317 results about "Point cloud processing" patented technology

The invention discloses a scattered workpiece recognition and positioning method based on point cloud processing, and the method is used for solving a problem of posture estimation of scattered workpeics in a random box grabbing process. The method comprises two parts: offline template library building and online feature registration. A template point cloud data set and a scene point cloud are obtained through a 3D point cloud obtaining system. The feature information, extracted in an offline state, of a template point cloud can be used for the preprocessing, segmentation and registration of the scene point cloud, thereby improving the operation speed of an algorithm. The point cloud registration is divided into two stages: initial registration and precise registration. A feature descriptor which integrates the geometrical characteristics and statistical characteristics is proposed at the stage of initial registration, thereby achieving the uniqueness description of the features of a key point. Points which are the most similar to the feature description of feature points are searched from a template library as corresponding points, thereby obtaining a corresponding point set, andachieving the calculation of an initial conversion matrix. At the stage of precise registration, the geometrical constraints are added for achieving the selection of the corresponding points, therebyreducing the number of iteration times of the precise registration, and reducing the probability that the algorithm falls into the local optimum.

The invention relates to the field of point cloud processing, and specifically discloses a three-dimensional point cloud road boundary automatic extraction method. The method comprises the steps of S1, screening seed points for the whole acquired three-dimensional point cloud data set P so as to perform super voxel classification; S2, extracting boundary points between adjacent non-coplanar super voxels by using an alpha-shape algorithm; S3, extracting road boundary points by using a graph cut based energy minimization algorithm; S4, removing outliers based on an Euclidean distance clustering algorithm; and S5, fitting the extracted road boundary points into a smooth curve. The method can directly operate on the large-scale three-dimensional point cloud, can be applied to different scenes, is high in calculation speed and good in algorithm robustness, and can extract road boundaries quickly.

The invention discloses a method and system for processing point cloud data. The method includes: calculating distance of a current point in point cloud after topology construction and neighborhood points, mean values mu of the distance and a standard deviation sigma for expressing data dispersion degree; judging whether the distance between one neighborhood point and the current point p is in an interval of mu+-alpha-sigma, if on yes judgment, neglecting, otherwise deleting the point; building a bounding box data structure used for expressing point cloud grid space for the point cloud, and calculating distance between a central point p0 of a current bounding box and other points in the current bounding box, a mean value mu0 of the distance and a standard deviation sigma0; judging whether the distance between one point in the current bounding box and the central point p0 is in an interval of mu0+-alpha0-sigma0, if on yes judgment, deleting the point, otherwise neglecting. The point cloud processing method improves accuracy of noise point filtration and precision of point cloud data and effectively reduces redundancy of the point cloud data.

The invention discloses a robot grinding system with three-dimensional vision and a control method thereof. The robot grinding system comprises a workpiece fixing device, a grinding robot, an opticaldevice, pneumatic grinding equipment, a power supply and a control device. The control method of the robot grinding system with the three-dimensional vision comprises the steps that the grinding robotdrives the optical device to horizontally scan a curved surface workpiece; a laser image collected by a 3D camera is transmitted to the control device in the power supply and the control device; thecontrol device performs the three-dimensional point cloud processing to the image obtained by the 3D camera by using a corresponding function, and the preliminary filtering denoising is carried out; the point cloud segmentation and a slice algorithm are used for performing the further processing on the obtained three-dimensional point cloud; and processing results are analyzed and sent to the control device according to rules, and the robot is driven to perform the wholly grinding operation on the curved surface workpiece in the correct attitude. The robot grinding system with three-dimensional vision and the control method thereof are mainly used for automatic grinding in the field of curved surface structure manufacturing, and the production efficiency and production quality are improvedby improving the automation degree of the industrial grinding field.

The invention provides a swing laser radar-based indoor three-dimension point cloud map generation system. The system includes a three-dimension point cloud data acquisition module and a data processing module; the three-dimension point cloud data acquisition module comprises a laser radar and a swing implementing mechanism, wherein the laser radar is used for acquiring point cloud data of the surrounding environment of a mobile platform, and the swing implementing mechanism is used for carrying the laser radar, driving the laser radar to swing in a reciprocating manner and sending swing angle data to the data processing module; and the data processing module, adopted as a point cloud data processing and storage device, controls the start / stop state of the three-dimensional point cloud acquisition module, merges the scanning data of the laser radar and the swing angle data so as to generate indoor three-dimension point cloud data, splices the indoor three-dimension point cloud data to generate an indoor three-dimension point cloud map, and stores the indoor three-dimension point cloud map. The swing laser radar-based indoor three-dimension point cloud map generation system of the invention has the advantages of structural simplicity, easiness in implementation, high performance-to-price ratio as well as high density and high precision of generated point cloud map data. The generated three-dimensional point cloud map can be directly applied to various application situations of indoor navigation and positioning.

The invention discloses a method for quickly acquiring an object surface model, which comprises the following steps of: (1) automatically controlling the movement of a bearing platform and acquiring qualified sequence images through software; (2) accurately marking a laser plane, a camera parameter and a bearing platform movement space through coordinate of each marking point of a marking block, position of a laser line on the image and relative relationship between an object space point and the laser line; (3) accurately extracting the image coordinate sequence of the laser line; (4) reading the marking parameter, and obtaining a three-dimensional coordinate point cloud of an object by using forward intersection of triangulation principle; and (5) obtaining a complete three-dimensional model by using Polyworks equal point cloud processing software for editing. The method can quickly reconstruct the three-dimensional surface model of the object, and can judge whether to meet the requirements by using the quickly generated accurate three-dimensional model.

The invention discloses a three-dimensional shoe last reconstruction process based on the scanning input, which comprises scanning a last by a three-dimensional scanner, generating a point cloud data model through the computer data processing, conducting the pre-processing for the point cloud data model to obtain an initial shoe last model, carrying out the automatic character identification, generating all sub face sheets of the shoe last through the surface slicing technology, stitching all the sub face sheets to form the surface of the whole shoe last, thereby gaining the shoe last model after the reconstruction. The invention transforms the shoe last data into a grid model after the point could processing, automatically obtains each character point and each character line of the shoe last through the automatic character identification technology, and eliminates the problems brought by the losses or damages of original character information files. The position with larger surface curvature change is a difficult point to generate a curved surface in the process of generating curved surface of the shoe last, and the surface slicing technology can be adopted to reduce the difficulty of curved components, and sub face sheets in different regions are adopted with appropriate curvature reconstruction methods, thereby guaranteeing the surface accuracy of the shoe last model.

A dynamic road surface detecting method based on a three-dimensional sensor is provided. The three-dimensional sensor receives a plurality of laser-emitting points reflected from a road surface to generate a plurality of three-dimensional sensor scan point coordinates which is transmitted to a point cloud processing module. The point cloud processing module transforms the three-dimensional sensor scan point coordinates to a plurality of vehicle scan point coordinates according to a coordinate translation equation, and then transforms a plurality of vehicle coordinate height values of the vehicle scan point coordinates to a road surface height reference line according to a folding line fitting algorithm. An absolute difference of two scan point height values of any two adjacent scan points on each of the scan lines is analyzed to generate a discontinuous point. The point cloud processing module links the discontinuous points to form a road boundary.

A system comprises an encoder configured to compress attribute information and / or spatial for a point cloud and / or a decoder configured to decompress compressed attribute and / or spatial information for the point cloud. To compress the attribute and / or spatial information, the encoder is configured to convert a point cloud into an image based representation. Also, the decoder is configured to generate a decompressed point cloud based on an image based representation of a point cloud. A processing / filtering element utilizes occupancy map information and / or auxiliary patch information to determine relationships between patches in image frames and adjusts encoding / decoding and / or filtering or pre / post-processing parameters based on the determined relationships.

The present invention relates to the technical field of depth learning, image processing and three-dimensional point cloud processing, and more particularly, to an object recognition and registrationmethod under an event-triggered camera and a three-dimensional laser radar fusion system. The invention provides a system for fusing the data of an event-triggered camera and the data of a lidar. Universal object detection is carried out by YOLO3 depth learning neural network. The minimum filter is used to fuse the depth information of lidar to detect the object and its depth information in real-time and accurately.

The invention relates to a Kinectv2-based complete object real-time three-dimensional reconstruction method. The method comprises data acquisition, depth completion, point cloud processing, ICP pointcloud registration, point cloud fusion and surface reconstruction. The beneficial effects of the invention are as follows: the Kinect2.0-based object complete real-time three-dimensional reconstruction method comprises the steps of data acquisition, deep completion, point cloud processing, ICP point cloud registration, point cloud fusion, curved surface reconstruction and the like, the effect of the method on complete real-time three-dimensional reconstruction of the object is improved; the defects that a plurality of pictures, monocular cameras or binocular cameras are generally used for depth acquisition in traditional three-dimensional reconstruction, the calculation amount in the process is large, real-time performance and precision are difficult to guarantee at the same time, professional high-precision three-dimensional scanning equipment is too high in price, and the application and popularization degree of the three-dimensional reconstruction technology is limited are overcome.

The invention provides a semantic segmentation and point cloud processing combined plant recognition and model construction method. The method comprises the following steps: 1, generating an orthoimage according to a landscape image obtained by oblique photography; 2, training a deep learning network, and performing semantic segmentation on the orthoimage by a neural network to identify a plant region; 3, generating a point cloud corresponding to the image, and realizing coordinate correspondence between point cloud data and the orthoimage through coordinate system conversion; 4, segmenting the point cloud data to obtain a plant area point cloud; 5, in combination with oblique photography images and point cloud data, plant species are further recognized through k-means point cloud clustering, target detection and other methods; 6, establishing a plant model library; 7, processing the point cloud of the plant area, determining parameters including plant types, positions, sizes and the like, and importing a plant model to replace the point cloud; and 8, converting the plant model into a required format. According to the invention, efficient and accurate recognition of plant species and construction of a three-dimensional plant scene with a sense of reality can be realized.

The invention relates to a three-dimensional space map construction method based on a Kinect vision technology. The three-dimensional space map construction method comprises the following steps: step S1, obtaining a three-dimensional coordinate of each point of a view field space based on a Kinect imaging technology of a light coding to form a single-width point cloud; carrying out point cloud processing to obtain point cloud data; step S2, unifying the point cloud data to the same coordinate system based on a point cloud alignment technology of an ICP algorithm; step S3, lifting a point cloud reading-writing speed and reducing a data storage space based on a point compression and visualization technology of Octree; and step S4, obtaining a disordered dispersed point cloud to form a model of a space object. The three-dimensional space map construction method has the advantages that the cost and the energy source consumption are greatly reduced; Kinect is provided with a RGB (Red, Green and Blue) colored camera so that a constructed map has more abundant color and texture information; the trouble of manually calibrating is avoided and the three-dimensional reconstruction efficiency is improved; the three-dimensional space map construction method can be applied to the fields of robot navigation, digital cities, three-dimensional measurement and the like.

The invention relates to an object six-degree-of-freedom pose estimation method based on color and depth information fusion. The object six-degree-of-freedom pose estimation method comprises the following steps of: acquiring a color image and a depth image of a target object, and carrying out instance segmentation on the color image; cutting a color image block containing a target object from thecolor image, and acquiring a target object point cloud from the depth image; extracting color features from the color image block, and combining the color features to the target object point cloud atthe pixel level; carrying out point cloud processing on the target object point cloud to obtain a plurality of point cloud local region features fusing the color information and the depth informationand a global feature, and combining the global feature into the point cloud local region features; and predicting the pose and confidence of one target object by means of each local feature, and taking the pose corresponding to the highest confidence as a final estimation result. Compared with the prior art, color information and depth information are combined, the object pose is predicted by combining the local features and the global features, and the object six-degree-of-freedom pose estimation method has the advantages of being high in robustness, high in accuracy rate and the like.

The invention provides a preparation method of an indoor occupation grid map based on RGB-D information, and relates to the technical field of robot indoor navigation. The preparation method of the indoor occupancy grid map based on the RGB-D information comprises the steps of loading a global three-dimensional point cloud map, performing ground and non-ground point cloud segmentation, performingground and non-ground point cloud processing, generating a two-dimensional occupancy grid map and the like. According to the preparation method, the precision of the prepared indoor map is approximateto that of a laser radar type generation method, and the cost is lower than that of a laser radar, and the practical value is higher; and the generated two-dimensional occupation grid maps with different height levels can meet the navigation requirements of robots with different working heights.

The invention discloses a point cloud processing method and device and a computer readable storage medium. Specifically, the target area of the obtained first point cloud data is segmented through atleast two segmentation surfaces to obtain second point cloud data, then the second point cloud data is projected to a preset projection plane according to a preset direction to obtain third point cloud data, and finally, the third point cloud data is denoised through a plurality of segmentation lines to obtain fourth point cloud data. According to the scheme provided by the embodiment of the invention, when the three-dimensional point cloud data is processed, most of the noise point data can be quickly removed, so that the processing effect of the noise point data is improved.

The invention discloses a free-form surface part processing system based on the multi-sensor integrated measurement technology. A non-contact type sensor and a contact type sensor are integrated to form a surface measuring component, a surface processing component is provided with a milling cutter, and the surface measuring component is connected with the surface processing component by a linear guide rail. A point-cloud processing component is used for realizing geometric processing for point-cloud data obtained by the non-contact type sensor, and provides processing G codes which can be directly actuated by a current work table to the surface processing component for processing. The surface measuring component measures a workpiece in a non-contact manner, and then measures a product which is obtained by means of fine processing in a contact manner. A quality detecting component is used for comparing errors of measured data obtained by the contact type sensor and obtains results of the quality of the product. The non-contact type sensor and the contact type sensor are integrated, an integrated function of 'measuring, processing and detecting' is realized on the same machine tool, manual intervention is omitted in a total process, and processing efficiency and automation degree are improved.

A point cloud data processing device comprises: a non-plane region removing unit (101) for removing point cloud data involved in a non-plane region, which cause a large computational load, from point cloud data associating a two-dimensional image of an object to be measured with three-dimensional coordinate data of a plurality of points constituting this two-dimensional image; a plane labeling unit (102) for assigning a label for specifying a plane to the point cloud data obtained after removing the data of the non-plane regions; a contour calculation unit (103) for calculating the contour of the object by using a local plane based on a local region continuing from the plane to which the label is assigned; and a point cloud data reacquisition request processing unit (106) for requesting for the reacquisition of point cloud data in order to enhance accuracy.

The invention relates to a rapid 3D face point cloud generation method and device based on a structured light module, and the method comprises the steps: S100, recognizing and positioning an image region where a face is located through an RGB camera or an infrared camera of a structured light module, and marking a face region in the RGB image or the infrared image; S200, opening an infrared codingprojector of the structured light module, acquiring an infrared coding image by an infrared camera, and mapping the face region into the infrared coding image to obtain an effective face coding region; S300, decoding the effective face coding area to obtain depth information corresponding to the effective face coding area; And S400, converting the depth information of the effective face coding area into 3D face point cloud through coordinates. The method is favorable for greatly reducing the computing resources of the structured light module in the application processes of 3D selfie, face unlocking and 3D scanning reconstruction of the smart phone, and is favorable for real-time point cloud processing.

Provided is a method for point cloud data processing and crossover object obtaining of a transmission line. A terrain and ground object information database is established according to actual information of terrain and ground objects in three-dimensional point cloud data; original three-dimensional point cloud data is obtained by utilizing three-dimensional laser scanning; and the ground objects are obtained. On the original basis of not changing the original point cloud data, no modeling rendering processing is performed, spatial information at any point is not changed, he point cloud data of the various ground objects is obtained accurately; and according to concrete requirements of design disciplines, a topographic map suitable for three-dimensional design of a high-voltage transmission line is formed, which is achieved internationally for the first time and is a breakthrough in the technical field of track design.

The present invention relates to the technical field of image processing, point cloud processing and sensor data registration, and more particularly, to a registration method of an event triggered camera and a three-dimensional laser radar. 1, designing a calibration object suitable for registration of an event triggering camera and a three-dimensional laser radar; 2, simultaneously starting an event triggering camera and a three-dimensional laser radar to obtain data of the two sensors at the same time; 3, using image processing method such as edge extraction and specific pattern recognitionto locate a point of a calibration object in that image; 4, locating the same point of the calibration object in the point cloud by using the method of point cloud segmentation based on RANSAC; 5, calculating a transformation matrix on six degrees of freedom in the space according to the results obtained in the step 3 and the step 4; Step 6: the registration result is evaluated by the registrationerror and the edge-based value function proposed by the invention.

The invention discloses a complex structure point cloud processing algorithm bases on Poisson reconstruction. Curved surface reconstruction carried out through a point cloud processing technical method based on Poisson reconstruction adopts an indicator function to describe a curved surface; a vector field based on normal vector is constructed through Gaussian filtering; and a Poisson equation is solved through a multi-grid method to obtain a transition portion of the indicator function of directional point cloud to finish the curved surface reconstruction. The curved surface reconstruction technology based on Poisson is mainly formed by five parts: defining an octree, setting function space, estimating the vector field, solving the Poisson equation and extracting contour surface. With the fast development of a 3D laser scanning device, the point cloud obtaining technology also has considerable progress. The point cloud processing algorithm can be widely applied to the fields of reverse engineering, hybrid modeling, visual inspection, medical images and archaeological and cultural relic modeling and the like.

Provided in the invention is a tunnel deformation monitoring and analysis method based on a grid projection point cloud processing technology. For a tunnel surface point cloud model obtained by a three-dimensional laser scanner, a tunnel design surface model is constructed and unit grid processing is carried out; projection of unit grids on a tunnel surface point cloud model is carried out; fitting of point clouds in a projection range is carried out to form a plane quadrangle; a distance between the center of the plane quadrangle and the projection center at the central axis of the tunnel iscalculated; differences of deformation directions of point clouds of corresponding projection areas of the unit grids are determined; and deformation monitoring data screening is carried out and colorprocessing is carried out on the tunnel design surface model based on the deformation amounts of the corresponding positions of the unit grids, thereby realizing overall analysis and visual processing of the tunnel deformation monitoring. According to the invention, the tunnel surface point cloud model obtained by the three-dimensional laser scanner is processed by using the grid projection pointcloud processing technology, so that tunnel surface point cloud deformation caused by influences of tunnel surface peeling, block falling or attachments can be identified and the monitoring and analysis errors can be avoided. On the basis of the unit meshing and deformation amount visual processing of the tunnel design surface model, the overall analysis of the deformation of the tunnel structureis realized conveniently and thus the visual effect is enhanced.

The invention relates to a water surface multi-target rapid identification tracking method based on a laser radar, and belongs to the technical field of environmental perception. According to the method, the laser radar point cloud is corrected to a real world coordinate system through information such as the position, the speed and the direction of the GPS; downsampling is carried out on the laser radar point cloud at the specified interval; the point cloud is divided into mutually independent tracking targets by using an Euclidean distance clustering method; a k-d tree point cloud search model is constructed for subsequent laser radar point cloud data; the point cloud data in a neighborhood with a specified radius is intercepted by taking the target as a center; and the point cloud is divided into one or more to-be-selected updating targets through an Euclidean distance clustering method, the most similar updating target is selected to replace the original tracking target through aniterative nearest point registration algorithm, and the moving direction and distance of the tracking target are obtained through a registration matrix. According to the method, a plurality of water surface targets can be rapidly detected and tracked so that the operation time of point cloud processing is shortened, and the identification efficiency and tracking precision of the target are greatlyimproved.

The invention relates to the technical field of intelligent warehousing. The invention discloses a tray pose positioning method, device and equipment and a storage medium. The method comprises the following steps of: acquiring an RGBD image and a point cloud image of a preset tray area, performing image recognition processing on the RGBD image; obtaining tray identification information, establishing a corresponding coordinate system based on the tray identification information, wherein the processor is used for receiving the tray identification information, determining a tray center point coordinate according to the tray identification information, extracting tray point cloud data from the point cloud image according to the tray identification information, determining a tray course angle according to the tray point cloud data, and obtaining tray pose data according to the tray center point coordinate and the tray course angle. According to the method, deep learning and point cloud processing are combined, and the tray pose is positioned according to the obtained tray center point coordinate and the tray heading angle, so that high-precision and high-robustness positioning of various types of tray poses is realized, and the universality and robustness of tray pose detection are improved.

The invention discloses a three-dimensional-point-cloud-data-based prevention method of a coal mine roof disaster. System equipment comprises a ground analysis device arranged on the ground, an underground center station, a plurality of three-dimensional laser scanners having same parameters, and a three-dimensional point cloud data processing chip, wherein the center station, the plurality of three-dimensional laser scanners, and the three-dimensional point cloud data processing chip are arranged underground. All underground devices are intrinsically safe ones. According to the method, the possibility of roof disaster occurrence is analyzed based on the point cloud data difference. The coal mine roof disaster risk is predicted by using a method of analyzing three-dimensional point cloud data. The structure is simple; the devices are easy to arranged; the adaptability and recognition rate are high; calculation is simple; the operation efficiency is high; manual intervention is reduced; the low cost is low and the adaptability of the system is high; and hardware and software can be upgraded or repaired conveniently. A three-dimensional voxel filtering algorithm is employ; because of high stability and reliability of the algorithm, data compression, being filtering, can be carried out on the collected three-dimensional point cloud data of the roof rapidly and effectively and the searching can be accelerated greatly. On the basis of a three-dimensional normal distribution transform algorithm, rapid convergence can be realized and thus problems of poor registration convergence and frequent occurrence of local optimum of the common algorithm of the three-dimensional point cloud data can be solved; and the time is saved and complexity of the common algorithm can be reduced. The method can be applied to monitoring of the roof falling disaster in a complex environment; the monitoring precision and efficiency can be improved effectively; and great convenience is provided for safety production.

The invention relates to a digital pre-assembling method of a building steel structure based on three-dimensional scanning and reverse modeling. The method comprises the following steps: step one, carrying out component manufacturing; step two, carrying out component scanning, step three, carrying out point cloud processing; step four, carrying out reverse modeling of a component; step five, carrying out component error checking; step six, carrying out structural assembling; step seven, carrying out overall assembling error checking; and step eight, finishing pre-assembling. According to the invention, all information of components including a bolt hole and small hidden parts can be reflected completely; with reverse modeling, massive stored point cloud data are processed by fitting to obtain an entity model; the data volume is reduced for follow-up pre assembling and entity model displaying; the computer burden is reduced; and the pre assembling efficiency of the computer is improved substantially. On the basis of the optimization-algorithm-based digital pre assembling technique, optimized pre assembling is carried out on the fitted entity component model.

The invention discloses an autonomous cruising system and an autonomous cruising method of a water surface robot. A user can control the water surface robot to establish a water area map through a monitoring station through wireless network communication, learn the water surface environment, and prepare for autonomous navigation of a subsequent water surface robot; an integrated navigation technology is applied to autonomous navigation of the water surface robot to realize autonomous planning of a driving path; the combination of laser scanning point cloud processing and computer vision calculation is used for autonomous searching of a floating target on the water surface, so that autonomous planning of the driving path in the operation process is realized. The disclosed autonomous cruising method for the floating target on the water surface, which utilizes the synergistic effect of a water surface target detection laser radar and a water surface target detection camera, is particularly suitable for the situation that the number and the positions of the target points are uncertain, and the inverted reflection in water and water wave interference caused by illumination can be effectively eliminated. According to the method, the autonomous operation capability of the water surface robot is enhanced, the workload of manned operation is effectively reduced, and the water surface operation efficiency is improved.

The invention discloses a three-dimensional point cloud processing method and device based on geometric unwrapping, and equipment. The method comprises the steps of obtaining and processing point cloud data of an object; carrying out geometric unwrapping on the point cloud data so as to divide the point cloud data into points with large geometric changes and points with small geometric changes according to the geometric change degree; and inputting the point cloud data into a trained convolutional neural network for feature extraction and classification segmentation, the neural network model being used for extracting local features and global features of the point cloud data; and learning mutual supplementary information between the points with large geometric changes and the points with small geometric changes by using a geometric attention module. According to the invention, the accuracy of point cloud data classification tasks and the cross combination rate of segmentation tasks canbe improved.