a magnetic resonance image are mapped into a four-dimensional Riemannian manifold space. Meanwhile, the feature points are further mapped into a higher-dimensional Hilbert space based on the kernel method, so that a richer description of data distribution is obtained. Meanwhile, a real geometric distance between feature points is obtained, so that the feature points are matched.

The invention discloses an image correction method suitable for a bending deformation two-dimensional code, and specially aims at the problem of two-dimensional code nonlinear deformation because of the photographing angle and the bending surface. The method is mainly divided into two steps of a two-dimensional code area extraction and bending deformation correction. The step of two-dimensional code area extraction comprises two-dimensional code image preprocessing, two-dimensional code position detection pattern positioning and two-dimensional code closed extraction. The step of bending deformation correction comprises original image reference corresponding point extraction, standard image reference corresponding point extraction and construction of a bending deformation correction transformation model. The two-dimensional code corner point detection method is improved by using the slope and intercept knowledge of plane geometry to realize accurate positioning of the two-dimensional code corner points, a "roller" method is designed to solve the edge points of the bending deformation two-dimensional code, and the side length relationship of the two-dimensional code and the positiondetection pattern is calculated through combination of curve fitting to construct the transformation model so that adaptive correction of the bending deformation two-dimensional code under the general condition can be realized.

The invention discloses a controllable dynamic shape interpolation method with physical third dimension. According to the method, a rotation-strain space definition is provided, and a set of conversion mode from a three-dimensional model to a rotation-strain space coordinate is established, so that the primary nonlinear deformation can be expressed by linear operation of the coordinate in the rotation-strain space; and the method greatly reduces the algorithm complexity and improves the solving efficiency by combining the traditional mode shape decomposition to perform dimension reduction. The method comprises the following steps of: decomposing a base vector acquired by the traditional mode shape decomposition into rotation and strain parts, and constructing a deformation subspace by using rotation-strain corresponding to the main component bases; obtaining a rich dynamic deformation effect by editing and regulating mode shape coordinate motion interpolation curve parameters in the subspace; and finally, obtaining a new model pose sequence through a reconstruction process according to the interpolated mode shape coordinate.

This algorithm provides a marker-free approach to establishing the pixel correspondence among the IR images taken at different times, which is the basis for quantitatively characterizing the variation of the heat energy and patterns pixel-wise on a breast surface. The idea is to use the corner points of the heat pattern and the branch points of the skeletons of the heat pattern on the body surface as the initial fiducial points for the longitudinal IR image registration. The Thin-Plate Spline technique is used to model the nonlinear deformation between two IR images taken at two different times. Mutual information between the TPS-transformed image and the target image is employed as the metric quantifying the quality of the longitudinal IR image registration. To optimize the registration, Nelder-Mead simplex method is used to locally modify the pairings of the fiducial points in the source and target IR images to maximize the mutual information.

The invention discloses a calculation method for the stratum initial cracking pressure of deep shale, and belongs to the technical field of exploration. The calculation method for the stratum initial cracking pressure of the deep shale is used for calculating the stratum initial cracking pressure of the deep shale and comprises the steps that the size of a casing pipe, the size of a cement loop and the size of a borehole are acquired, and a stratum-shaft-perforation three-dimensional model is constructed by combining perforation parameters; material attributes are given to the stratum-shaft-perforation three-dimensional model on the basis of material parameters of the casing pipe and the cement loop and stratum rock mechanical parameters, wherein the stratum rock mechanical parameters have the transverse isotropic homogeneity characteristic and the nonlinear deformation characteristic; boundary loads are applied to the stratum-shaft-perforation three-dimensional model, and the stratum initial cracking pressure is obtained.

The invention provides a Ti-Nb-Mo-Sn near-beta titanium alloy with low elastic modulus and high strength. The alloy comprises the following chemical components in percentage by weight: 15-35% of niobium, 1-8% of molybdenum, 1-8% of stannum and 0-1.2% of titanium dioxide powder. The Ti-Nb-Mo-Sn near-beta titanium alloy with low elastic modulus and high strength has the beneficial effects that the systemic alloy has good cold machining performance and a low machining cementation index and can coldly deform in a large scale by using cold machining processes such as cold rolling, cold wire drawing and the like; the systemic alloy has the characteristic of nonlinear deformation after being thermally machined or coldly machined and has large restorable elastic strain, low initial modulus and average modulus and high strength; the systemic alloy has low elastic modulus, high yield strength and breaking strength and high linear elasticity after being smelted, thermally machined or coldly machined and thermally treated; and the systemic alloy can be used for preparing medical equipment, physical training equipment and industrial equipment, and can be used for biomedical applications such as artificial bones, bone joints, implanted tooth roots, bony plates and the like.

The invention discloses a robust target tracking method based on a phantom adversarial network, and relates to a computer vision technology. The method comprises the following steps: firstly, proposing a new illusion confrontation network, aiming at learning nonlinear deformation between sample pairs, and applying the learned deformation to a new target to generate a new target deformation sample.In order to effectively train the proposed phantom confrontation network, deformation reconstruction loss is proposed. The invention provides the target tracking method based on the illusion confrontation network based on the offline training, and the method can effectively alleviate the overfitting problem of the deep neural network in the target tracking process due to the online updating. In addition, in order to further improve the deformation migration quality, a selective variable migration method is provided, and the tracking precision is further improved. The target tracking method provided by the invention obtains a competitive result on the current mainstream target tracking data set.

An apparatus comprising two or more adjacent wide angle cameras, such as CMOS cameras, for imaging a document, and methods for combining corresponding images into one image while correcting radial, scale, rotation, translation and other linear or nonlinear deformations.

The invention relates to a pressure sensor structure of a compound diaphragm, comprising a monocrystalline silicon layer and a porous silicon layer, wherein one side of the monocrystalline silicon layer is concave and the porous silicon layer grows on the concave inner surface of the monocrystalline silicon layer. On the premise that the sensitivity of a low-range sensor is not influenced, the compound diaphragm of porous silicon and monocrystalline silicon is adopted for reducing the nonlinear deformation of the diaphragm, thereby reducing the flexibility of the diaphragm and finally achieving the purpose of enhancing the sensitivity or reducing the nonlinearity of the low-range sensor.

An anti-fatigue mat/shock pad has a mat body with a top surface, a bottom surface, a left side edge, a right side edge and a top and bottom edges. The top and bottom surface of the mat have raised regions thereon which are made up of longitudinally extending raised ribs arranged in parallel fashion with respect to one another and spaced apart a selected distance by longitudinal channels. The longitudinal ribs on the top surface are offset from the longitudinal ribs on the bottom surface in an alternating fashion so that they present a staggered array when viewed in cross section. This particular pattern of raised ribs on the top and bottom surfaces of the mat produces a non-linear deformation pattern under impact forces in use, such as would be caused by athletes running or jumping on an athletic field.

A printed circuit board manufacturing system and a manufacturing method thereof are disclosed. A method of manufacturing printed circuit board, comprising: providing a substrate that comprises a pad and an insulation layer covering the pad; acquiring an image of the substrate; acquiring location information of the pad by analyzing the image of the substrate; forming a via hole by removing a part of the insulation layer that corresponds the location information of the pad; and forming a via by filling the via hole with a conductive material, provides improved process conformity, even if the substrate has partial or nonlinear deformation, by considering the location information of the pad in the via hole forming. The improved conformity may allow more flexibility to substrate design and more integrity for circuitries on printed circuit board.

A data post-processing method for dam and landslide deformation GBSAR monitoring comprises the steps of obtaining N monitoring image subsets, processing the interiors of the image subsets, and obtaining N average images; taking one average image as a main image and the rest N1 average images as slave images to carry out interference processing to obtain N1 interference images; selecting a PS point, and establishing a Delaunay triangulation network; carrying out differential phase modeling and parameter estimation on each PS point; calculating according to the estimated linear deformation rate increment and residual phase increment to obtain linear deformation and nonlinear deformation corresponding to each PS point; and adding the linear deformation amount and the nonlinear deformation amount to obtain an overall deformation amount, and obtaining a dense deformation field of the whole monitoring object in a global range by adopting an interpolation mode. The data post-processing method provided by the invention solves the data analysis problems of long long-period monitoring time baseline and large information amount of dams and landslides, and is good in adaptability and high in precision in global space deformation monitoring information analysis.