101 results about "Plane segmentation" patented technology

The invention discloses an indoor three-dimensional model reconstruction method and system with structural semantics, and the method comprises the steps: collecting indoor three-dimensional point cloud data through a mobile depth scanning device, and carrying out the preprocessing of the three-dimensional point cloud data; extracting a wall surface component of the functional space; obtaining a function space boundary; door and window parts are extracted and parameterized; carrying out plane segmentation on the global point cloud data to obtain a global plane, and optimizing the geometric andincidence relation of the wall surface part and the door and window part according to a plane structure global optimization algorithm; and outputting the parameterized data to an indoor three-dimensional model with structural semantics. According to the method, based on the mobile depth scanning equipment, the low-precision point cloud and the camera track subjected to semantic marking are fused for rapid three-dimensional reconstruction of the indoor model with structural semantics, and better robustness and reconstruction precision are achieved.

The invention relates to an automatic laser scanning data physical plane extraction method with multi-scale characteristics fused. Existing problems solved by the method of the invention mainly comprise: 1, an optimal neighborhood adaptive determination method based on information entropy is developed, and influences on three-dimensional point local geometric characteristic calculation by factors such as point density changes, noise and data missing are overcome; 2, point-based characteristics (such as dimension characteristics and a normal vector) and region-based characteristics (such as roughness, compactness, a scale and a length-width ratio) are used comprehensively to carry out extraction of a physical plane region, and thus accuracy of physical plane extraction is improved; and 3, the existing plane segmentation method application range is expanded. The method of the invention is applicable to physical plane extraction for laser scanning data with an airborne type, a vehicle-mounted type, a fixed station type and the like. The overall technical flow chart is as follows.

The invention provides a cargo pallet detection method and system based on point cloud plane contour matching. The method includes the following steps: S1, original point cloud filtering; S2, point cloud plane segmentation; S3, generating a grid map through projection; S4, contour extraction; S5, contour matching; and S6, acquiring the relative position and posture. The system includes an original point cloud filtering module, a point cloud plane segmentation module, a module for generating a grid map through projection, a contour extraction module, a contour matching module, and a relative position and posture acquisition module. The method and the system are of low cost and high detection accuracy, and are basically not affected by illumination. There is no need for label laying or other operation, and there is no need to transform the storage environment. The method and the system are suitable for pallets of various colors and specifications, and have the advantages of high accuracy, low price, strong robustness, and the like.

The invention provides an optical image and SAR image automatic registration method within multilevel multi-feature constraint. The optical image and SAR image automatic registration method within the multilevel multi-feature constraint comprises the following steps that optical images and SAR images are preprocessed, multi-scale level set segmentation is conducted, and a plane segmentation result is obtained; when similar plane targets exist, a coordinate set of centroid points of areas seemingly provided with the same name is calculated; when similar plane targets do not exist, multi-scale analysis is conducted on the images by means of wavelet transformation, extraction of lower-layer linear characteristics and point set matching are conducted on the thickest image, and lower-layer registration transformation parameters are extracted; high-layer linear characteristics is extracted, a control-point matching degree function is defined according to the lower-layer transformation parameters, and high-layer point set matching is conduced; finally, a matched point pair is precisely judged out through the KNN image from the structure, a wrong matched point pair is eliminated, transformation parameters of the matched point pair are obtained according to a polynomial transformational model, and a final registration result is obtained.

The invention relates to a point cloud data automatic registration method and device based on plane features, and belongs to the technical field of three-dimensional laser scanning. The method comprises the following steps: firstly, performing plane segmentation on point cloud data through a plane segmentation algorithm to obtain plane pieces; calculating attribute information of each plane; establishing a corresponding relation between the plane pieces through the attribute information of the plane pieces, the inter-plane mutual relation and the geometric constraint of rotation and translation; and obtaining a corresponding plane pair set for solving the coordinate conversion parameters, solving the coordinate conversion parameters by utilizing the plane parameters, and selecting an optimal solution, namely a final registration result, according to the established point cloud registration overall consistency measurement. The method comprehensively utilizes the plane attribute information, the inter-plane constraints and the spatial geometric constraints to ensure the accuracy and high efficiency of the registration process, does not depend on additional intensity or color information, can effectively perform registration on point cloud data acquired by various platforms, and is high in applicability.

The invention provides a real-time object detection method based on combination of three-dimensional point cloud segmentation and local feature matching. The method comprises the following steps that: a normal vector of each point in a three-dimensional point cloud is calculated rapidly by combining an integral image so as to carry out horizontal plane segmentation on an object; the object is segmented; target object feature extraction is carried out; and the extracted target object feature is compared with an object feature database. According to the method provided by the invention, object detection and positioning are realized by rapid segmentation of the three-dimensional point cloud; and with an SURF-feature-matching-based method, target object identification is realized. Because of the integrated structure characteristics of orderly point clod data, a method combining the integral image with regional growth is provided, so that rapid three-dimensional object detection is realized.

The invention discloses a pallet recognition and positioning method based on an RGB-D sensor. S1. Acquiring a depth image through a sensor; S2. Carrying out plane segmentation on the point cloud of the depth image to obtain one or more planes to form a plane Set; S3. Determine the relevant planes that may contain pallets from the plane set; S4. Match the relevant planes according to the preset pallet template, identify the position of the pallet in the relevant planes and perform positioning. The invention has the advantages of high recognition efficiency, precise positioning and the like.

The invention discloses a semantic SLAM method and system based on object and plane features, and belongs to the technical field of computer vision, and the method comprises the steps: obtaining the RGB-D image stream of a scene; performing frame-by-frame tracking on the RGB-D image stream to obtain a key frame image; constructing a local map about the scene by utilizing the key frame image; carrying out plane segmentation on the depth map of the key frame image to obtain a current plane, constructing a global plane map by utilizing the current plane, carrying out object detection on the key frame image to obtain a detection frame and confidence, reconstructing point cloud of an object by utilizing the detection frame and the confidence, and merging feature points in the detection frame into the object to obtain a global object map; and performing loop-back detection by using the key frame image to obtain a loop-back frame, and performing loop-back correction and optimization on the plane constraint and the object constraint by using the loop-back frame to obtain a plane map and an object map of the scene. According to the method, the SLAM optimization performance can be improved,and the semantic description of the environment is enhanced.

The invention provides an unmanned aerial vehicle cluster cooperative task planning method for air maneuvering combat, and belongs to the technical field of unmanned aerial vehicle task planning. Themethod comprises the following steps of: firstly, establishing a task planning model taking the minimum flight range, the minimum fixed cost, the maximum target income and the minimum threat cost of an unmanned aerial vehicle for executing a task as target functions; and then mapping the task planning optimization process into the life cycle growth and development process by referring to the biological life cycle characteristics, and forming a track point set by one point on each segmentation plane by adopting a space plane segmentation method. The method is suitable for an unmanned anti-radiation attacker to execute attack tasks in a complex task environment under a three-dimensional map, and can also be used for multi-unmanned-aerial-vehicle cooperative task planning of reconnaissance and reconnaissance tasks. The technical problems to be solved by the invention are as follows: simulating a real task scene, adding all dynamic constraints of the unmanned aerial vehicle, and obtainingmulti-unmanned-aerial-vehicle cooperative task distribution and track planning meeting complex constraints in a three-dimensional environment.

The invention discloses a calibration method and device for a hand-eye relation of a polishing operation arm. The calibration method comprises the steps that collected three-dimensional point cloud data of the surface of a calibration block is acquired; preprocessing is carried out on the three-dimensional point cloud data to remove corresponding noise points; a consistency algorithm is adopted tocarry out plane segmentation on the three-dimensional point cloud data with the noise points being removed, and an estimation planar model is obtained; according to a preset convex hull algorithm, the estimation planar model is solved, and a corresponding convex polygon is obtained; and by fitting vertexes of the convex polygon, an outer parameter rotation translation matrix RW(Cam) of a camera is calculated so that a transformational relation MCam(Tool) of a three-dimensional scanning camera coordinate system relative to the robot actuator tail end in robot hand-eye calibration can be determined. The technical problems that traditional visual sensor 3D calibration is not suitable for rapid automatic identification required by a polishing system is solved.

The invention relates to a camera and laser radar joint calibration method based on an L-shaped calibration plate. The method includes: firstly, installing a camera and a laser radar on to-be-calibrated equipment; secondly, placing the L-shaped calibration plate on the ground in the field of view of equipment to be calibrated; starting a camera and a laser radar, carrying out data collection, and obtaining an image containing the L-shaped calibration plate and point cloud data; performing angular point detection on the image data to obtain pixel coordinates of feature angular points on two planes of the L-shaped calibration plate; performing plane segmentation and fitting on the point cloud data to obtain an equation of two planes of the L-shaped calibration plate, and further obtaining coordinates of feature angular points on the planes in a laser radar coordinate system through geometric information; and finally, on the basis of the pixel coordinates of the detected angular points and the coordinates in the laser radar coordinate system, calculating and obtaining pose changes of the camera and the laser radar. According to the method, the position of the angular point feature in the point cloud data can be accurately obtained, and the accuracy of joint calibration of the camera and the laser radar is greatly improved.

The invention discloses an on-line diagnosis method of mechanical fault for a gas insulated switchgear, and belongs to the technical field of power transmission and transformation equipment conditionmonitoring. The method takes the opening and closing operation of a breaker as an excitation source, a sensor can be placed on a to-be-monitored member, and vibration signals can be extracted; and S-transform energy equalization based vibration signal time-frequency plane segmentation thought needs to determine an appropriate segmentation ruler. The method can divide a frequency sub-band and a time sub-band and extract the energy values of each characteristic subinterval under a normal and fault state by adopting a normalized vibration frequency energy density spectrum and instantaneous energydensity spectrum; and the energy values of the characteristic subinterval can be taken as a data set, and classification and identification can be performed on faults by adopting an SVM optimized bya genetic algorithm. The method can effectively diagnose the GIS mechanical faults and enhance the safety and reliability of GIS operation, and has important theoretical and practical values.

The invention discloses a lateral multi-plane segmentation-based oblique photography monomer object extraction method. The method comprises the following steps of: selecting a subordinate tile set ofa terrain object according to a user selection / extraction range, and loading the tile set in a scene; constructing a multi-plane set: generating a lateral plane according to terrain boundaries sketched by a user under different human eye optimum view angles, under each view angle, drawing sets of boundary lines of the terrain object to form a segment set, and constructing planes according to triangles so as to obtain the multi-plane set; dividing data layers according to LOD, under each data layer, cutting a TIN surface of an original photography product obtained via oblique photography through the multi-plane set, and carrying out separation and reconstruction to generate independent TIN surfaces of a monomer object; under each data layer, respectively reconstructing terrain objects and adding textures according to the independent TIN surfaces of the monomer object; and storing a newly generated monomer object as standard oblique photography. According to the method, the problem of object edge aliasing in oblique photography monomerization is solved.

The invention discloses a point cloud plane segmentation method based on quick adjacency voxel query, which comprises the following steps: constructing an octree according to point cloud data, puttingeach point into a corresponding octree node according to a coordinate value, wherein each node corresponds to a voxel; coding nodes of the octree by adopting a binary coding mode, and quickly querying adjacent voxels according to voxel codes; calculating the significance characteristic of each voxel; carrying out coarse segmentation on voxels by using a region growing algorithm; and performing fine segmentation on each point in the unsegmented voxel according to the distance from the point in the unsegmented voxel to the nearest plane in the adjacent voxel. According to the method, the problems of over-segmentation and under-segmentation of the point cloud plane are effectively reduced, the recall rate and accuracy of point cloud plane segmentation are improved, the influence of point cloud noise is small, and the calculation efficiency is high.

The invention discloses a vision-based target object identification and positioning method, a terminal and a storage medium, comprising: scene image data and scene point cloud data of a current frameof a depth camera are acquired; image recognition is carried out for the scene image data, and the target object is recognized; the target point cloud data corresponding to the target object is distinguished from the scene point cloud data; plane segmentation is carried out for the points contained in the target point cloud data, and the planes contained in the target object in the current frame are extracted; the target surface is selected from the extracted planes; the posture of the target object is calculated based on the target surface. The technical scheme of the invention adopts first image recognition, then the point cloud is analyzed and processed, and finally the complex object is accurately identified and located by calculating the pose of the target surface with features, whichavoids the complex calculation. The method has the advantages of small calculation amount, high computational efficiency and strong practicability.

The invention discloses a laser LiDAR power line point cloud automatic point supplementing method. The method comprises the steps of 1 performing power line point cloud grading and coordinate conversion to obtain a power line point cloud set; 2 performing single-gear power line point cloud plane segmentation and gridding; and 3 performing point supplement calculation, namely performing point supplement on the power lines in the plane segmentation subsets. The technical problems of low laser LiDAR power line point cloud density and serious point cloud loss are solved, the workload of manual data processing is reduced, the point location precision of point supplement is improved, and data support is provided for subsequent functional applications such as power line reconstruction and dangerous point detection.

The invention provides an airborne point cloud roof plane segmentation method based on an octree and boundary optimization. Multi-scale initial plane patch extraction based on an octree is carried out, and the method comprises the following steps: iteratively dividing the non-sequence point cloud into plane patches with different scales according to a spatial position according to a data structureof the octree, and selecting a plane patch with planarity from the plane patches as an initial plane patch obtained by data preprocessing; aggregating the initial plane pieces according to the adjacent relationship and the parameter similarity through hierarchical clustering to form an initial plane; merging the points which do not belong to the planar sheet into an initial plane through regionalgrowth; and plane boundary point re-classification based on energy optimization comprises the steps of converting a plane boundary point re-classification problem into an energy optimization problemthrough an energy function, optimizing a plane boundary obtained by region growth, and obtaining a final plane segmentation result. According to the method, the problems of poor robustness and inaccurate boundary of seed points growing in a region in point cloud plane segmentation are solved, and an optimal plane segmentation result is obtained.

The invention discloses a plane segmentation method based on superpixels and a graph convolutional network. The method comprises the following steps: preprocessing an input color image to obtain a preprocessed image; then, according to the image resolution, segmenting the preprocessed image into a proper number of superpixels, and converting the superpixels into an undirected graph structure; constructing a graph convolutional network, and training the graph convolutional network by using a data set; and finally, predicting a graph formed by the superpixels by using the trained graph convolutional network, and carrying out plane classification on each superpixel so as to complete plane segmentation. According to the method, the image is segmented into the superpixels, so that the edge information in the original image can be well reserved, the learning burden of a subsequent image neural network is reduced, and the problem that the segmented plane edge is too large in gap with the actual situation is prevented; and a specific label is extracted through a specific algorithm by utilizing an existing data set and serves as a training set for subsequent supervised neural network learning, so that the problem that the data set is not segmented for a plane is solved, and huge cost of manual marking is avoided.

The present invention relates to a road image segmentation method based on a normal feature. A binocular camera is used to obtain a depth image which, in combination with the internal reference of the camera, is converted into a planar normal image. A road plane is determined on the planar normal image by using the normal characteristics to complete the road plane segmentation processing. The segmented road plane is marked as a safe driving area, and the remaining area is marked as an anti-collision area.

The invention provides a method for quickly extracting an indoor three-dimensional line segment structure based on point cloud data. The method comprises the steps of: measuring and collecting point cloud data of an indoor building, subjecting the point cloud data to projection, and forming two-dimensional projection point cloud data; extracting line segments in the two-dimensional projection point cloud data, calculating a spatial geometric equation of a vertical face through inversion by utilizing coordinates of end points of the line segments, and then acquiring a facade through segmentation; removing facade points from the two-dimensional projection point cloud data, performing down-sampling on the remaining points, acquiring a horizontal plane through segmentation, and combining the facade and the horizontal plane to obtain a final plane segmentation result; for each extracted plane in the plane segmentation result, fitting precise parameters of the plane, projecting points on theplane to the fitted fitting surface, and extracting edge points of each plane to obtain an edge point set; and setting the farthest distance from a point to a straight line, extracting three-dimensional line segments, and further performing line segment combination according to the parallelism, the colinearity and the coincidence between the line segments to obtain a final three-dimensional linesegment structure extraction result.

The invention discloses a semi-automatic segmentation 2D-3D sequential optimization registration method. The method comprises the steps that a CT sequence containing bone tissue is traversed, corner points are extracted through a Curveur Scale (CSS) algorithm, the bone tissue range is iteratively expanded, and plane segmentation is utilized to obtain volume data of all segments; a sequential optimization registration method is adopted, the initial registration parameters are obtained in a solution space according to the sequence, and finally a step acceleration method is used in a very small range for optimization search. According to the present invention, the problems that the preoperative CT and intraoperative X-ray shooting time are different, and the bone tissues generate the relativemotion, and the registration time is increased along with the increase of the number of segments, are solved, and the time complexity is reduced to O (6n).

The invention provides a box-type structure detection method and device based on artificial intelligence, electronic equipment and a computer readable storage medium. The method comprises the following steps: carrying out feature extraction processing on a to-be-detected image comprising a box-type structure to obtain the features of the box-type structure; performing plane segmentation processingon the to-be-detected image based on the features of the box-type structure to obtain a segmentation result of a frame in the corresponding box-type structure in the to-be-detected image; performinglinear detection processing on the to-be-detected image based on the characteristics of the box-type structure to obtain an edge line of the frame; determining a plane area included in the frame basedon the segmentation result and the edge line of the frame; and determining a detection result of the box-type structure based on the plane area included in the frame. According to the invention, thesegmentation result of the frame in the box-type structure and the edge line of the frame can be combined, and the detection accuracy of the box-type structure is improved.

The invention provides a monocular dense simultaneous localization and mapping (SLAM) method. A faster and more robust plane area detection algorithm is adopted, and highly stable and reliable plane areas are found out before reconstruction, then the reconstruction is performed, and the time is greatly reduced; before the reconstruction, the highly stable and reliable plane areas are found out, only these areas are reconstructed, only a small part of reconstruction results will be discarded in the next filtering process, and the reconstruction efficiency is greatly improved; due to the presence of noise in the plane area detection algorithm, a multi-plane segmentation method is used to reconstruct the plane areas; the monocular dense SLAM method of the invention can run in real time without using a GPU; and on a TUM data set, the dense reconstruction algorithm has a running time of only about 35% of the DPPTAM, and this algorithm is advantageous over the DPPTAM in terms of map precision and integrity rate.

The embodiment of the invention discloses a vehicle driving information method and device based on plane segmentation and a vehicle-mounted terminal. The method comprises the steps that a to-be-detected vehicle area in a road image frame is input into a vehicle plane division model, and the vehicle plane division model determines segmentation information used for segmenting planes in all directions of a to-be-detected vehicle according to pre-trained model parameters; when determining that the to-be-detected vehicle area has a transverse plane according to the segmentation information, segmenting the to-be-detected vehicle area according to the segmentation information to obtain the transverse plane in the road image frame; and selecting a first corresponding feature point from the associated transverse plane between the road image frame and the previous road image frame, and determining the vehicle speed information of the vehicle to be detected relative to the current vehicle according to the position difference between the first corresponding feature points and the time difference between the road image frame and the previous road image frame. By applying the scheme provided bythe embodiment of the invention, the vehicle driving information of the to-be-detected vehicle can be determined more accurately.

The embodiment of the invention relates to the technical field of environmental perception, and discloses a stair detection method and device, and the method comprises the steps: obtaining the point cloud data around a current scene where a mobile robot is located, determining the ground in the current scene according to the point cloud data, and building a ground coordinate system; performing plane segmentation on the point cloud data to determine a plane in the current scene, classifying the plane in the ground coordinate system, screening out a horizontal step surface in the plane, and finally according to the horizontal step surface, establishing and outputting the visualized three-dimensional model of the stair structure and the spatial pose information of the three-dimensional model.According to the stair detection method provided by the embodiment of the invention, the mobile robot can accurately detect the pose of the stair structure in the current scene and establish the three-dimensional model of the stair structure.

The invention provides an object measurement method, a virtual object processing method, an object measurement device, a virtual object processing device, a computer readable storage medium and electronic equipment, and relates to the field of measurement. The method comprises the steps of obtaining a depth image of a scene, and determining point cloud data of the scene according to the depth image; performing plane segmentation processing on the point cloud data to determine surface information of an object in the scene; determining three-dimensional vertex coordinates of the object in combination with the surface information of the object; and calculating measurement parameters of the object by utilizing the three-dimensional vertex coordinates of the object. Furthermore, a virtual object associated with the measurement parameter may also be generated and displayed on the electronic device for viewing by the user. The accuracy of object measurement can be improved.

The invention relates to a central circle apparatus in a plane segmentation production-line rotation device, belonging to the technical field of mechanical equipment. The central circle apparatus is characterized in that two to eight groups of rotation seat components are evenly arranged on the circumference of a rotation circle. The rotation seat components comprise a permanent seat, a guide lever, a lifting seat, a pneumatic bag type cylinder, an inside locating wheel, a small rotary drum, an outside locating wheel, and other components. The guide lever and the pneumatic bag type cylinder are arranged on the permanent seat; the lifting seat is sleeved on the guide lever; the inside locating wheel, the small rotary drum and the outside locating wheel are respectively arranged on the lifting seat. The central circle apparatus has simple, compact and reasonable structure, can effectively solve the off-tracking problem of plate array rotation, can improve the safety and the reliability of the plate array rotation, can reduce safety incipient fault, can reduce the working intensity and improve the working efficiency, can be used conveniently and is suitable for rotating and centering of large-sized plate array.

The invention discloses a multi-modal region-consistent significance object detection method based on foreground and background priori. The method comprises the following steps: 1) obtaining a color image and a depth image of a scene, while generating a point cloud; 2) using the color image and the depth image for the scene segmentation; 3) using an object region-based detection method for detecting all regions of the image to obtain a focal region, taking the focal region as foreground priori in the color image; 4) detecting a plane region of the point cloud as a background priori, obtaining point cloud data from the color image and the corresponding depth image, detecting a plane structure in the point cloud data, then calculating the boundary length of a scene connection boundary, calculating a background connectivity of plane segmentation, and obtaining over-segmentation background weight; 5) calculating to obtain a global region contrast SG(rk), obtaining a significance map through the formula below: S(rk)=SF(rk)*SB(rk)*SG(rk). The multi-modal region-consistent significance object detection method effectively improves the accuracy.

The invention discloses a mark point multi-angle scanning method, system and device. The method comprises the steps: obtaining the mark point data of a scanned mark point at different visual angles, carrying out the reconstruction of the mark point data, determining a mark point plane according to the mark point reconstruction data obtained through reconstruction, and dividing the mark point planeinto at least two angle regions; in the continuous scanning process at different visual angles, projecting the center of the scanner instrument scanned each time to a marking point plane, and calculating the number of times of falling into each angle area; counting the number of times of falling into each angle area in real time, and completing scanning when the number of times meets a preset threshold value of a corresponding angle; after scanning is completed, data structure conversion is conducted on the mark point reconstruction data, global optimization is conducted on the converted markpoint reconstruction data through a bundle adjustment method, and an obtained global optimal solution serves as a scanning result to be output. According to the invention, the splicing precision of the overall mark points can be improved, so that laser scanning can be better applied to the field of industrial measurement.

The invention provides an airborne building LiDAR point cloud-based facade line automatic extraction method and equipment. The method comprises the following steps: step 1, carrying out monomer segmentation on LiDAR point clouds of buildings to obtain point clouds of all the buildings; step 2, according to the point cloud of each building obtained in the step 1, utilizing a normal constraint plane segmentation method to extract building facade points; step 3, according to the building facade points obtained in the step 2, performing horizontal projection and straight line fitting to obtain facade lines, performing pairwise intersection processing on the facade lines, and extracting facade grids by adopting a minimum ring search method; and step 4, for the facade grids obtained in the step 3, completing extraction of effective facade line segments in combination with building point cloud distribution characteristics. The method can automatically extract the vertical face line of the multi-storey building, does not have field measurement work, effectively improves the working efficiency, and improves the measurement precision compared with a manual measurement means.