CN104809759A - Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter - Google Patents

Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter Download PDF

Info

Publication number
CN104809759A
CN104809759A CN201510158959.2A CN201510158959A CN104809759A CN 104809759 A CN104809759 A CN 104809759A CN 201510158959 A CN201510158959 A CN 201510158959A CN 104809759 A CN104809759 A CN 104809759A
Authority
CN
China
Prior art keywords
dimensional
large
texture
destructuring
scene
Prior art date
Application number
CN201510158959.2A
Other languages
Chinese (zh)
Inventor
邱纯鑫
朱晓蕊
李小春
王斌
王秋云
Original Assignee
哈尔滨工业大学深圳研究生院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 哈尔滨工业大学深圳研究生院 filed Critical 哈尔滨工业大学深圳研究生院
Priority to CN201510158959.2A priority Critical patent/CN104809759A/en
Publication of CN104809759A publication Critical patent/CN104809759A/en

Links

Abstract

The invention discloses a large-area unstructured three-dimensional scene modeling method based on a small unmanned helicopter. The large-area unstructured three-dimensional scene modeling method based on the small unmanned helicopter targets at three-dimensional reconstruction of unstructured objects, for example, the Great Wall of the Ming Dynasty in Shanxi Province, and takes the terrain complexity of Great Wall ruins. The method comprises, during a original data collecting stage, utilizing the unmanned helicopter carrying an LiDAR (light detection and ranging) system and a high-quality digital camera to perform data collection on Kuan'gong city walls solve the problem of point cloud collection and original photo collection of large-area objects; during a data analyzing and processing state, performing point cloud matching and overall optimization to effectively overcome the difficulty in point cloud registration. Experiments show that an unmanned helicopter dynamically monitoring system is feasible and can effectively solves three-dimensional reconstruction of large-area unstructured large objects.

Description

Based on the large area destructuring scene three-dimensional modeling method of small-sized depopulated helicopter

Technical field

The present invention relates to geographical mapping technical field, particularly relate to a kind of large area destructuring scene three-dimensional modeling method.

Background technology

Three-dimensional reconstruction refers to sets up to three-dimensional body the mathematical model being applicable to computer representation and process, being the basis processing, operate and analyze its character under computer environment to it, is also the gordian technique setting up the virtual reality expressing objective world in a computer.

The general three-dimensional reconstruction step based on camera is generally: 1. Image Acquisition, 2. camera calibration, 3. feature point extraction, 4. Stereo matching, 5. three-dimensional reconstruction.Above-mentioned technology is view-based access control model method, requires harsh, cannot carry out in an outdoor environment the geometrical property of illumination and camera; Further, due to the limited line-of-sight range of camera, data acquisition cannot be carried out to large area object.

Usually, the technical step based on the three-dimensional reconstruction of laser scanner is: 1. obtain original scanning cloud data, positional information, attitude information and digital picture; 2. pair cloud data carries out pre-service, comprises removal noise, removes erroneous point, removably millet cake; 3. pair some mysorethorn executes registration, obtains preliminary point cloud model; 4. constructive geometry model; 5. texture mapping.The bottleneck of this technology is: the acquisition operational difficulties of one, original point cloud and data image, particularly feels simply helpless especially for large area destructuring object; Two, under the condition of big data quantity, continuous ICP coupling can cause cumulative errors, and this can make follow-up work not continue.The general building of general house and city is all structurized, for structuring scene, collecting part data, such as roof data, just can construct whole three-dimensional model, but destructuring scene (such as bright Great Wall) is needed to gather all data.

Expert both domestic and external and scholar do a lot in airborne three-dimensional reconstruction: the people such as Vivek Verma, Lu Wang and Liang Cheng, by LiDAR data, realize the three-dimensional modeling on roof; The people such as Shugen Wang achieve the three-dimensionalreconstruction work to simple building by LiDAR and video camera; The people such as Charalambos Poullis and Jinhui Hu by LiDAR system, realize carrying out modeling to a large urban area; The people such as Cornelius Wefelscheid have carried out three-dimensional reconstruction with UAVs to a house, and the method for their reconstruct, based on image, not based on laser scanner, and all carries out three-dimensional reconstruction to structurized object; People's laser scanners such as Zhichao Zhang and video camera carry out three-dimensional reconstruction to the sculpture in city, in processing procedure, they merge the overlapping region in point cloud model, make point cloud model more accurate, in the texture stage, they adopt SURF to realize picture match, and finally they have also been obtained good effect; The people such as Chia-Yen Chen use LiDAR system to carry out large-scale laser scanning, author rebuilds proposition two kinds of methods for 3D on a large scale and improves ICP matching precision: one, reject the point that those difference are very large, set a differential threshold, the point within this threshold value should not participate in ICP matching process; Its two, selected characteristic point carries out ICP coupling, and their matching result is also gratifying; After ICP coupling, the people such as cumulative bad error can be larger, Rainer Kummerle propose a kind of graphics-optimized algorithm: G2O; The people such as Sabry F.El-Hakim have also been made the three-dimensional reconstruction of the site of ancient culture remains, they reconstruct sage my audience hall periphery how bright, they are by united for multiple three-dimensional rebuilding method use, such as: for large building, unworkable with the modeling technique of image, the modeling technique based on distance should be adopted namely to adopt laser scanner.But, for the thing that some volumes are little, if go to be reconstructed with laser scanner, the thing of a lot of detail can be lost, now, only get well again with the modeling technique based on image.Based on above-mentioned thought, the three-dimensional model that different technologies reconstructs out by the team of Sabry F.El-Hakim is put into together, and effect is pretty good, but they do not do global optimization to a cloud; Zhang Tao studies cutting out based on the space triangular of complexity point cloud and three-dimensional, the Triangulation Algorithm of the improvement of a kind of complexity point cloud is proposed, the space lattice after processing is made to reflect topological connection relation between body surface discrete point, simultaneously also illumination and material process and texture are increased to D Triangulation, substantially increase rendering speed, improve rendering effect.Zheng Keqiang mainly completes the reconstruct of Three-dimensional unstructured scene, effectively establishes model of geometrical features and the 3 d grid model of scene.Above-mentioned event is not all rebuild non-structured object on a large scale.Wang Changhan etc. do a lot in historical relic three-dimensional reconstruction, and they have carried out three-dimensional reconstruction work with laser to Thousands Hands Guanyin, but they do not reconstruct a complete three-dimensional stereo model, have just reconstructed a Thousands Hands Guanyin's part.Yu Mingxu etc. utilize three-dimensional laser scanner and some system softwares to carry out three-dimensional reconstruction to two tree fossils, and they not carry out global optimization to three-dimensional point cloud, and scanned object volume is little, and analyzing spot cloud number of times is few.

Summary of the invention

In order to solve the problems of the prior art, the invention provides a kind of large area destructuring scene three-dimensional modeling method based on depopulated helicopter, the point cloud collection and the original photo that solve large area object gather a difficult problem, effectively overcome the difficulty of a cloud, solve a three-dimensional reconstruction work difficult problem for the large object of destructuring on a large scale.

The present invention is achieved through the following technical solutions:

Based on a large area destructuring scene three-dimensional modeling method for depopulated helicopter, described method adopts segment reconstruction, finally splices and carry out three-dimensional modeling to large area destructuring scene, said method comprising the steps of:

A. depopulated helicopter lift-launch LiDAR system and high-quality digital camera gather laser point cloud data and the image data of destination object;

B. described cloud data is processed: first carry out ICP and mate single step rectification, then G2O figure global optimization, then carry out triangle gridding process, obtain surface mesh normal vector;

C. described image data is processed: first carry out pre-service, then calculate image mapping relations, obtain texture sequence;

D. utilize the texture of redundancy to carry out multitexture mapping, realize the blurred transition between texture: the texture sequence that step C obtains is attached on the 3-D geometric model that step B obtains according to texture coordinate one by one, forms three-dimensional digital model.

E. the three-dimensional digital model of destination object is exported.

As a further improvement on the present invention, in described step B, described cloud data is processed and also comprise a cloud denoising: based on determining radius encircle sphere denoise algorithm, take impact point as the center of circle, if counted in radii fixus region be less than quantity threshold values, be then considered to noise and removed.

As a further improvement on the present invention, in described step B before carrying out triangle gridding process, need use based on the method for cluster to point cloud compressing process.

As a further improvement on the present invention, described triangle gridding process based on the PowerCrust algorithm of delaunay criterion as the method for trigonometric ratio.

As a further improvement on the present invention, in order to obtain more level and smooth accurate model, need to be encrypted grid, cryptographic algorithm is based on three rank B-spine interpolation.

As a further improvement on the present invention, described method also comprises step F: spliced by the three-dimensional digital model of multistage destination object, forms the three-dimensional digital model of complete destination object

The invention has the beneficial effects as follows: the large area destructuring scene three-dimensional modeling method based on small-sized depopulated helicopter that the present invention proposes, in the raw data acquisition stage, LiDAR system and high-quality digital camera is carried with depopulated helicopter, reconcile amounts tribute city city wall carries out data collection task, and the some cloud collection and the original photo that solve large area object gather a difficult problem.Data analysis and processing stage, have employed point cloud matching and global optimization, effectively overcome the difficulty of a cloud.Experiment proves, this cover unmanned plane dynamic monitoring system is feasible, efficiently solves the three-dimensional reconstruction work of the large object of destructuring on a large scale.

Accompanying drawing explanation

Fig. 1 is the flow chart of data processing figure of method of the present invention;

Fig. 2 is the principle schematic of G2O algorithm;

Fig. 3 is texture schematic diagram;

Fig. 4 is Great Wall data acquisition schematic diagram;

Fig. 5 is depopulated helicopter laser scanning schematic diagram;

Fig. 6 is gap section city wall point cloud chart;

Fig. 7 is gap section city wall 3-D geometric model vertical view;

Fig. 8 is gap section city wall 3-D geometric model side view;

Fig. 9 is the side view that gap section city wall sticks texture.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

The attached flow chart of data processing figure that Figure 1 shows that large area destructuring scene three-dimensional modeling method of the present invention, input is laser point cloud data, camera picture; During process, have employed the methods such as point cloud matching, G2O graphics-optimized, the denoising of some cloud, point cloud compressing, some cloud trigonometric ratio, texture mapping; Output is three-dimensional model.

Point cloud matching is the basis of three-dimensional reconstruction, which determines the precision of model.Point cloud matching field uses maximum methods to be ICP (Iterative Closest Point: iterative closest point method), it is the thought of Corpus--based Method, calculate the Euclidean distance between all matching double points, the distance r.m.s. of all-pair is as evaluation function, when this function convergence is to minimal value, namely think coupling.The shortcoming of ICP coupling is that it can only find local optimum, for a model of cloud integrated distribution, the point cloud of such as globoid or right cylinder distribution, this algorithm effect is best, and the some cloud on Great Wall for the distribution of linearly wide area, this algorithm can not the guarantee point cloud overall situation move towards correctness, reuse ICP and can cause cumulative bad error.Therefore present invention uses G2O (General Graph Optimization: a kind of optimized algorithm for overall nonlinearity erron based on figure), it can on the basis of ICP coupling, the overall situation distribution of a cloud is finely tuned, makes point cloud matching reach global optimum.Accompanying drawing 2 describes the principle of G2O.

There is noise in scanning process, itself there is edge effect in scanner, in addition Ruins of Great Wall landforms are changeful, this brings noise to scan-data, makes scan-data flatness very poor, and the common filtering algorithm based on smoothed data well can not adapt to the uneven some cloud of this roughness, for this reason, the present invention adopts based on determining radius encircle sphere denoise algorithm: take impact point as the center of circle, if counted in radii fixus region be less than quantity threshold values, be then considered to noise and removed.

A large amount of splicing regions is there is between some cloud scanning plane, cause a cloud redundancy, in addition, because Ruins of Great Wall surfaceness is uneven, and there is a lot of sharp point, and the difference at scanning visual angle, cause cloud data Density inhomogeneity, be unfavorable for generating uniform grid model, and grid model may be caused to there is the hole that should not exist, before gridding, need to simplify process.General simplifies algorithm based on point cloud local curvature, less point can be retained like this in level and smooth place, more point is retained in the place that curvature is larger, to retain the more feature of the original, but this algorithm is based on the more level and smooth surface of entirety, and the cloud data on Great Wall is unsmooth and uneven, so the method that present invention uses based on cluster is simplified, this algorithm can cause less details disappearance, but meets the demand of this project.

Be gridding method by the point maximum method made in face, and in gridding method be triangle gridding the most widely, it possesses the characteristic of self-adaptation multiresolution, and wide adaptability, so this task choosing Triangulation Algorithm is as the method extracting outside surface.Triangulation Algorithm can be divided into two classes: based on the approach method of implicit function and the method for geometry based on delaunay criterion.First kind method carrys out structured surface by implicit function matching, be applicable to the more smooth and model of rule, be not suitable for not only rough but also irregular Great Wall point cloud, so, the present invention have selected PowerCrust algorithm based on delaunay criterion as the method for trigonometric ratio, it is widely used, and robustness is comparatively strong, and can ensure the correctness of global Topological Structure.In order to obtain more level and smooth accurate model, the present invention also encrypts grid, and cryptographic algorithm is based on three rank B-spine interpolation.

Surface mesh normal vector is requisite important parameter in 3D modeling, is the foundation of shadowing in texture acquirement, is also the reference of many grain table weight and the reference of illumination calculation.The present invention utilizes the cross product on any both sides of triangle to determine initial normal vector, utilize Multi-extended adjusting method vector in PowerCrust algorithm towards, make it consistent outwardly.

Being different from indoor can manual control environment light field, outdoor environment is complicated and changeable, such as cloudy day, fine day, photo environment brightness is different, causes the inconsistent of texture lightness, tone, and this can bring difficulty to follow-up grain table, so, the initial photo of shooting can not be directly used in extraction texture, needs first to carry out certain pre-service, comprises the global adaptation of brightness, tone.

Calculate the indispensable step before image mapping relations are texture, image mapping relations are the prerequisites obtaining texture coordinate.Image mapping relations are obtained by a series of space coordinate transformation, and these coordinate transforms are based on the inside and outside parameter of camera.So first need to demarcate camera.Camera calibration is a very proven technique at present, present invention uses the camera calibration tool box of Matlab to complete the demarcation of camera.It image mapping principle is combined by two conversion: the conversion that world coordinates is tied to the conversion of camera coordinates system, camera coordinates is tied to image coordinate system.Parameters of these conversion derive from the internal reference (being provided by demarcation) of pose data that GPS and IMU (Inertial Measurement Unit) measure and camera.The degree of accuracy of these parameters directly affects the precision of texture.Accompanying drawing 3 exponent is as mapping relations, and wherein, P and p is world coordinates and the image coordinate of same point, and O is camera coordinates system, and f is focal length, and C ' is image principle point location.

3D grid model vertex data is projected to the texture and texture coordinate thereof that on pretreated image, just can get each tri patch according to image mapping relations.

The object of texture strengthens the authenticity in 3D model-based vision.The work that texture is done be exactly by previous step obtain texture sequence according to texture coordinate one by one " subsides " on the 3D geometric model that step obtains before.

Because the texture image obtained exists lap, so texture sequence exists redundancy, namely a corresponding multiple texture of tri patch possibility, exists one here and selects excellent or fusion problem.In some cases, they take the texture selecting visual angle maximum as texture to be mapped, and then carry out grain table.But cause data waste like this, these data may be used for grain table.In practical operation, matching error is there is between adjacent texture image, so, if maximum visual angle texture is not derive from same piece image, may occur the problem of " misalignment " between adjacent texture triangle, solution of the present invention is exactly utilize the texture of redundancy to carry out multitexture mapping, realizes the blurred transition between texture, data are neither caused to waste, perfect again grain table.

Field experiment devises following data acquisition plan: adopt depopulated helicopter as carrier, machine carries the positioning equipments such as GPS and IMU, is used for measuring the posture information of carrier; The mode of the two dimensional laser scanning instrument of lighter in weight and rotary head collocation is adopted to complete the measurement of Ruins of Great Wall landform distributed in three dimensions data; Adopt high-resolution camera to complete the image data acquiring of Ruins of Great Wall.Then according to the data collected, set up the emulation three-dimensional model on Great Wall, can reflect three-dimensional geometry data and the texture general picture on Great Wall, can provide virtual roaming, the dynamic monitoring for Great Wall provides visual support.

Considering the huge area of short distance flying power and the Shaanxi bright Great Wall town North Platform money Gong Cheng of depopulated helicopter, in order to complete the three-dimensional reconstruction to North Platform money tribute city, town enclosure wall, adopting segment reconstruction, the method for finally splicing.To carry out data collection task to gap section metope, operating diagram as shown in Figure 4.

In figure 4, unmanned plane carries out laser scanning work 13 different positions to Great Wall body of wall, these 13 positions be Position Approximate and position number by actual conditions (body of wall length, the complexity etc. of wall) and determine, as long as its physical location can satisfy condition: after laser completes a scan period, minority ground point can be obtained, most of metope point and wall summit.Unmanned plane carries out data acquisition successively according to a position, 13 in accompanying drawing 4, that is: unmanned plane moves to reach destination locations, and keep hovering, laser image data, moves to the next position then.The namely data acquisition scheme of so-called " flight-stop-scanning ".After unmanned plane completes scanning according to the order of 1-12, get back near position 1, gather the 13rd group of data to complete the data collection task of this section of body of wall.If unmanned plane drift occurs in the process of laser image data, shakes greatly, then need Resurvey data herein.

In order to realize carrying out data acquisition better, need roughly to know the flying height of unmanned plane and the horizontal range of unmanned plane distance body of wall.As shown in Figure 5, wherein, in figure, D is object physical length to depopulated helicopter laser scanning schematic diagram, and h is object true altitude, and d is the horizontal range of helicopter distance object, and H is the vertical range of depopulated helicopter distance object.

tan α = d H + h tan β = d H

According to the The Cloud Terrace rotation angle range [α, β] set in advance, appropriate flying height and flying distance can be calculated.

Before a cloud gathers, also need to do examine on the spot, roughly the length of measurement target object, width and height, and then determine flying height and the flying distance of unmanned plane according to α and β, so just can better collection point cloud data, the quality that raising is tested and efficiency.

In field experiment, reconcile amounts Gong Chengdongqiang has carried out data acquisition near southern wall gap section city wall, accompanying drawing 6 is the point cloud chart of this section of city wall, accompanying drawing 7 is gap section city wall 3-D geometric model vertical views, accompanying drawing 8 is gap section city wall 3-D geometric model side views, and accompanying drawing 9 is side views that gap section city wall sticks texture.Point Cloud Processing result is out the figure as accompanying drawing 7 and accompanying drawing 8, and they are texture not.And accompanying drawing 9 is through the effect be attached to by texture after such model, it reflects the landforms on Great Wall truly.

Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to protection scope of the present invention.

Claims (6)

1., based on a large area destructuring scene three-dimensional modeling method for depopulated helicopter, described method adopts segment reconstruction, finally splices and carry out three-dimensional modeling to large area destructuring scene, it is characterized in that: said method comprising the steps of:
A. depopulated helicopter lift-launch LiDAR system and high-quality digital camera gather laser point cloud data and the image data of destination object;
B. described cloud data is processed: first carry out ICP and mate single step rectification, then G2O figure global optimization, then carry out triangle gridding process, obtain surface mesh normal vector;
C. described image data is processed: first carry out pre-service, then calculate image mapping relations, obtain texture sequence;
D. utilize the texture of redundancy to carry out multitexture mapping, realize the blurred transition between texture, the texture sequence obtained by step C is attached on the 3-D geometric model that step B obtains according to texture coordinate one by one, forms three-dimensional digital model;
E. the three-dimensional digital model of destination object is exported.
2. large area destructuring scene three-dimensional modeling method according to claim 1, it is characterized in that: in described step B, described cloud data is processed and also comprise a cloud denoising: based on determining radius encircle sphere denoise algorithm, take impact point as the center of circle, if counted in radii fixus region be less than quantity threshold values, be then considered to noise and removed.
3. large area destructuring scene three-dimensional modeling method according to claim 1, is characterized in that: in described step B before carrying out triangle gridding process, needs use based on the method for cluster to point cloud compressing process.
4. large area destructuring scene three-dimensional modeling method according to claim 1, is characterized in that: described triangle gridding process based on the PowerCrust algorithm of delaunay criterion as the method for trigonometric ratio.
5. large area destructuring scene three-dimensional modeling method according to claim 1, is characterized in that: in order to obtain more level and smooth accurate model, need to be encrypted grid, cryptographic algorithm is based on three rank B-spine interpolation.
6. large area destructuring scene three-dimensional modeling method according to claim 1, is characterized in that: described method also comprises step F: spliced by the three-dimensional digital model of multistage destination object, forms the three-dimensional digital model of complete destination object.
CN201510158959.2A 2015-04-03 2015-04-03 Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter CN104809759A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510158959.2A CN104809759A (en) 2015-04-03 2015-04-03 Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510158959.2A CN104809759A (en) 2015-04-03 2015-04-03 Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter

Publications (1)

Publication Number Publication Date
CN104809759A true CN104809759A (en) 2015-07-29

Family

ID=53694564

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510158959.2A CN104809759A (en) 2015-04-03 2015-04-03 Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter

Country Status (1)

Country Link
CN (1) CN104809759A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105389849A (en) * 2015-11-23 2016-03-09 公安部交通管理科学研究所 Vehicle collision angle analysis system based on three-dimensional reconstruction technology
CN105678833A (en) * 2016-01-11 2016-06-15 塔里木大学 Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling
CN106023312A (en) * 2016-05-13 2016-10-12 南京大学 Automatic 3D building model reconstruction method based on aviation LiDAR data
CN106097433A (en) * 2016-05-30 2016-11-09 广州汉阈数据处理技术有限公司 Object industry and the stacking method of Image model and system
CN106204443A (en) * 2016-07-01 2016-12-07 成都通甲优博科技有限责任公司 A kind of panorama UAS based on the multiplexing of many mesh
CN106643494A (en) * 2016-12-22 2017-05-10 上海华测导航技术股份有限公司 Mine windrow volume measurement method and system
CN106969721A (en) * 2017-02-20 2017-07-21 深圳大学 A kind of method for three-dimensional measurement and its measurement apparatus

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060061584A1 (en) * 2004-09-20 2006-03-23 Kristiansen Stig R Method, system and device for efficient distribution of real time three dimensional computer modeled image scenes over a network
US20140125654A1 (en) * 2003-02-14 2014-05-08 Everyscape, Inc. Modeling and Editing Image Panoramas
CN104320616A (en) * 2014-10-21 2015-01-28 广东惠利普路桥信息工程有限公司 Video monitoring system based on three-dimensional scene modeling
CN104318622A (en) * 2014-10-25 2015-01-28 大连理工大学 Triangular mesh modeling method of indoor scene inhomogeneous three dimension point cloud data

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140125654A1 (en) * 2003-02-14 2014-05-08 Everyscape, Inc. Modeling and Editing Image Panoramas
US20060061584A1 (en) * 2004-09-20 2006-03-23 Kristiansen Stig R Method, system and device for efficient distribution of real time three dimensional computer modeled image scenes over a network
CN104320616A (en) * 2014-10-21 2015-01-28 广东惠利普路桥信息工程有限公司 Video monitoring system based on three-dimensional scene modeling
CN104318622A (en) * 2014-10-25 2015-01-28 大连理工大学 Triangular mesh modeling method of indoor scene inhomogeneous three dimension point cloud data

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王斌: "基于LiDAR散乱点云的长城遗址三维模型构建", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105389849A (en) * 2015-11-23 2016-03-09 公安部交通管理科学研究所 Vehicle collision angle analysis system based on three-dimensional reconstruction technology
CN105389849B (en) * 2015-11-23 2018-06-05 公安部交通管理科学研究所 Vehicle collision angle resolved systems based on three-dimensional reconstruction
CN105678833A (en) * 2016-01-11 2016-06-15 塔里木大学 Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling
CN106023312A (en) * 2016-05-13 2016-10-12 南京大学 Automatic 3D building model reconstruction method based on aviation LiDAR data
CN106023312B (en) * 2016-05-13 2019-02-22 南京大学 Three-dimensional building object model automatic reconstruction method based on aviation LiDAR data
CN106097433A (en) * 2016-05-30 2016-11-09 广州汉阈数据处理技术有限公司 Object industry and the stacking method of Image model and system
CN106204443A (en) * 2016-07-01 2016-12-07 成都通甲优博科技有限责任公司 A kind of panorama UAS based on the multiplexing of many mesh
CN106643494A (en) * 2016-12-22 2017-05-10 上海华测导航技术股份有限公司 Mine windrow volume measurement method and system
CN106969721A (en) * 2017-02-20 2017-07-21 深圳大学 A kind of method for three-dimensional measurement and its measurement apparatus

Similar Documents

Publication Publication Date Title
Zhang et al. LOAM: Lidar Odometry and Mapping in Real-time.
JP4685313B2 (en) Method for processing passive volumetric image of any aspect
Haala et al. An update on automatic 3D building reconstruction
Wallace et al. Development of a UAV-LiDAR system with application to forest inventory
Harwin et al. Assessing the accuracy of georeferenced point clouds produced via multi-view stereopsis from unmanned aerial vehicle (UAV) imagery
US20040105573A1 (en) Augmented virtual environments
Berger et al. A survey of surface reconstruction from point clouds
Baltsavias et al. High‐quality image matching and automated generation of 3D tree models
Cheng et al. 3D building model reconstruction from multi-view aerial imagery and lidar data
US20080238919A1 (en) System and method for rendering of texel imagery
Poullis et al. Automatic reconstruction of cities from remote sensor data
Meng et al. Ground filtering algorithms for airborne LiDAR data: A review of critical issues
Mancini et al. Using unmanned aerial vehicles (UAV) for high-resolution reconstruction of topography: The structure from motion approach on coastal environments
Sohn et al. Extraction of buildings from high resolution satellite data
KR20100087034A (en) Method and apparatus of taking aerial surveys
Mayer Object extraction in photogrammetric computer vision
CN102332179B (en) Three-dimensional model data simplification and progressive transmission methods and devices
EP2973409A2 (en) Determining object volume from mobile device images
Matei et al. Building segmentation for densely built urban regions using aerial lidar data
Johnson‐Roberson et al. Generation and visualization of large‐scale three‐dimensional reconstructions from underwater robotic surveys
AU2011312140B2 (en) Rapid 3D modeling
Gross et al. Extraction of lines from laser point clouds
CN102317973A (en) Fusion of a 2d electro-optical image and 3d point cloud data for scene interpretation and registration performance assessment
CA2678491A1 (en) Exemplar/pde-based technique to fill null regions and corresponding accuracy assessment
CN103021017B (en) Three-dimensional scene rebuilding method based on GPU acceleration

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20150729

RJ01 Rejection of invention patent application after publication