69 results about "Geometric consistency" patented technology

The invention discloses a block model building method for complex geological structure, and relates to the geological field of three-dimension modeling comprising the following steps: a, triangular meshing description is carried out on planes or fault planes; b, after step a, level intersection in block modeling for complex geological structure is converted into triangle intersection in space, whether the two triangles are intersected is judged, points of intersection are found from the two intersected triangles, and a line of intersection is obtained by connecting the points of intersection; c, geometric consistency and topological consistency are carried out on inside of each intersected triangle; d, after step c, enclosing block is extracted to obtain an interface formed by peripheral sides of the enclosing block, and is defined geological attribute to form three dimension model block. The method in the invention completely solves the problems of three dimension model input and blocking in complex geological regions, avoids using equations set with high solving difficulty and calculated amount, and is simple and practical.

An automated 3D Optical Metrology Scanning and Computer Aided Inspection System for dimensional inspection of precision manufactured parts. The system example and implemented configuration is based within a relocatable cabinet providing ambient light and optional temperature control. The cabinet further includes a part placement area having an optical metrology scanner positioned over a multi-axis robotic arm positioned in the part placement area. The robotic arm is constructed and arranged to grip and manipulate parts within a field of view of the optical metrology scanner. The robotic arm provides adequate multi-axis control to rotate and tilt and translate tp manipulate the part to allow substantially every surface of the part to be scanned. Dimensional comparison and analysis software application provide geometric conformance / deviation plus extraction of the dimensions indicated in the part computer aided design (CAD) model.

The invention relates to a method suitable for multi-view-angle automatic global registration of ground laser point cloud data. The method involves two key modules: a module for extracting semantic feature points and a module for feature matching. The method comprises: step 1, extracting the semantic feature points: obtaining the semantic feature points in a series of modes of data slicing, distance clustering, geometric element fitting and the like; step 2, matching the semantic feature points: matching the semantic feature points by constructing a triangular geometric constraint condition and a semantic constraint condition, and removing error matching in a geometric consistency clustering mode; and finally, constructing a weighted undirected graph by taking a reciprocal of a feature point number as a weight, and finally obtaining global registration parameters of all stations by taking a minimum spanning tree of the weighted undirected graph as a registration path, thereby realizing global optimal registration. According to the invention, the method suitable for multi-view-angle automatic global registration of the ground laser point cloud data is constructed; the method can effectively resist influences of noise, point density and coverage; and the method improves the laser scanning operation efficiency, thereby having very high practical values.

The invention provides a fast retrieval method based on an inverted technology. The method comprises the following step of: 1, taking out all pieces of data which meet requirements of a corresponding centre from an inverted table by using geometrical consistency according to a quantization table to form a data set A; 2, quickly sorting the data in the data set A and counting the occurrence frequency of the data by using an array-based bucket sorting method; 3, calculating matching rate according to content consistency, and filtering out interference data to obtain a data set B; and 4, according to an accuracy requirement, in the data set B, quickly searching an optimum set C as an output result. By the method which is quite high in operating speed, the data which is concerned by a user can be quickly and accurately found in disordered mass data, and the accuracy of a query result can be adjusted according to practical application. In the method, time complexity is O(n), relevant to the scale n of the data set A and irrelevant to the scale of a total data set, so that query efficiency is greatly improved.

A block model constructing method for complex geological structures is provided. The method comprises the following steps: a) making a triangle mesh description of a layer plane or a fault plane; b) judging whether two triangles intersect, and finding out the intersection points if they intersect; c) performing a geometric consistency and topological consistency processing within every intersectant triangle; d) extracting an enclosing block to acquire an interface constituted of peripheral edges of the enclosing block, and defining a geological attribute of the enclosing block to form a three-dimensional model block. The method needs not to solve difficult equation sets with large computation, thus simplifying the constructing method.

The invention discloses a visual odometer method based on an end-to-end semi-supervised generative adversarial network, and relates to a visual odometer calculation method. The method comprises: constructing a feature generation network; constructing a discrimination network; performing adversarial training; solving dynamic scene problems. The method specifically comprises: marking point positioninformation and extracting feature descriptors through an SIFT feature algorithm, and obtaining related frames and matching feature points by randomly generating a homography matrix so as to generatecorresponding training labels; the generation network can input an original image to generate corresponding feature point positions and corresponding depth descriptions; and the discriminant network combining the semantic-geometric consistency loss function, the feature point cross entropy loss function and the discriminant loss function to form an adversarial with the generative network. Throughtraining of the GAN, the generation network can generate point position information and depth description which cannot be distinguished by the discrimination network, so that manual design of local features is avoided.

An integrated automated 3D Optical Scanning and Computer Aided Inspection System for dimensional inspection of precision manufactured parts. The system is based within a portable cabinet having lockable caster wheels for portability and retractably support feet for enhanced stability. The cabinet further includes a part placement area having an optical metrology scanner positioned over a multi-axis robotic arm positioned in the part placement area. The robotic arm is constructed and arranged to grip and manipulate parts within a field of view of the optical metrology scanner. The robotic arm provides multi-axis to rotate and tilt a base to allow substantially every surface of the part to be scanned. Dimensional comparison and analysis software application provide geometric conformance / deviation plus extraction of the dimensions indicated in the part computer aided design (CAD) model.

A method for feature matching in image recognition is provided. First, image scaling may be based on a feature distribution across scale spaces for an image to estimate image size / resolution, where peak(s) in the keypoint distribution at different scales is used to track a dominant image scale and roughly track object sizes. Second, instead of using all detected features in an image for feature matching, keypoints may be pruned based on cluster density and / or the scale level in which the keypoints are detected. Keypoints falling within high-density clusters may be preferred over features falling within lower density clusters for purposes of feature matching. Third, inlier-to-outlier keypoint ratios are increased by spatially constraining keypoints into clusters in order to reduce or avoid geometric consistency checking for the image.

The invention belongs to the technical field of plant modeling and computer graphics processing, in particular to a process-type plant model reconstruction method based on multi-view images, aiming atsolving the problems of large error and low precision between the plant model constructed by the prior method and the real data. The method comprises the following steps: acquiring the plant multi-view picture information; performing dense depth estimation based on color and geometric consistency; acquiring 3D point cloud data based on multi-view depth information fusion; a three-dimensional treebranch structure model of the plant is constructed by using a pre-constructed branch structure growth parameter representation model and the three-dimensional point cloud data to restrict plant growth. The invention provides a solution for reconstructing a complete tree model from a multi-view image. The obtained point clouds and models have a high degree of consistency with the original image, which skillfully combines plant process modeling and data-driven plant reconstruction method, not only to ensure the accuracy of the model, but also to maintain the biological significance of plants.

The invention provides a global positioning method based on geometric information, and relates to the technical field of autonomous positioning of mobile robots. According to the method, the current sensing information of the mobile robot is obtained, a three-dimensional point cloud map is generated by adopting an SLAM algorithm, the three-dimensional point cloud map is divided into global sub-maps, the robot moves to establish a local map, a global descriptor is calculated, an optimal local sub-map is found in the global sub-maps, the local descriptors of the optimal local sub-map and the global sub-maps are calculated, a corresponding relation of the local descriptors is found by utilizing a kd-tree, and error matching is removed by utilizing a geometric consistency algorithm; for the found corresponding relation, rough pose transformation of the local map and the global sub-maps is obtained by utilizing an absolute positioning algorithm; finally, fine matching is carried out by utilizing an ICP or NDT algorithm, the laser scanning and a global map are aligned finally, and the pose relation of the robot relative to the global map is obtained. The method is wide in application scene, the global positioning can be realized only by a laser range finder, and the method has the characteristics of easy operation and high precision.

The invention relates to a manufacture method of a polarization maintaining optical fibre, which comprises the following steps: a mandrel and stress application performs are depositionally made with a deposition method, the mandrel and a quartz sleeve are melted to be draught into an optical fibre perform with a fibre core and a covering, and the two stress application performs are respectively inserted into two holes of the perforated optical fibre perform to form a combined optical fibre perform. The combined optical fibre perform is put into a fibre drawing tower for drawing to be made into a panda type polarization maintaining optical fibre. During drawing, the temperature in the drawing tower is controlled within 1700 DEG C to 2000 DEG C, the optical fibre drawing tension is controlled at over 30g, and the optical fibre drawing speed is controlled at over 300m / min. The invention has simple process and high reliability. The high-tension and high-speed drawing reduces the fluctuation of the optical fibre drawing speed so as to improve the geometrical consistency of the optical fibre, optimize the service performances of the polarization maintaining optical fibre, and improve the drawing production efficiency, and thus the capacities of tension screening, dynamic fatigue and tensile resistance of the polarization maintaining optical fibre meet or are even superior to relative index requirements.

The invention discloses a three-dimensional reconstruction method capable of obviously increasing the time efficiency of method realization. The three-dimensional reconstruction method comprises the following steps of: (1) feature point matching: carrying out feature extraction on an image by virtue of a sift method, determining a matching relationship by virtue of the nearest neighbour matching method, determining matching by the ratio of a next nearest neighbour Euclidean distance to the nearest neighbour Euclidean distance, and accurately matching by a homographic and self-adaptive non-maximum value suppression processing method; (2) quasi-dense matching: processing the feature matching points obtained in the step (1) by a quasi-dense matching method; (3) re-sampling; (4) generation for an initial surface patch: generating an initial quasi-dense spatial surface patch from the quasi-dense matching points through spatial triangularization; (5) surface patch diffusion: diffusing the initial surface patch by virtue of the characteristic of normal and position similarity of the adjacent surface patches, and gradually obtaining dense spatial surface patches; (6) surface patch filter: eliminating exterior points by virtue of geometric consistency constraint and gray-level consistency constraint.

The invention discloses an image splicing method based on SIFT feature point matching. The method comprises steps of performing dimensionality reduction of feature descriptors by using a bidirectional 2DPCA method, so as to improve the efficiency and precision of matching; performing crude purification of feature points by the geometrical consistency between the matching feature point pairs, so as to improve the operation efficiency of the random sampling consistency method; performing image fusion by adopting Contourlet transform and eliminating the chromatic aberration of the fusion image, so as to achieve high quality splicing effect of two images. The image splicing method based on SIFT feature point matching solves the problems that the conventional image splicing method is low in matching precision and efficiency, and the fusion image is not clear.

The invention discloses a multiple dimensioned dense matching method and system. The multiple dimensioned dense matching method includes the following steps: constructing image pyramids for multiple images, and calculating a relevant matching image set of each image, wherein each image pyramid has a zeroth to (n-1)th layer and each layer corresponds to a dimension; constructing a depth range of each image; calculating a depth map, a normal vector map and a visibility map of each layer sequentially from the (n-1)th layer of each image pyramid, converting to the next layer of each image pyramid,and calculating the depth map, the normal vector map and the visibility map of the present layer by taking the depth map, the normal vector map and the visibility map of the previous layer of each image pyramid as initial values of an algorithm of the present layer until to the zeroth layer of each image pyramid; and carrying out depth map fusion according to the depth map, the normal vector mapand the visibility map of the zeroth layer of each image pyramid to generate point clouds. The method and system of the invention use a depth fusion framework to carry out dense matching, consider depth information, normal vector information and visibility information, and combines constrains of optical consistency and geometric consistency, thereby improving the precision. Meanwhile, a multiple dimensioned strategy is used, and so the memory requirements are reduced and the efficiency is improved.

Based on low-order overall situation geometry consistency check error match detection method comprises the steps of using SIFT (scale invariant feature transform) and BoF (bag of features) for extracting and matching feature points in two images to be compared; calculating and combining the square distance matrixes of the two images; dissolving the combined square distance matrix to be a square distance matrix A formed by real matching pairs and a difference matrix E caused by error matching; calculating the sums of elements in each line of the E matrix, sequencing the sums, calculating the second order difference of all sequenced sums, taking the point with the maximum second difference value as a threshold value, and judging the corresponding feature points, higher than the threshold value, of all lines as the error matching pairs; removing the error matching pairs, calculating the similarity between the images according to the real matching points, and outputting an image index result according to the sizes of the similarity through sequencing. The method is simple and efficient, the feature coordinates are used as the only input information, the similarity conversion causing the image difference can be handled, and all error matching pairs can be rightly detected.

The invention discloses a dense three-dimensional reconstruction method and system based on multi-scale geometric consistency guidance, and belongs to the field of computer vision, and the method comprises the steps: building an image pyramid based on an image set; carrying out depth estimation on the most rough scale of the image pyramid by utilizing photographing consistency, and acquiring a depth map on the most rough scale; taking the most rough depth image on the scale as a depth image on the current scale, sequentially carrying out upsampling and error depth estimation correction on thedepth image on the current scale, and carrying out optimization by utilizing geometric consistency to obtain an optimized depth image on the next scale; taking the optimized depth map of the next scale as a depth map of the current scale, and then carrying out upsampling, correction and optimization until optimized depth maps of all original images in the image set are obtained; and fusing the optimized depth maps of all the original images to obtain dense three-dimensional point cloud. The depth information of the weak texture area can be estimated, and the depth information of the detail area can be kept.

The invention provides a two-dimensional high lift device parametrization design method based on an NURBS curve and meeting engineering constraints. Firstly, the positions of slat and main wing interfaces and flap and main wing interfaces are determined on a two-dimensional airfoil profile, and control nodes are arranged, deflection angles of slats and flaps and rotary shaft positions are determined according to aerodynamic design requirements; secondly, the control nodes are respectively fitted to obtain geometric profiles of the slat and main wing interfaces and the flap and main wing interfaces; thirdly, intersection points of the interfaces and the two-dimensional airfoil profile are found again and are utilized to split the two-dimensional airfoil profile, the geometric profiles of the interfaces are spliced with the split two-dimensional airfoil profile to form slat profiles, flap profiles and main wing profiles; finally, according to the deflection angles of the slats and flaps and rotary shaft positions, fixed-shaft rotation is conducted on the slats and flaps to obtain a high lift device profile. The method adopts a non-uniform rational B-spline method to parametrize a two-dimensional high lift device, a complicated geometric profile is fit at higher precision, and the problem that the engineering constraints and geometrical consistency constraints are difficult to meet is solved.

The invention provides a three-dimensional augmentation device parameterization method meeting engineering constraints based on an FFD method, and the method comprises the steps: firstly, measuring the position parameters of a three-dimensional augmentation device, arranging a control frame on a slat and a flap of an initial configuration, carrying out parameterization disturbance on the object surface in the control frame, and changing the geometric shapes of the slat and the flap in the control frame; Reversing a control frame arranged on the initial configuration slat and the flap to the main wing according to the position parameters to obtain local coordinates of corresponding object plane grid points on the main wing in the corresponding control frame; and reversing the deformed control frame on the slat and the flap to the main wing according to the position parameters to obtain a global coordinate vector after the corresponding object plane grid point on the main wing is deformed. According to the method, the FFD method is used for guaranteeing the fitting precision of the geometric shape of the lifting device, the intersection interface part of the main wing in the method is processed, the geometric consistency constraint is met, finally, the generated three-section wing shape is subjected to fixed-axis rotation and translation operation, and the engineering constraintproblem is solved.

The invention discloses an unsupervised semantic segmentation method for a cross-domain remote sensing image. The method comprises the following steps: acquiring an unlabeled target domain remote sensing image to be segmented; inputting the unlabeled target domain remote sensing image to be segmented into an unsupervised semantic segmentation model which is trained in advance, wherein the unsupervised semantic segmentation model comprises a geometric consistency constraint module, a domain adaptation network module and a semantic segmentation network module; and outputting a segmentation result image consistent with the unlabeled target domain remote sensing image to be segmented in size. According to the method, the segmentation performance of a segmentation model trained on labeled source domain data on a target domain can be improved, so that the dependence of a semantic segmentation task on large-scale labeled data is reduced, and meanwhile, the generalization performance of the semantic segmentation model on different image domains is improved; the method can achieve the accurate segmentation of the to-be-segmented unlabeled target domain remote sensing image, and is higher in segmentation precision.

The invention provides a mobile visual search framework based on a CRBM and Fisher network and relates to the image retrieval of a mobile terminal. The method comprises the following steps of 1) constructing and training a continuous restricted Boltzmann machine network; 2) constructing and training a Fisher layer network. According to the invention, the sub-space feature information of the localfeature nature of non-Gaussian distribution is found by adopting the nonlinear dimensionality reduction algorithm CRBM in the aggregation global compact binary feature algorithm. Meanwhile, more efficient global features are obtained through adopting a fisher-based network structure aggregation fisher vector. A compact self-adaptive feature is obtained by adopting the scalar quantization algorithmand the bit self-adaptation algorithm. Therefore, the length of the transmitted image feature information can be selected according to different self-adaptive selections of the network bandwidth of amobile terminal. During the retrieval stage, global features are roughly matched to obtain a candidate set, and the accurate matching of the geometric consistency test is conducted by adopting localfeatures. Therefore, the mobile visual search framework can be adapted to large-scale image retrieval tasks.

The invention relates to an accumulation energy based three-dimensional image line clipping redirecting method. The method comprises the following steps: 1, inputting a three-dimensional image and a corresponding vision difference map; and initializing an energy matrix and an energy accumulation matrix of the input image; 2, finding out the optimal seam set of the left and right vision images based on the dynamic planning method, and recording the route; 3, accumulating the energy of each pixel on the optimal seam set route on the adjacent eight pixels; 4, detecting and accumulating the energyto adjacent two pixels in the maximum edge information direction; 5, deleting the optimal seam set; updating the energy matrix of the image, the energy accumulation matrix M and the corresponding vision difference image; and 6, iterating the abovementioned steps until the target size is met, and thus obtaining the re-directed left and right vision images and the corresponding vision difference images. With the adoption of the method, the vision distortion of three-dimensional images, caused by seam operation concentrated in some area of the images, can be effectively reduced; the geometric consistency of the left image and the right image can be maintained; and the three-dimensional image line cutting clipping performance is improved.

The invention discloses an L1 norm total geometrical consistency check-based wrong matching detection method, which comprises the steps: SIFT (scale-invariant feature transform) and a bag-of-features model are used for extracting and matching characteristic points of two images; the squared distance matrix of the two images is calculated; a golden section search method is used for solving: min[Lambda is larger than 0]||D[1]-Lambda.D[2]||[1]; the difference matrix E is calculated, wherein E=||D[1]-Lambda*.D[2]||[1]; the sum of all elements in each row in the matrix E is calculated, the sums are sequenced from large to small, the second order difference of the sequenced sums is calculated, the point with the maximum second order difference value is taken as a threshold value, and characteristic points which correspond to all rows and are higher than the threshold value are determined as wrong matching pairs; after the wrong matching pairs are removed, the similarity of the images is calculated according to the true matching points of the two images, and then an image search result is output according to the similarity from large to small. According to the detection method disclosed by the invention, complicated backgrounds, partly shielding and various similar geometrical transformation among the similar images are allowed; only the coordinate information of characteristic points is utilized, and the detection method is simple and efficient.

The invention relates to a laser SLAM loopback detection system and method based on a graph descriptor. The system comprises a semantic segmentation module, a full-graph descriptor extraction module,a full-graph descriptor matching module, a vertex descriptor extraction module, a vertex descriptor matching module and a geometric consistency verification module. According to the method, semantic information extracted from point cloud data is utilized to form two graph descriptors, namely a full graph descriptor and a vertex descriptor, so as to characterize a point cloud frame and a semantic object. Compared with a traditional algorithm for extracting the descriptors from the pixel level, the invention can avoid heavy normal vector calculation tasks and avoid the problems caused by great change of the visual angle, and can detect loopback more quickly and robustly. The potential loopback candidate frames are roughly screened out through the full-graph descriptors, then detail information of the query frame and the loopback candidate frames is matched more finely through the vertex descriptors, the method is a coarse-to-fine search process, and theoretical guarantee is provided forthe real-time performance of the method.

The invention relates to a manufacture method of a polarization maintaining optical fibre, which comprises the following steps: a mandrel and stress application performs are depositionally made with a deposition method, the mandrel and a quartz sleeve are melted to be draught into an optical fibre perform with a fibre core and a covering, and the two stress application performs are respectively inserted into two holes of the perforated optical fibre perform to form a combined optical fibre perform. The combined optical fibre perform is put into a fibre drawing tower for drawing to be made into a panda type polarization maintaining optical fibre. During drawing, the temperature in the drawing tower is controlled within 1700 DEG C to 2000 DEG C, the optical fibre drawing tension is controlled at over 30g, and the optical fibre drawing speed is controlled at over 300m / min. The invention has simple process and high reliability. The high-tension and high-speed drawing reduces the fluctuation of the optical fibre drawing speed so as to improve the geometrical consistency of the optical fibre, optimize the service performances of the polarization maintaining optical fibre, and improve the drawing production efficiency, and thus the capacities of tension screening, dynamic fatigue and tensile resistance of the polarization maintaining optical fibre meet or are even superior to relative index requirements.

The invention discloses a compact visual descriptor deep neural network generation model in visual retrieval, and relates to image retrieval. The model is implemented by the following steps of: constructing a Fisher layer network; constructing a grouping and secondary classification module; training a loss function on the basis of a maximum boundary condition; for image library images and a queryimage, firstly extracting local features of the images, carrying out aggregation and binary embedding on the local features of the images by using a trained network structure so as to obtain binary codes of the images, carrying out matching in an image library according to the binary code of the query image, returning images with high similarities as a matched candidate set, carrying out geometricconsistency inspection on the candidate set by using the local features so as to carry out accurate matching, and returning a final query result. According to the model, the flexible Fisher network is used to aggregate local features of images so as to generate more efficient global feature Fisher vectors, and the grouping and secondary classification module is used to carry out binary coding onthe Fisher vectors so as to obtain compact global binary features.

The invention relates to a phase unwrapping method based on multi-view geometric consistency. Firstly a multi-view structured light system is established, each view includes a set of measuring units,and calibration of the system is performed so as to acquire the projection parameter matrix of each camera; then each structured light unit uses a template to project to a measuring target, and each view measures and calculates a reference phase diagram and an acquisition phase diagram; and finally a three-dimensional mapping target energy function is established through the camera parameters andthe deformation phase under the constraint of the multi-view geometric consistency and the compensation phase of each pixel point is obtained so as to realize phase unwrapping. The correlation and theconstraint between the views are considered and the obtained phase unwrapping result is more accurate; and multiple views are synchronously unwrapped so that the workload of phase unwrapping can be greatly reduced.

The invention discloses an image semantic feature matching method based on geometric consistency. The method comprises the steps of semantic feature extraction, feature matching initialization, feature matching positioning optimization, image semantic alignment and the like, wherein the semantic feature extraction is to extract high-level semantic features by using a convolutional neural network to construct a five-layer semantic feature pyramid; the feature matching initialization is to design a semantic feature matching constraint rule on the top layer of a feature pyramid based on geometricconsistency, and construct an energy function; the feature matching positioning optimization is used for improving the positioning precision of feature matching, and the accuracy of feature matchingpairs is improved layer by layer through a pyramid back propagation algorithm; and finally, geometric transformation model parameters between the to-be-matched images is estimated by adopting a localgeometric transformation model, and image deformation is performed to realize image semantic alignment. The method can improve the precision of semantic feature matching, and achieves the alignment ofthe geometric attitude and orientation of the foreground target.

The invention discloses a visual ADAS-ECU hardware-in-loop simulation test system based on a high-fidelity game engine. The system comprises an ADAS camera imaging simulation module, an ADAS computercontrol module, a central control module, a physical scene module which is used for providing a simulated physical scene and is based on the high-fidelity game engine, and a vehicle dynamics simulation module and a geometric consistency video stream generation module which are connected with the physical scene module, and a setting and evaluation module.

The invention provides an ADS-B anti-interference anti-cheat ground single-station system. According to the system, on the basis of reserving all functions of an original ADS-B system, a monopulse anddifference angle measurement processing technology is adopted for performing accurate detection on the direction of a target source, and potential false cheat targets in most directions can be judged, so that the ADS-B anti-interference anti-cheat purpose is achieved; meanwhile, the system also has the characteristics of being small in size, light in weight, easy to erect, convenient in equipmentmaintenance, high in practicability, and small angle resolution of less than 0.6 degrees; and while angle measurement processing is performed by using monopulse and difference technologies, the judgment of the target geometric consistency and the target intention consistency is also fused, so that the ADS-B anti-interference anti-cheat effect can be more effectively realized.