CN113012206B - Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics - Google Patents

Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics Download PDF

Info

Publication number
CN113012206B
CN113012206B CN202110179598.5A CN202110179598A CN113012206B CN 113012206 B CN113012206 B CN 113012206B CN 202110179598 A CN202110179598 A CN 202110179598A CN 113012206 B CN113012206 B CN 113012206B
Authority
CN
China
Prior art keywords
point
points
eave
vehicle
airborne
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN202110179598.5A
Other languages
Chinese (zh)
Other versions
CN113012206A (en
Inventor
阳凡林
郭亚栋
王贤昆
宿殿鹏
马跃
亓超
李邵禹
刘骄阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
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 Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN202110179598.5A priority Critical patent/CN113012206B/en
Publication of CN113012206A publication Critical patent/CN113012206A/en
Application granted granted Critical
Publication of CN113012206B publication Critical patent/CN113012206B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • G06T7/344—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4023—Scaling of whole images or parts thereof, e.g. expanding or contracting based on decimating pixels or lines of pixels; based on inserting pixels or lines of pixels
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/757—Matching configurations of points or features

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Mathematical Analysis (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computing Systems (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Algebra (AREA)
  • Medical Informatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses an airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics, which belongs to the technical field of mobile measurement, and comprises the steps of denoising and filtering airborne and vehicle-mounted LiDAR point cloud data, and extracting building point clouds and characteristic angular points from ground object points; establishing a local neighborhood similarity measure model, realizing quick automatic matching of airborne and vehicle-mounted LiDAR pseudo-homonymy points, and realizing coarse registration by using pseudo-homonymy points; then, expanding the contour line of the vehicle-mounted LiDAR building based on a direction prediction algorithm to construct a potential eave feature point set; and finally, respectively realizing airborne and vehicle-mounted LiDAR point cloud true homonymy point pair matching and iterative fine registration by utilizing an ICP algorithm twice. The invention realizes the high-precision registration of the airborne LiDAR and the vehicle-mounted LiDAR, and effectively solves the problems of few homonymous features, low registration precision and the like of the airborne LiDAR and the vehicle-mounted LiDAR data under the influence of eaves.

Description

Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics
Technical Field
The invention belongs to the technical field of mobile measurement, and particularly relates to an airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics.
Background
A laser radar (Light Detection and Ranging) system integrates a laser scanner, a global satellite navigation system, an inertial navigation system and other sensors, and can quickly obtain three-dimensional high-precision laser point cloud. The LiDAR system can acquire terrain three-dimensional data information in a high-speed laser scanning measurement mode with high efficiency, high precision, automation and directness. Compared with the traditional measuring method, the LiDAR measurement is basically not limited by illumination and weather conditions, has the advantages of short period, high speed, low cost, high efficiency and the like, fully meets the timeliness requirements of terrain three-dimensional modeling and updating, and has higher research value and important practical significance.
Airborne LiDAR (ALS) and Vehicle-mounted LiDAR (VLS) technologies are widely used in road extraction, cultural heritage protection, building three-dimensional modeling, smart city construction, environmental monitoring and other works. However, due to the scanning distance and the scanning angle, it is difficult for a single measurement system to acquire the complete information of the ground object: ALS generally can only obtain top information of ground features, and side elevation information is rare or even missing; due to the scanning view angle problem, the VLS often can only acquire the side elevation information of the ground features such as buildings, and the top information of the ground features is difficult to acquire. How to obtain omnibearing stereo point cloud data and realize the registration of ALS and VLS visual angle data is a hot point problem of LiDAR data processing application.
The prior art has respective advantages in performance, but is mostly suitable for registration of a building without an eave area. The difference of the shape and the size of the building eave and the vertical face causes the spatial position deviation of a top surface feature (ALS) and a side surface feature (VLS), and the multi-view point cloud registration based on the building features needs to solve the eave influence.
Disclosure of Invention
Aiming at the technical problems in the prior art, the invention provides an airborne and vehicle-mounted LiDAR point cloud registration method considering eave features, the design is reasonable, and the problems that the same-name features of the airborne and vehicle-mounted LiDAR are few, the prior art is not suitable for eave areas and the like are solved. The registration method can effectively realize multi-source point cloud registration on the premise of fully considering the eave characteristics, and has good applicability to airborne and vehicular LiDAR point cloud registration under the influence of the eave.
In order to achieve the purpose, the invention adopts the following technical scheme:
an airborne and vehicle-mounted LiDAR point cloud registration method considering eave features comprises the following steps:
step 1: denoising and filtering point cloud data of Airborne LiDAR (ALS) and Vehicle-mounted LiDAR (VLS), and extracting building point cloud and characteristic corner points from ground object points;
step 2: establishing a local neighborhood similarity measure model, realizing quick automatic matching of airborne and vehicle-mounted LiDAR pseudo-homonymy points, and realizing coarse registration by using pseudo-homonymy points;
and step 3: expanding the outline of the vehicle-mounted LiDAR building based on a direction prediction algorithm to construct a potential eave feature point set;
and 4, step 4: and respectively realizing the matching and Iterative fine registration of true homonymous point pairs of airborne and vehicle-mounted LiDAR point clouds by utilizing an Iterative Closest Point (ICP) algorithm twice.
Preferably, in step 2, the local neighborhood similarity measure model is established as follows:
step 2.1: calculating the local gravity center of the feature points;
sequentially selecting one point ap from airborne LiDAR characteristic points, and searching for the same point ap with the selected point apK nearest feature points ap, apn of the building1，…，apnk-1Calculating the center of gravity O of the k pointsap；
One point vp is sequentially selected from vehicle-mounted LiDAR characteristic points, and k nearest characteristic points vp and vpn belonging to the same building are searched1，…，vpnk-1Calculating the gravity center O of the k pointsvp；
Step 2.2: evaluating local neighborhood similarity;
calculating a correlation coefficient of the ALS characteristic point coordinate difference sequence and the VLS characteristic point coordinate difference sequence by using formulas (1) and (2);
Figure GDA0003504828540000021
Figure GDA0003504828540000022
in the formula (I), the compound is shown in the specification,
Figure GDA0003504828540000023
gix、giyrespectively being the abscissa and ordinate g 'of the ith neighborhood point of the ALS point cloud current point'ix、g'iyRespectively the horizontal and vertical coordinates of the ith neighborhood point of the current point in the VLS point cloud;
calculating ap to OapAzimuthal angle of and vp to OvpAnd calculating ap to OapEuclidean distance of and vp to OvpThe difference in Euclidean distance of (c);
step 2.3: constructing a similarity measure model;
each feature point in the VLS and the current ALS feature point are matched and calculated, and when the absolute value of the horizontal and vertical correlation coefficient is larger than a threshold value TρAp to OapAzimuthal angle of and vp to OvpIs less than a threshold value TaWhen and ap to OapEuclidean distance of and vp to OvpHas an absolute difference value of Euclidean distance smaller than a threshold value Td1When the formula (3) is satisfied, the data is stored at the beginningStarting matching set SsimPerforming the following steps;
Figure GDA0003504828540000031
in the formula, angle apOapIs ap to OapAngle of < vpOvpIs vp to OvpThe azimuth of (d); | ap-OapL is ap to OapThe Euclidean distance of; i vp-OvpL is vp to OvpThe Euclidean distance of; | | is an absolute value;
repeating the step 2.1 to the step 2.3 until all ALS characteristic points are processed;
step 2.4: random sampling consistency matching is carried out to determine an optimal matching pair;
from the set SsimRandomly selecting a pair of matching points and calculating the horizontal and vertical coordinate difference, and counting the horizontal and vertical coordinate differences of other points and the absolute difference value thereof at a threshold value Td2Number within range, establishing pseudo-homonymous point pair set S according to matching point pair with maximum numberpcpOne-to-many or many-to-one matching phenomenon can occur in the set, and the point pairs of max (rho (x) + rho (y)) are selected to ensure one-to-one matching, so that pseudo-homonymous point pairs are obtained;
step 2.5: coarse registration;
interpolating elevation of pseudo-homonymous point pairs by using filtered ground point clouds to obtain [ X ]als Yals Zals]TAnd [ X ]vls YvlsZvls]TRespectively representing ALS and VLS characteristic point coordinates of the pseudo homonymy point pairs, wherein coordinate conversion relations are represented by formulas (4), (5) and (6), and coordinate conversion parameters are solved by using least squares;
Figure GDA0003504828540000032
Figure GDA0003504828540000033
Figure GDA0003504828540000034
wherein R (α, β, γ) represents a rotation matrix, α, β, γ being Euler rotation angles around the Z, Y and X axes, respectively; [ T ]xTy Tz]TRepresenting the amount of translation in the direction X, Y, Z; λ represents the scale parameter, and since ALS and VLS are registered as rigid transformations, let λ be 1.
Preferably, in step 3, the vehicle-mounted LiDAR building corner points are extended based on a direction prediction algorithm, and the step of constructing the potential eaves feature point set is as follows:
step 3.1: determining the position of a potential feature point set of the eave;
the eave condition of a certain corner of the facade of the building comprises three conditions of no eave, existence on one side and existence on two sides, so that eave characteristic points of the building have 4 possibilities, the eave characteristic points are called as potential eave characteristic points, the eave potential characteristic points at a certain position comprise intersection points of contour lines and adjacent contour lines, intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the contour lines and the adjacent contour lines, intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the contour lines and intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the adjacent contour lines, and the eave width outwards along the vertical line direction;
step 3.2: judging the outward expansion direction;
firstly, sequencing ALS data by taking a feature point with the minimum horizontal coordinate of a current building as a starting point and clockwise, and determining a starting point and an end point of each section of contour line; selecting pseudo homonymous point pairs one by one, respectively searching characteristic points adjacent to the pseudo homonymous point pairs, reordering VLS characteristic points according to a slope similarity principle, and determining a starting point and an end point of a contour line;
making predictions from the geometric relationships: when the abscissa of the starting point is smaller than the abscissa of the end point, the upper side of the straight line where the contour line is located is the outer side, and the contour line is corrected along the Y-axis direction; when the abscissa of the starting point is larger than the abscissa of the end point, the lower side of the straight line where the contour line is located is the outer side, and the contour line is corrected downwards along the Y axis;
step 3.3: expanding a vehicle-mounted LiDAR contour line;
setting a linear equation Ax + By + C of the contour line as 0, and determining the relation between the width d of the eave and the translation distance of the contour line along the Y axis according to geometry; in order to simplify the calculation formula, the linear equation coefficient A is changed into a non-negative form, and the VLS contour line is expanded by using a formula (7); obtaining 4 plane positions of potential eave feature points of each corner through the intersection points of adjacent contour lines, and determining the elevation of the potential eave feature points by adopting an elevation interpolation method;
Figure GDA0003504828540000041
in the formula, A, B, C represents the coefficient of the general formula of the linear equation, Xstart、XendThe horizontal coordinates of the starting point and the end point of the contour line and the width of the building eave are shown as d.
Preferably, in step 4, the method respectively realizes the matching and iterative fine registration of the airborne and vehicle-mounted LiDAR point cloud true homonymy point pairs by using the two-time iterative closest point algorithm, and comprises the following steps:
step 4.1: taking a coarse registration result as an initial value of ICP, taking a potential eave feature point set as a search point set and constructing a KD tree to improve search efficiency, and taking minimization of Euclidean distance E (R, T) between homonymous points as an optimal convergence condition of the ICP algorithm, so that a true homonymous point matching problem is expressed as a point pair corresponding to the minimum value of a solving function E (R, T), wherein E (R, T) is as formula (8);
Figure GDA0003504828540000042
step 4.2: obtaining minimum E (R, T) after iterative optimization, wherein the point pair for resolving the registration parameters is a true homonymy point pair of ALS and VLS building characteristic points, namely ALShAnd VLSh；
Step 4.3: the method comprises the steps that vehicle-mounted LiDAR point cloud data is used as a search point set, a true homonymy point matching result based on an ICP method is used as an initial value of the ICP, minimization of Euclidean distance between homonymy points is used as an optimal convergence condition of the ICP algorithm, and a point cloud fine registration problem can be expressed as a registration parameter corresponding to the minimum value of a solving function E (R, T);
step 4.4: and obtaining the minimum E (R, T) after iterative optimization, and then the ALS and VLS data registration parameters are shown as the formula (9):
Figure GDA0003504828540000051
where R, T is the ALS and VLS final registration parameter, RRi、TTiA rotation and translation matrix for each calculation of the iterative process, j being the final iteration number, Ti＝RiTi-1+TTi。
The invention has the following beneficial technical effects:
compared with the prior art, the method has the advantages that a local neighborhood similarity measurement model is constructed, and quick matching of pseudo-homonymous point pairs of the airborne and vehicle-mounted LiDAR point clouds is realized; constructing a potential eave feature point set by externally expanding a vehicle-mounted LiDAR contour line based on a direction prediction algorithm, and realizing high-precision registration of airborne and vehicle-mounted LiDAR point clouds by utilizing an ICP algorithm twice; the problems of few homonymous features of airborne and vehicle-mounted LiDAR, low registration accuracy and the like under the influence of effective eaves.
Drawings
FIG. 1 is a flow chart of the present invention that accounts for the registration of on-board and on-board LiDAR for eaves features.
Fig. 2 is a schematic diagram of potential eave feature point positions in the invention.
FIG. 3 is a schematic diagram of the direction prediction based on-board LiDAR building contour line dilation in the present invention.
FIG. 4 is a detailed flow chart of the airborne and vehicle-mounted LiDAR point cloud registration method of the present invention that accounts for eaves characteristics.
Detailed Description
The invention is described in further detail below with reference to the following figures and detailed description:
in order to achieve the purpose, the invention adopts the following technical scheme:
a flow of an airborne and vehicle-mounted LiDAR point cloud registration method considering eave features is shown in FIG. 1, and comprises the following steps:
step 1: denoising and filtering point cloud data of Airborne LiDAR (ALS) and Vehicle-mounted LiDAR (VLS), and extracting building point cloud and characteristic corner points from ground object points;
denoising and filtering the airborne and vehicle-mounted LiDAR point cloud data respectively, and extracting building point cloud and characteristic angular points by using the prior art;
step 2: establishing a local neighborhood similarity measure model, realizing quick automatic matching of airborne and vehicle-mounted LiDAR pseudo-homonymy points, and realizing coarse registration by using pseudo-homonymy points;
different local neighborhood feature descriptors reflect different local neighborhood similarity degrees, and the local neighborhood feature descriptors provide powerful data support for matching pseudo-homonymous point pairs. In order to comprehensively and comprehensively realize automatic matching of the onboard and vehicle-mounted LiDAR pseudo-homonymous point pairs, firstly, calculating a correlation coefficient of KNN characteristic points, and setting an azimuth angle difference threshold value and an Euclidean distance difference threshold value from the characteristic points to the gravity center to reduce mismatching; considering that buildings may have local similarity, the RANSAC algorithm is used to ensure that pseudo-homonymous points are optimally matched.
In a further embodiment, step 2 specifically includes the following steps:
step 2.1: calculating the local gravity center of the feature points;
one point ap is selected from the airborne LiDAR feature points in sequence, and k nearest feature points ap belonging to the same building are searched for, apn1，…，apnk-1Calculating the center of gravity O of the k pointsap；
One point vp is sequentially selected from vehicle-mounted LiDAR characteristic points, and k nearest characteristic points vp and vpn belonging to the same building are searched1，…，vpnk-1Calculating the gravity center O of the k pointsvp；
Step 2.2: evaluating local neighborhood similarity;
calculating a correlation coefficient of the ALS characteristic point coordinate difference sequence and the VLS characteristic point coordinate difference sequence by using formulas (1) and (2);
Figure GDA0003504828540000061
Figure GDA0003504828540000062
in the formula (I), the compound is shown in the specification,
Figure GDA0003504828540000063
gix、giyrespectively being the abscissa and ordinate g 'of the ith neighborhood point of the ALS point cloud current point'ix、g'iyRespectively the horizontal and vertical coordinates of the ith neighborhood point of the current point in the VLS point cloud;
calculating ap to OapAzimuthal angle of and vp to OvpAnd calculating ap to OapEuclidean distance of and vp to OvpThe difference in Euclidean distance of (c);
step 2.3: constructing a similarity measure model;
each feature point in the VLS and the current ALS feature point are matched and calculated, and when the absolute value of the horizontal and vertical correlation coefficient is larger than a threshold value TρAp to OapAzimuthal angle of and vp to OvpIs less than a threshold value TaWhen and ap to OapEuclidean distance of and vp to OvpHas an absolute difference value of Euclidean distance smaller than a threshold value Td1When the initial matching set S is stored, namely the formula (3) is satisfiedsimPerforming the following steps;
Figure GDA0003504828540000071
in the formula, angle apOapIs ap to OapAngle of < vpOvpIs vp to OvpThe azimuth of (d); | ap-OapL is ap to OapThe Euclidean distance of; i vp-OvpL is vp to OvpThe Euclidean distance of; | | is an absolute value;
repeating the step 2.1 to the step 2.3 until all ALS characteristic points are processed;
step 2.4: random sampling consistency matching is carried out to determine an optimal matching pair;
from the set SsimRandomly selecting a pair of matching points and calculating the horizontal and vertical coordinate difference, and counting the horizontal and vertical coordinate differences of other points and the absolute difference value thereof at a threshold value Td2Number within range, establishing pseudo-homonymous point pair set S according to matching point pair with maximum numberpcpOne-to-many or many-to-one matching phenomenon can occur in the set, and the point pairs of max (rho (x) + rho (y)) are selected to ensure one-to-one matching, so that pseudo-homonymous point pairs are obtained;
step 2.5: coarse registration;
interpolating elevation of pseudo-homonymous point pairs by using filtered ground point clouds to obtain [ X ]als Yals Zals]TAnd [ X ]vls YvlsZvls]TRespectively representing ALS and VLS characteristic point coordinates of the pseudo homonymy point pairs, wherein coordinate conversion relations are represented by formulas (4), (5) and (6), and coordinate conversion parameters are solved by using least squares;
Figure GDA0003504828540000072
Figure GDA0003504828540000073
Figure GDA0003504828540000074
wherein R (α, β, γ) represents a rotation matrix, α, β, γ being Euler rotation angles around the Z, Y and X axes, respectively; [ T ]xTy Tz]TRepresenting the amount of translation in the direction X, Y, Z; λ represents the scale parameter, and since ALS and VLS are registered as rigid transformations, let λ be 1.
During specific implementation, the model based on local neighborhood similarity measure has the advantages of strong stability, high efficiency, high reliability and the like, and provides guarantee for quick matching of airborne and vehicle-mounted LiDAR pseudo-homonymous point pairs.
And step 3: the method comprises the steps that angle points of a vehicle-mounted LiDAR building are expanded outwards based on a direction prediction algorithm, and a potential eave feature point set is constructed;
because airborne and onboard LiDAR have different scanning perspectives, there is a verge deviation in the extracted building corner points. The eave condition of a certain corner of the building facade may be no eave, one edge exists and two edges exist, so that the eave characteristic points of the building facade have 4 possibilities, as shown in fig. 2, which is called as potential eave characteristic points in the patent.
In a further embodiment, step 3 specifically includes the following steps:
step 3.1: determining the position of a potential feature point set of the eave;
in FIG. 2, L'1、L’2Are respectively a contour line L1And L2The characteristic line of the width of one eave is expanded, and potential eave characteristic points comprise L1And L2、L1And L'2、L’1And L2、L’1And L'2The intersection point of (a).
Step 3.2: judging the outward expansion direction;
firstly, sequencing feature points of ALS data in a clockwise direction by taking the feature point with the minimum horizontal coordinate of the current building as a starting point, and determining a starting point and an end point of each section of contour line. And selecting the pseudo homonymous point pairs one by one, respectively searching the adjacent characteristic points, reordering the VLS characteristic points according to the slope similarity principle, and determining the starting point and the end point of the contour line.
Predictions can be made from the geometric relationships: when the abscissa of the starting point is smaller than the abscissa of the ending point, the upper side of the straight line on which the contour line is located is the outer side (e.g., L in FIG. 3)1A is a starting point and b is an end point), and the contour line is corrected along the Y axis; when the abscissa of the starting point is larger than the abscissa of the ending point, the lower side of the straight line where the contour line is located is the outer side (e.g. L in FIG. 3)2C is the starting point a is the end point), the contour line should be corrected downward along the Y-axis.
Step 3.3: expanding a vehicle-mounted LiDAR contour line;
and setting the linear equation Ax + By + C of the contour line as 0, and determining the relation between the width d of the eave and the translation distance of the contour line along the Y axis according to geometry. In order to simplify the calculation formula, the coefficient A of the linear equation is converted into a non-negative form, and the VLS contour line is expanded by using a formula (7). And 4 plane positions of potential eave characteristic points of each angle are obtained through the intersection points of the adjacent contour lines, and the elevation of the potential eave characteristic points is determined by adopting an elevation interpolation method.
Figure GDA0003504828540000081
In the formula, A, B, C represents the coefficient of the general formula of the linear equation, Xstart、XendThe horizontal coordinates of the starting point and the end point of the contour line and the width of the building eave are shown as d.
In specific implementation, the method obtains four potential eave feature points of each building angular point by predicting the outward expansion criterion through reasonable design direction and utilizing the intersection points between adjacent contour lines and the outward expansion contour lines, realizes the construction of a potential eave feature point set, and lays a foundation for the matching of subsequent true homonymy point pairs. A detailed flow chart of the onboard and onboard LiDAR point cloud registration method that accounts for eaves features is shown in figure 4.
And 4, step 4: and respectively realizing the matching and Iterative fine registration of true homonymous point pairs of airborne and vehicle-mounted LiDAR point clouds by utilizing an Iterative Closest Point (ICP) algorithm twice.
Based on the coarse registration result, potential eave feature points are used as a search set, and an ICP algorithm is used for obtaining true homonymy point pairs; and on the basis, the ICP algorithm is used again to realize iterative fine registration.
In a further embodiment, step 4 specifically includes the following steps:
step 4.1: taking a coarse registration result as an initial value of ICP, taking a potential eave feature point set as a search point set and constructing a KD tree to improve search efficiency, and taking minimization of Euclidean distance E (R, T) between homonymous points as an optimal convergence condition of the ICP algorithm, so that a true homonymous point matching problem can be expressed as a point pair corresponding to the minimum value of a solving function E (R, T), wherein E (R, T) is as formula (8);
Figure GDA0003504828540000091
step 4.2: obtaining minimum E (R, T) after iterative optimization, wherein the point pair for resolving the registration parameters is a true homonymy point pair of ALS and VLS building characteristic points;
step 4.3: the method comprises the steps that vehicle-mounted LiDAR point cloud data is used as a search point set, a true homonymy point matching result based on an ICP method is used as an initial value of the ICP, minimization of Euclidean distance between homonymy points is used as an optimal convergence condition of the ICP algorithm, and a point cloud fine registration problem can be expressed as a registration parameter corresponding to the minimum value of a solving function E (R, T);
step 4.4: and obtaining the minimum E (R, T) after iterative optimization, and then the ALS and VLS data registration parameters are shown as a formula (9).
Figure GDA0003504828540000092
Where R, T is the ALS and VLS final registration parameter, RRi、TTiA rotation and translation matrix for each calculation of the iterative process, j being the final iteration number, Ti＝RiTi-1+TTi。
In specific implementation, the ICP iteration for the first time can realize true homonymy point pair matching, powerful quality guarantee is provided for iterative precise registration, and then the ICP algorithm is utilized again to realize precise registration. The method has good applicability to registration of airborne and vehicle-mounted LiDAR point cloud data under the influence of eaves.
In summary, the invention provides an airborne and vehicle-mounted LiDAR point cloud registration method taking into account eave features, the method comprising: firstly, denoising and filtering airborne and vehicle-mounted LiDAR point cloud data, and extracting building point cloud and characteristic angular points from ground object points; establishing a local neighborhood similarity measure model, realizing quick automatic matching of airborne and vehicle-mounted LiDAR pseudo-homonymy points, and realizing coarse registration by using pseudo-homonymy points; then, expanding the corner points of the vehicle-mounted LiDAR building based on a direction prediction algorithm to construct a potential eave feature point set; and finally, respectively realizing airborne and vehicle-mounted LiDAR point cloud true homonymy point pair matching and iterative fine registration by utilizing an ICP algorithm twice. By the method, high-precision registration of the airborne LiDAR and the vehicle-mounted LiDAR is realized, and the problems of few homonymous features, low registration precision and the like of the airborne LiDAR and the vehicle-mounted LiDAR under the influence of eaves are effectively solved.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make modifications, alterations, additions or substitutions within the spirit and scope of the present invention.

Claims (4)

1. An airborne and vehicle-mounted LiDAR point cloud registration method considering eave features is characterized in that: the method comprises the following steps:
step 1: denoising and filtering data of airborne LiDAR and vehicle-mounted LiDAR point clouds, and extracting building point clouds and characteristic angular points from ground object points;
step 2: establishing a local neighborhood similarity measure model, realizing rapid automatic matching of the onboard LiDAR and the vehicle-mounted LiDAR pseudo homonymous points, and realizing coarse registration by using the pseudo homonymous points;
and step 3: expanding the outline of the vehicle-mounted LiDAR building based on a direction prediction algorithm to construct a potential eave feature point set;
and 4, step 4: and respectively realizing the matching and iterative fine registration of the airborne and vehicle-mounted LiDAR point cloud true homonymy point pairs by using a twice iterative nearest point algorithm.
2. The method of airborne and onboard LiDAR point cloud registration accounting for eave features of claim 1, wherein: in step 2, the local neighborhood similarity measure model is established as follows:
step 2.1: calculating the local gravity center of the feature points;
one point ap is selected from the airborne LiDAR feature points in sequence, and k nearest feature points ap belonging to the same building are searched for, apn1，…，apnk-1Calculating the center of gravity O of the k pointsap；
From vehicleSequentially selecting one point vp from LiDAR characteristic points, and searching k nearest characteristic points vp and vpn belonging to the same building1，…，vpnk-1Calculating the gravity center O of the k pointsvp；
Step 2.2: evaluating local neighborhood similarity;
calculating a correlation coefficient of the ALS characteristic point coordinate difference sequence and the VLS characteristic point coordinate difference sequence by using formulas (1) and (2);
Figure FDA0003504828530000011
Figure FDA0003504828530000012
in the formula (I), the compound is shown in the specification,
Figure FDA0003504828530000013
gix、giyrespectively being the abscissa and ordinate g 'of the ith neighborhood point of the ALS point cloud current point'ix、g′iyRespectively the horizontal and vertical coordinates of the ith neighborhood point of the current point in the VLS point cloud;
calculating ap to OapAzimuthal angle of and vp to OvpAnd calculating ap to OapEuclidean distance of and vp to OvpThe difference in Euclidean distance of (c);
step 2.3: constructing a similarity measure model;
each feature point in the VLS and the current ALS feature point are matched and calculated, and when the absolute value of the horizontal and vertical correlation coefficient is larger than a threshold value TρAp to OapAzimuthal angle of and vp to OvpIs less than a threshold value TaWhen and ap to OapEuclidean distance of and vp to OvpHas an absolute difference value of Euclidean distance smaller than a threshold value Td1When the initial matching set S is stored, namely the formula (3) is satisfiedsimPerforming the following steps;
Figure FDA0003504828530000021
in the formula, angle apOapIs ap to OapAngle of < vpOvpIs vp to OvpThe azimuth of (d); | ap-OapL is ap to OapThe Euclidean distance of; i vp-OvpL is vp to OvpThe Euclidean distance of; | | is an absolute value;
repeating the step 2.1 to the step 2.3 until all ALS characteristic points are processed;
step 2.4: random sampling consistency matching is carried out to determine an optimal matching pair;
from the set SsimRandomly selecting a pair of matching points and calculating the horizontal and vertical coordinate difference, and counting the horizontal and vertical coordinate differences of other points and the absolute difference value thereof at a threshold value Td2Number within range, establishing pseudo-homonymous point pair set S according to matching point pair with maximum numberpcpOne-to-many or many-to-one matching phenomenon can occur in the set, and the point pairs of max (rho (x) + rho (y)) are selected to ensure one-to-one matching, so that pseudo-homonymous point pairs are obtained;
step 2.5: coarse registration;
interpolating elevation of pseudo-homonymous point pairs by using filtered ground point clouds to obtain [ X ]als Yals Zals]TAnd [ X ]vls Yvls Zvls]TRespectively representing ALS and VLS characteristic point coordinates of the pseudo homonymy point pairs, wherein coordinate conversion relations are represented by formulas (4), (5) and (6), and coordinate conversion parameters are solved by using least squares;
Figure FDA0003504828530000022
Figure FDA0003504828530000023
Figure FDA0003504828530000024
wherein R (α, β, γ) represents a rotation matrix, α, β, γ being Euler rotation angles around the Z, Y and X axes, respectively; [ T ]x TyTz]TRepresenting the amount of translation in the direction X, Y, Z; λ represents the scale parameter, and since ALS and VLS are registered as rigid transformations, let λ be 1.
3. The method of airborne and onboard LiDAR point cloud registration accounting for eave features of claim 1, wherein: in step 3, the vehicle-mounted LiDAR building corner points are extended based on a direction prediction algorithm, and the steps of constructing a potential eave feature point set are as follows:
step 3.1: determining the position of a potential feature point set of the eave;
the eave condition of a certain corner of the facade of the building comprises three conditions of no eave, existence on one side and existence on two sides, so that eave characteristic points of the building have 4 possibilities, the eave characteristic points are called as potential eave characteristic points, the eave potential characteristic points at a certain position comprise intersection points of contour lines and adjacent contour lines, intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the contour lines and the adjacent contour lines, intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the contour lines and intersection points of the contour lines, which extend the eave width outwards along the vertical line direction, of the adjacent contour lines, and the eave width outwards along the vertical line direction;
step 3.2: judging the outward expansion direction;
firstly, sequencing ALS data by taking a feature point with the minimum horizontal coordinate of a current building as a starting point and clockwise, and determining a starting point and an end point of each section of contour line; selecting pseudo homonymous point pairs one by one, respectively searching characteristic points adjacent to the pseudo homonymous point pairs, reordering VLS characteristic points according to a slope similarity principle, and determining a starting point and an end point of a contour line;
making predictions from the geometric relationships: when the abscissa of the starting point is smaller than the abscissa of the end point, the upper side of the straight line where the contour line is located is the outer side, and the contour line is corrected along the Y-axis direction; when the abscissa of the starting point is larger than the abscissa of the end point, the lower side of the straight line where the contour line is located is the outer side, and the contour line is corrected downwards along the Y axis;
step 3.3: expanding a vehicle-mounted LiDAR contour line;
setting a linear equation Ax + By + C of the contour line as 0, and determining the relation between the width d of the eave and the translation distance of the contour line along the Y axis according to geometry; in order to simplify the calculation formula, the linear equation coefficient A is changed into a non-negative form, and the VLS contour line is expanded by using a formula (7); obtaining 4 plane positions of potential eave feature points of each corner through the intersection points of adjacent contour lines, and determining the elevation of the potential eave feature points by adopting an elevation interpolation method;
Figure FDA0003504828530000031
in the formula, A, B, C represents the coefficient of the general formula of the linear equation, Xstart、XendThe horizontal coordinates of the starting point and the end point of the contour line and the width of the building eave are shown as d.
4. The method of airborne and onboard LiDAR point cloud registration accounting for eave features of claim 1, wherein: in step 4, the method respectively realizes the matching and iterative fine registration of airborne and vehicle-mounted LiDAR point cloud true homonymy point pairs by using a twice iterative nearest point algorithm, and comprises the following steps:
step 4.1: taking a coarse registration result as an initial value of ICP, taking a potential eave feature point set as a search point set and constructing a KD tree to improve search efficiency, and taking minimization of Euclidean distance E (R, T) between homonymous points as an optimal convergence condition of the ICP algorithm, so that a true homonymous point matching problem is expressed as a point pair corresponding to the minimum value of a solving function E (R, T), wherein E (R, T) is as formula (8);
Figure FDA0003504828530000041
step 4.2: obtaining minimum E (R, T) after iterative optimization, wherein the point pair for resolving the registration parameters is a true homonymy point pair of ALS and VLS building characteristic points, namely ALShAnd VLSh；
Step 4.3: the method comprises the steps that vehicle-mounted LiDAR point cloud data is used as a search point set, a true homonymy point matching result based on an ICP method is used as an initial value of the ICP, minimization of Euclidean distance between homonymy points is used as an optimal convergence condition of the ICP algorithm, and a point cloud fine registration problem can be expressed as a registration parameter corresponding to the minimum value of a solving function E (R, T);
step 4.4: and obtaining the minimum E (R, T) after iterative optimization, and then the ALS and VLS data registration parameters are shown as the formula (9):
Figure FDA0003504828530000042
where R, T is the ALS and VLS final registration parameter, RRi、TTiA rotation and translation matrix for each calculation of the iterative process, j being the final iteration number, Ti＝RiTi-1+TTi。
CN202110179598.5A 2021-02-07 2021-02-07 Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics Active CN113012206B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110179598.5A CN113012206B (en) 2021-02-07 2021-02-07 Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110179598.5A CN113012206B (en) 2021-02-07 2021-02-07 Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics

Publications (2)

Publication Number Publication Date
CN113012206A CN113012206A (en) 2021-06-22
CN113012206B true CN113012206B (en) 2022-03-25

Family

ID=76385169

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110179598.5A Active CN113012206B (en) 2021-02-07 2021-02-07 Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics

Country Status (1)

Country Link
CN (1) CN113012206B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113837326B (en) * 2021-11-30 2022-03-25 自然资源部第一海洋研究所 Airborne laser sounding data registration method based on characteristic curve
CN115951371B (en) * 2023-03-14 2023-06-09 山东科技大学 Method for determining equivalent points of airborne LiDAR sounding navigation belt
CN118071842A (en) * 2024-04-24 2024-05-24 山东科技大学 Ship identity recognition system based on camera calibration and deep learning algorithm

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103324916A (en) * 2013-06-07 2013-09-25 南京大学 Registration method for vehicle-mounted LiDAR data and aviation LiDAR data based on building outlines
CN105844629A (en) * 2016-03-21 2016-08-10 河南理工大学 Automatic segmentation method for point cloud of facade of large scene city building
CN106600690A (en) * 2016-12-30 2017-04-26 厦门理工学院 Complex building three-dimensional modeling method based on point cloud data
US20190279420A1 (en) * 2018-01-19 2019-09-12 Sofdesk Inc. Automated roof surface measurement from combined aerial lidar data and imagery

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103324916A (en) * 2013-06-07 2013-09-25 南京大学 Registration method for vehicle-mounted LiDAR data and aviation LiDAR data based on building outlines
CN105844629A (en) * 2016-03-21 2016-08-10 河南理工大学 Automatic segmentation method for point cloud of facade of large scene city building
CN106600690A (en) * 2016-12-30 2017-04-26 厦门理工学院 Complex building three-dimensional modeling method based on point cloud data
US20190279420A1 (en) * 2018-01-19 2019-09-12 Sofdesk Inc. Automated roof surface measurement from combined aerial lidar data and imagery

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
"一种序列影像和地面激光点云的配准方法";刘燕萍等;《激光杂志》;20171025(第10期);第1-3节 *
"利用球心投影与线特征的点云与全景影像配准";岳明宇等;《遥感信息》;20170215(第01期);第1-3节 *
"基于三维激光扫描与无人机倾斜摄影技术的古建筑重建";何原荣等;《遥感技术与应用》;20191220(第06期);第1-3节 *
"基于多源LiDAR数据的LoD3城市建筑物模型自动重建研究";李瑶;《硕士电子期刊》;20181215(第12期);第1-4章 *
"基于方向预测规则化的机载激光雷达建筑物正交轮廓线提取";郭亚栋等;《激光与光电子学进展》;20190902;第1-11页 *
"基于空-地多源数据融合的建筑物精细建模研究";李永强等;《测绘工程》;20150825(第08期);第1-4节 *
"联合机载-车载LiDAR点云的建筑物立面建模研究";杨清科;《硕士电子期刊》;20200715(第07期);第1-5章 *

Also Published As

Publication number Publication date
CN113012206A (en) 2021-06-22

Similar Documents

Publication Publication Date Title
CN113012206B (en) Airborne and vehicle-mounted LiDAR point cloud registration method considering eave characteristics
CN111882612B (en) Vehicle multi-scale positioning method based on three-dimensional laser detection lane line
WO2021232463A1 (en) Multi-source mobile measurement point cloud data air-ground integrated fusion method and storage medium
CN110717983A (en) Building facade three-dimensional reconstruction method based on knapsack type three-dimensional laser point cloud data
CN102506824B (en) Method for generating digital orthophoto map (DOM) by urban low altitude unmanned aerial vehicle
CN103703490B (en) For generation of the equipment of three-dimensional feature data and the method for generation of three-dimensional feature data
CN110006408B (en) LiDAR data cloud control aerial image photogrammetry method
CN103324916B (en) Vehicle-mounted and aviation LiDAR data method for registering based on building profile
CN111354083B (en) Progressive building extraction method based on original laser point cloud
CN115372989A (en) Laser radar-based long-distance real-time positioning system and method for cross-country automatic trolley
CN112799096B (en) Map construction method based on low-cost vehicle-mounted two-dimensional laser radar
CN112465732A (en) Registration method of vehicle-mounted laser point cloud and sequence panoramic image
Shen et al. A new algorithm of building boundary extraction based on LIDAR data
CN112767461A (en) Automatic registration method for laser point cloud and sequence panoramic image
CN109146990B (en) Building outline calculation method
CN112270698A (en) Non-rigid geometric registration method based on nearest curved surface
Rebelo et al. Building 3D city models: Testing and comparing Laser scanning and low-cost UAV data using FOSS technologies
CN117029870A (en) Laser odometer based on road surface point cloud
CN111899291A (en) Automatic registration method for coarse-to-fine urban point cloud based on multi-source dimension decomposition
Monnier et al. Registration of terrestrial mobile laser data on 2D or 3D geographic database by use of a non-rigid ICP approach.
RU2736506C1 (en) Method for automatic local increase in accuracy of aerial laser scanning using ground-based laser scanning data
CN114089376A (en) Single laser radar-based negative obstacle detection method
Choi et al. Automatic Construction of Road Lane Markings Using Mobile Mapping System Data.
CN112767460B (en) Spatial fingerprint image registration element feature description and matching method
Zhu et al. Research on urban construction land change detection method based on dense dsm and tdom of aerial images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant