314 results about "Plane fitting" patented technology

An image matching method and system for use with multiple images of a scene captured from different angles. Image matching is performed by identifying a plurality of segments within at least two images, determining an initial disparity values for pixels in the images and then determining initial disparity planes for the segments by fitting a plane to initial disparity values for the segments. A refined disparity plane set is created by iteratively refitting the disparity planes by using various fitting cost functions and weighted linear systems. A labeling of each segment to a disparity plane is made by minimizing a global energy function that includes energy terms for segment to disparity plane matching as well as penalizing disparity plane discontinuities between adjacent image segments.

The invention relates to an indoor scene three-dimensional reconstruction method based on a single depth vision sensor. The method is technically characterized by including the following steps of firstly, continuously scanning a whole indoor scene through the single depth vision sensor; secondly, conducting preprocessing including denoising, hole repairing and the like on collected depth data to obtain smooth depth data; thirdly, calculating point cloud data corresponding to the current depth frame according to the depth data collected in the second step; fourthly, conducting registration on point cloud obtained through different viewpoint depth frames to obtain complete point cloud of the indoor scene; fifthly, conducting plane fitting, achieving segmentation of the special point cloud, and establishing an independent and complete three-dimensional model of each object in the indoor scene. Scanning devices used by the method are simple; scanned data information is comprehensive, and the point cloud registration accuracy calculation efficiency is effectively improved; finally, a complete and high-quality three-dimensional model set with a geographic structure and a color map can be established for the indoor scene.

The invention discloses a fusion calibration method of a three-dimensional laser radar and a binocular visible light sensor. According to the invention, the laser radar and the binocular visible lightsensor are used to obtain the three-dimensional coordinates of the plane vertex of the square calibration plate, and then registration is carried out to obtain the conversion relation of the two coordinate systems. In the calibration process, an RANSAC algorithm is adopted to carry out plane fitting on the point cloud of the calibration plate, and the point cloud is projected to a fitting plane,so that the influence of measurement errors on vertex coordinate calculation is reduced. For the binocular camera, the vertex of the calibration plate is obtained by adopting an angular point diagonalfitting method; for the laser radar, a distance difference statistical method is adopted to judge boundary points of the point cloud on the calibration board. By utilizing the obtained vertex coordinates of the calibration plate, fusion calibration can be accurately carried out on the three-dimensional laser radar and the binocular visible light sensor, a rotation matrix and a translation vectorof coordinate systems of the three-dimensional laser radar and the binocular visible light sensor are obtained, and a foundation is laid for realizing data fusion of three-dimensional point cloud anda two-dimensional visible light image.

The invention discloses a method for overall calibration of a galvanometer type line laser scanning three-dimensional measuring system. First two lenses whose effective focal lengths are f1 and f2 are selected, and a camera is utilized to shoot a plane target and feature point coordinates are extracted, and main point coordinates of an image are calculated according to a varifocal method; then the lens whose focal length is f1 is used as a lens of the system, and for the feature point coordinates on the plane target which are obtained by the lens, intrinsic parameters and extrinsic parameters of the camera are calibrated; next, by changing galvanometer control voltage and driving a one-dimensional translation table to move, line laser stripes in different depth directions and different optical planes are obtained, and through operation of plane fitting and intersecting line obtaining, a transformation relation between a target coordinate system and a world coordinate system is obtained; and finally, coordinates under a system model are all unified to an image pixel coordinate system and the world coordinate system, thereby completing calibration of the whole system. The method has the advantages of simple calibration process, high calibration precision, fast calibration speed and the like.

The invention discloses a random trihedron-based radar-camera system external parameter calibration method. According to the method, an external parameter of a system can be solved with only two frames of data by using a trihedron scene a natural environment. The method comprises the following steps of: stipulating a world coordinate system by using a trihedron; performing planar fitting on the trihedron observed in a radar system to obtain a parameter of each plane and solving conversion relation of the world coordinate system and the radar coordinate system and relative motion between the two frames of data under the radar coordinate system; and in a camera system, solving an essential matrix by using matching characteristic points extracted by front and rear frames, then solving relation motion under the camera coordinate system, solving the plane parameter under the camera coordinate system by using the parameter under the radar coordinate system, finally solving an external parameter of a radar-camera, and performing final optimization by using coplanarity of points on corresponding planes under the two coordinate systems. A scene required by the method is simpler; and the method has the characteristics of high anti-interference performance, simple experimental equipment and high flexibility.

The invention discloses an obstacle detection method and system. The method comprises steps of acquiring point cloud data acquired by N line laser radar; dividing the point cloud data into Q regions including a first region, a second region, ..., and the Q region; carrying out plane fitting on the point cloud data in the first to Q regions to correspondingly obtain first to Q planes; successively acquiring coordinates of obstacles in the point cloud data of the first to Q regions; and carrying out voxelization processing on the coordinate data of the obstacles to obtain coordinate sets of dynamic obstacles and coordinate sets of static obstacles. According to the invention, large-scale compression of three-dimensional point cloud data is achieved; data processing quantity is greatly reduced; the environment is further divided into three parts: a road region (a feasible region), a static obstacle region and a dynamic obstacle region; and better environment information is provided for effective behavior decisions of the robot.

The invention discloses a calibration method of a line-structured light vision measuring sensor based on the space plane restriction, comprising the following calibrating steps: placing a two-dimensional planar target in a camera field range; collecting a group of images of multiple characteristic circles at the different positions on a two-dimensional target panel; determining the target topological relation; and synchronously completing the internal parameter calibration and the light planar calibration of a camera. In the process of the light planar calibration, substantial quantity of line-structured light planar three-dimensional points can be easily extracted by utilizing a camera imaging model and the restraining provided by the space plane formed by each position target; and planar fitting is carried out on the three-dimensional point set based on a principal element analytical method, so as to obtain an equation of the light plane in the coordinate system. The calibration method of the invention has simple operation and reliable result, is suitable for field calibration, and can satisfy the requirements of the line-structured light vision high-precision detecting tasks.

Systems and methods are provided for orienting a hair harvesting tool relative to a follicular unit to be harvested. Further, systems and methods of implanting follicular units are provided, which are based on the orientation of the existing follicular units or based on pre-determined insertion angle. The provided systems and methods use image processing and image-generated topological skill models. In some embodiments surface fit is performed using, for example, a planar fit, quadric fit, cubic fit, mesh fit and parametric fit.

The invention relates to a road boundary detection and tracking method based on a three-dimensional laser radar. The method comprises the following steps: converting polar coordinate data of a laser radar scanning point into rectangular coordinate data; Extracting ground points by using a random sampling consistency plane fitting method and a principal component analysis method; Roughly extractingroad boundary feature points by adopting a multi-feature, wide-threshold and hierarchical feature point extraction method; Dividing the roughly extracted road boundary feature points into a left roadboundary feature point and a right road boundary feature point by using a projection-based method by referring to the idea of linear discriminant analysis; Respectively filtering the left road boundary point and the right road boundary point by using a random sampling consistency method to obtain final road boundary points; Fitting the left road boundary line and the right road boundary line by adopting a quadratic polynomial random sampling consistency method; And tracking the road boundary line by adopting a Kalman filtering method based on amplitude filtering. The method can be used for processing various complex scenes and working conditions, and can be widely applied to unmanned driving.

The invention provides a defocus depth estimation and full focus image acquisition method of a dynamic scene. The method comprises the following steps of: acquiring first depth maps and globally inconsistent fuzzy kernels of a plurality of defocused images, and employing an image deblurring algorithm based on defocus depth estimation to carry out feedback iterative optimization to obtain a full focus image and a second depth map of each moment; after carrying out color segmentation on the full focus image of each moment, and carrying out plane fitting on the depth map, and carrying out refinement of space to obtain a third depth map, and carrying out optimization again to obtain an optimized full focus image; after carrying out optical flow estimation on the full focus image, carrying out smoothing on the third depth map, refining the third depth map in the time to obtain a depth estimation result with a consistent time. According to the method, a more precise dynamic scene depth estimation result and a full clear image can be obtained, and realization is easy.

The invention discloses a road vision detection method combined with laser point cloud data, belongs to the field of unmanned driving, and solves the problem of multi-sensor fusion in the prior art. The method comprises the steps of performing data-level fusion on the point cloud data and the image data; performing road division on each frame of image in the image data, and performing conversion after division to obtain an illumination-invariant space image; according to the M estimation sample consistency, performing plane fitting on the point cloud data after data level fusion to obtain a point cloud data coarse division result; performing road feature extraction on each frame of image, the illumination-invariant space image and the point cloud data coarse division result of the image data after data level fusion and the point cloud data after data level fusion corresponding to each frame of image; respectively sending the extracted road features into two random forests for training,and obtaining probability output of each pixel point and probability output of each point cloud data; and constructing each point cloud-image mixed conditional random field to obtain a road area. Themethod is used for road vision detection.

The invention discloses a parcel volume measuring method and system, a storage medium and a mobile terminal, and relates to the field of computer vision. Target point cloud is obtained, and the targetpoint cloud at least comprises the point cloud of a support surface and the point cloud of the top surface of a to-be-measured parcel; a plane equation of the support surface and the plane equation of the top surface of the to-be-measured parcel and the point cloud of the top surface of the to-be-measured parcel are obtained in a plane fitting mode, and the height of the to-be-measured parcel iscalculated according to the plane equation of the support surface and the plane equation of the top surface of the to-be-measured parcel; and the point cloud of the top surface of the to-be-measured parcel is projected on a plane corresponding to the plane equation of the to-be-measured parcel top surface to obtain a boundary point cloud; four frame angle points of the top surface frame of the to-be-measured parcel are obtained; according to the coordinates of the four frame angle points, the length and the width of the to-be-measured parcel are obtained through calculation, and the height ofthe to-be-measured parcel is combined, the volume of the to-be-measured parcel is obtained through calculation. The purpose of reducing the measurement error and simplifying the measurement requirement is achieved.

The invention provides a method for detecting a hook face circular hole based on a three-dimensional point cloud. The method comprises the steps that firstly, three-dimensional points on the edge of the circular hole to be detected on the point cloud are accurately extracted; then least square plane fitting is conducted on the edge points of the circular hole, so that an initial plane of a projection is obtained, the edge points are projected to the plane, and a coordinate system is converted, so that two-dimensional points of the projection are obtained; the radius of a least square circle and two orthogonal fixing rotary axes of the projection points are obtained; when the radius of the circle can not meet the judgment criterion of the circle, a least square plane respectively rotates around the two rotary axes, projection and iterative computation are conducted again until the radius of the circle meets the judgment criterion of the circle, so that the normal of the least square plane is the axial direction of the circular hole and the radius of the least square circle of the projection two-dimensional points is the radius of the circular hole. According to the method, the axial direction and the radius of the circular hole can be rapidly detected from the three-dimensional point cloud.

The invention discloses a quick binocular stereo matching method based on superpixel segmentation. The method comprises the following steps that: (1) adopting a SLIC (Simple Linear Iterative Clustering) superpixel segmentation method to carry out regional division on an original reference diagram and a target diagram; (2) on the basis of a local matching algorithm of adaptive weighting, calcuating an original parallax spatial diagram; (3) carrying out regional parallax plane fitting based on a confidence point; (4) applying a clustering algorithm to combine adjacent regional parallax planes; and (5) on the basis of a superpixel region, constructing an energy cost function to carry out stereo matching. A stereo matching unit is the superpixel region combined on the basis of the edge information of an image, so that the algorithm is guaranteed to be suitable for a large no-texture region, a depth boundary can be accurately positioned, a matching cost better than the matching cost of a traditional stereo matching method is obtained and is combined with a parallax postprocessing method to effectively obtain a high-accuracy parallax diagram, and the method exhibits good instantaneity.

The invention discloses a three-dimensional and four-dimensional automatic ultrasound image optimization and adjustment method. The method comprises the following steps: inputting image data; dividing soft tissue image areas; fitting a gain plane; and carrying out three-dimensional and four-dimensional gain compensation on the images. The method has the following advantages: the weights carried by pixels of the envelop data before logarithmic compression or image data after logarithmic compression are computed according to the statistical parameters, or the soft tissue areas and the non-soft tissue areas are separated by utilizing the signal to noise ratio, then gain plane fitting is carried out on the areas, the gain compensation value is computed and the gain compensation value is used for carrying out three-dimensional and four-dimensional uniformity adjustment on the ultrasound images, thus ensuring the image brightness to be uniform and consistent.

The invention discloses a building point cloud registration algorithm based on dimension reduction, comprising the following steps: respectively acquiring building point cloud data at two sampling points at different angles of view, and numbering all the points in the point clouds; selecting the overlapping areas of the building point clouds at the two angles of view, and carrying out building point cloud plane fitting on the overlapping areas by means of the least square method to obtain a projection surface of the building point clouds; projecting the building point clouds on the building projection surface, and determining the coordinates of the projection points of all the points in the point clouds after the points in the point clouds are vertically projected on the building projection surface; re-sampling the projection points, and determining a two-dimensional image of the building point clouds after dimension reduction; searching same-name points in the two-dimensional image through template matching based on difference of squares; indexing the searched same-name points of the two-dimensional image back to a three-dimensional point cloud according to the mark number; and getting rotation and translation variables by means of the unit quaternion method according to the points in the three-dimensional point cloud indexed based on the same-name points, and applying the rotation and translation variables to the overall building point cloud. By using the method, the efficiency of registration is improved.

The application relates to an object size measuring method and device based on three-dimensional vision technology. The method comprises the following steps: acquiring three-dimensional point cloud data of the object; performing plane fitting on the three-dimensional point cloud data to obtain at least one point cloud plane; determining a plane intersection line at which the plurality of point cloud planes intersect; and calculating the actual size of the object according to the plane intersection line. The method can not only enhance the reliability of the size measurement result, but also improve the efficiency of measuring the size of the object, thereby realizing the rapid measurement of the shape and structure of the packaging box by the object measuring system in the logistics application scenario.

The embodiment of the invention provides an obstacle recognition method and device. The method comprises following steps of acquiring three-dimensional point cloud data in the range that the proceeding ability of a robot is capable of reaching from three-dimensional point cloud data in the proceeding front space of the robot according to the proceeding ability parameters of the robot; carrying out plane fitting on the three-dimensional point cloud data in the range that the proceeding ability of the robot is capable of reaching so as to acquire plane fitting results; and according to the plane fitting results, recognizing obstacle data points in the three-dimensional point cloud data in the proceeding front space of the robot. According to the invention, consumption of system resources can be reduced; and obstacle recognition efficiency is improved.

The invention belongs to the field of computer vision and image measurement technology, more particularly relates to a selection method for a 3-D mosaic match point, the method includes the following steps: (1) a multi-sensor system is globally demarcated by using a target and data of a target feature point are simultaneously obtained; (2) a data point is processed a flat fitting for a same target, and target area are solved through using a projective point and a method of a minimum convex hull; (3) area data of all target areas are recorded for global demarcation; (4) actual measured data of all sensors are obtained when actual measurement is processed, and a solution which is combined with recorded target data is processed for overlap areas of the measured data between the neighboring sensors; (5) the quick 3-D mosaic is processed by using these overlap areas, finally panorama measurement data are obtained. The selection method for a 3-D mosaic match point of the invention is adopted, a determination of the target area is completed in a process of the global demarcation, and none of a hardware device is needed for a system and an unnecessary process is avoided.

The invention discloses a method for performing multi-body depth recovery and segmentation on a video, which comprises the following steps of: (1) performing energy minimization on the video by an iteration method to obtain an initial label of each frame of the video, wherein the initial label consists of the depth of pixels and segmentation information; (2) after performing image segmentation on each frame, optimizing the initial label of each frame of image by a multi-body plane fitting method to obtain the optimized labels of all segmentation blocks in each frame of image; (3) selecting a group of visible frames and a group of invisible frames from adjacent frames for each pixel on each frame by using the optimized labels; and (4) performing the energy minimization on each frame of the video by the iteration method to obtain the iterated labels of each frame of the video, and expanding the progression of the depth of the iterated labels further by using a hierarchical confidence propagation algorithm. By the method, the depth recovery and segmentation can be performed on videos in which multiple rigid objects move.

The invention discloses a laser radar and camera combined calibration method based on a mobile calibration plate. Firstly, calibration internal parameters are calculated for different poses of a calibration plate; a normal vector of a point cloud data fitting plane of the selected corresponding laser radar calibration plate is acquired, and calibration external parameters are calculated through equation solving to finally complete calibration. The method comprises the following specific steps: step 1, collecting laser radar and camera data of calibration plates at different angles in the viewfield of a camera as calibration data and verification data; 2, acquiring a camera internal reference calibration result of the acquired camera data by using a Zhang Zhengyou calibration method; 3, selecting a calibration plate position point cloud on the corresponding laser radar data; step 4, carrying out plane fitting on the point cloud data obtained in the step 3 and obtaining a correspondingnormal vector; and step 5, substituting the internal reference data obtained in the step 2 into a corresponding calculation formula to obtain a final external reference result. Compared with a single-point selection method, the method for selecting the plane data of the calibration plate is higher in accuracy and simpler to operate.

An identification method of a spatial position of a welding torch of a rotary electronic arc sensor comprises following steps: firstly, a coordinate system is found according to the rotation center of the rotary electronic arc and variations of the length of the electronic arc are transformed into variations of the coordinate system. Secondly, the length of the arc can be surfacely fitted by a method of least squares. Thirdly, a deviation and an inclined angle of the welding torch can be identified according to a rate of slope of the intersection line formed by the surface and the surface on which the coordinate system is positioned. The invention has the advantages of: (1) a simple and rapid surface fit method, strong practicability, and satisfying the requirements of the control of the position of the welding torch; (2) the realization of a mutual decoupling of the deviation and the inclined angle of the welding torch; (3) using three dimensional information detected by the rotary electronic sensor to a full extent with a large amount of information and high reliability and veracity of detected results.

The invention provides a method for measuring three-dimensional roughness of a rock mass structural surface based on a triangle area ratio method, and belongs to the field of the geotechnique in civil engineering. A digital photogrammetry system is utilized to acquire the structural surface, the three-dimensional space form of the whole structural surface is reflected according to the surface form of the exposed part of the structural surface, plane fitting is conducted on the structural surface through the least square method, the ratio of the total area of triangular meshes of the structural surface to the plane area is calculated and reflects the smooth degree of the fit structural surface. The advantage of a photogrammetry method that three-dimensional data of the structural surface can be quickly and accurately measured is fully utilized, the surface of the structural surface is reconstructed through a Delaunay triangulation method on the basis, and then the surface form of the structural surface is reflected through the three-dimensional triangular meshes. On the basis of reconstructing the surface of the structural surface in a three-dimensional mode, a specific area method is provided, and the magnitude of roughness is reflected according to the ratio of the area of the meshes of the structural surface to the fitting area.

The invention provides a method for calculating a monocular video depth map. The method is characterized by comprising a step of decomposing a video to be recovered into frames according to a frame, a step of extracting picture feature points of each frame, a step of matching the feature points and forming a feature point trajectory, a step of calculating a global rotation matrix and a translation vector, a step of optimizing camera parameters, a step of calculating dense optical flow of a selected frame, and a step of calculating a depth value of the selected frame to obtain a depth map. According to the method of the technical scheme, a depth estimation method of surface from motion (SFM) based on a physical mechanism is used, and a dense optical flow is used for matching. According to the method, no training sample is needed, optimization modes of segmentation, plane fitting and the like are not used, and the calculation quantity is small. At the same time, according to the method, a problem that the depth values of all pixels can not be obtained especially in a texture-free area in a process from sparse reconstruction to dense reconstruction in the prior art is solved, while the calculation efficiency is improved, the accuracy of the depth map is ensured.

An image matching method and system for use with multiple images of a scene captured from different angles. Image matching is performed by identifying a plurality of segments within at least two images, determining an initial disparity values for pixels in the images and then determining initial disparity planes for the segments by fitting a plane to initial disparity values for the segments. A refined disparity plane set is created by iteratively refitting the disparity planes by using various fitting cost functions and weighted linear systems. A labeling of each segment to a disparity plane is made by minimizing a global energy function that includes energy terms for segment to disparity plane matching as well as penalizing disparity plane discontinuities between adjacent image segments.

The invention provides a point-cloud feature point extraction method based on local sub-neighborhood division. The method includes the four steps of the initial feature point extraction stage, the local triangle construction stage, the local sub-neighborhood division stage and the robust feature point extraction stage. At the initial feature point extraction stage, potential initial feature points of input point cloud data are acquired. At the local triangle construction stage, in order to effectively extract true feature points from an initial feature point set, a triangle set capable of reflecting the local geometrical feature structure of the corresponding point is established in a local neighborhood of each initial feature point. At the local sub-neighborhood division stage, the normal of each constructed local triangle set is clustered, local neighborhood points at one point can be clustered, and thus sub-neighborhood division of the local neighborhood points at one point can be achieved. At the robust feature point extraction stage, local plane fitting is conducted on the data points in the divided sub-neighborhood, and accordingly the true feature points can be recognized by judging whether current points fall on intersection lines of a plurality of planes at the same time or not.

The invention discloses an evaluation method of the tooth profile deviation of an involute cylindrical spur gear under the influence of installation error. Firstly, establishing a machine coordinate system O0-X0Y0Z0 to measure the tested gear under the machine coordinate system. The measuring positions include the upper and lower section points of the measured gear mandrel, upper end of the gear and the measuring point of the tooth profile, so that the tooth profile measuring point under the influence of the installation error is obtained. According to the two cross-section points of the mandrel under the machine coordinate system, conducting the least-square circular fitting to work out rotating shaft equation to establish the coordinate system O1-X1Y1Z. The tooth-profile measuring pointis restored to the initial position and the error of the rotation axis of the tooth-profile measuring point is corrected. The least-square plane fitting is performed on the measuring point on the upper end face of the measured gear, and the coordinate system O2-X2Y2Z2 is established with the normal vector direction of the upper end face of the measured gear being the Z2 axis with the aim of correcting the inclination error of the tooth-profile measuring point. Finally the influence of the installation error on tooth profile deviation evaluation is corrected based on the tooth profile deviationvalue worked out by calculating the rotation angle to the measuring point of the design tooth profile around gear base circle center.

The invention discloses a binocular visual multi-line projection structured light calibration method. The method comprises the following steps that 1) the sequence numbers of light stripes are determined; 2) left and right cameras collect a lattice image, four corners of a lattice are clicked, initial estimated positions of other intersection points in the lattice are obtained, and sub-pixel coordinates of intersection points are obtained by using an Ostu threshold method and a gray scale weighted average method in a k*k neighborhood of the initial estimated positions of intersection points; 3) 3D coordinates of the intersection points are obtained; and 4) the steps 1-3 are repeated, the 3D coordinates of the intersection points obtained in multiple times are classified according to the sequence numbers of the light stripes, for all points belonging to the same structured light plane, a least square method is used to carry out planar fitting on the 3D coordinates of the intersection points belongs to the same structured light to obtain a plane equation, and calibration is not completed until all the structured light planes are in the planar equation of the system coordinate system. The method has the advantages that a calibration board has no requirements for planarization, manufacture is simple, calibration is flexible, no auxiliary equipment is needed, the method is suitable for onsite calibration and the versatility is high.

The invention discloses an obstacle detection system and method based on depth information, and mainly relates to the field of computer vision. The method comprises the following steps of firstly, collecting image information of the environment through a depth camera, and obtaining point cloud information of the environment and objects in the environment through the device; filtering the acquireddepth point cloud information, removing noise points and useless points in the image, and carrying out plane fitting on the acquired depth map; determining a preliminary point cloud set of the obstacle according to the depth map and the plane fitting result; extracting feature points of the point cloud information, and calculating descriptors of the feature points; and performing feature matchingon the descriptor of the scene and the descriptor of the target object to obtain an obstacle detection result in the scene. The method is not affected by environmental changes, color elements are added for detection according to the characteristics of the sensor, and the obstacle detection accuracy is improved.

In a method for reconstructing a curved surface of point cloud data using a computing device, point cloud data, a preset point distance is acquired and defined. A neighborhood point set for each point is calculated. The neighborhood point set of each point is fitted to be a plane, and a normal vector of the plane corresponding to each point is calculated. One or more singularity points in the neighborhood point set of each point is confirmed and corrected. A projection point set of each point is obtained by projecting the neighborhood points in the corrected neighborhood point set to the plane of each point. The projection point set of each point are meshed into triangles and the curved surface is reconstructed by integrating the plurality of triangles corresponding to the projection point set of each point.