CN114549879B

CN114549879B - Target identification and central point extraction method for tunnel vehicle-mounted scanning point cloud

Info

Publication number: CN114549879B
Application number: CN202210443723.3A
Authority: CN
Inventors: 贾洋; 孙璐; 李升甫; 夏季; 柯勇; 张弛; 杨宇雷; 李鹏; 汪致恒; 丁雨淋; 王东; 刘长风; 王义鑫
Original assignee: Sichuan Highway Planning Survey and Design Institute Ltd
Current assignee: Sichuan Highway Planning Survey and Design Institute Ltd
Priority date: 2022-04-25
Filing date: 2022-04-25
Publication date: 2022-07-08
Anticipated expiration: 2042-04-25
Also published as: CN114549879A

Abstract

The invention relates to the field of vehicle-mounted laser scanning surveying and mapping, and discloses a target identification and central point extraction method for tunnel vehicle-mounted scanning point cloud, which is beneficial to effectively identifying a target in the vehicle-mounted laser point cloud, realizes high-precision extraction of a target central coordinate, and provides important technical support for geometric correction of subsequent point cloud data. In the target point cloud identification process, three-dimensional laser scanning data of a tunnel wall is obtained by using vehicle-mounted mobile laser scanning equipment, and the target area point cloud data is rapidly extracted based on multi-level spatial position constraint and target geometric constraint; in the target center point extraction process, the incomplete characteristics of target point clouds are considered, the geometric center calculation error caused by target missing is effectively made up by using a standard template edge fitting and matching method, and the target center point coordinates are accurately calculated. The invention is particularly suitable for the mobile measurement operation of the highway tunnel in the mountainous area.

Description

Target identification and central point extraction method for tunnel vehicle-mounted scanning point cloud

Technical Field

The invention relates to the field of vehicle-mounted laser scanning surveying and mapping, in particular to a target identification and central point extraction method for tunnel vehicle-mounted scanning point clouds.

Background

With the development of a vehicle-mounted mobile three-dimensional laser measurement system, the technology, as a geographic space data acquisition mode, has become a main data source for large-scene three-dimensional point cloud data acquisition, has the characteristics of safety, accuracy and high efficiency, plays a vital role in the application of related fields such as road traffic and the like, and has great potential in the aspects of planning management, operation maintenance, digital modeling, asset element extraction and the like of roads.

However, the precision of the point cloud data acquisition of the technology depends on good continuous GNSS signals, which are abbreviated as Global Navigation Satellite System (Global Navigation Satellite System), and chinese is generally called Global Navigation Satellite System (GNSS). Therefore, in the surveying and mapping of an extra-long tunnel with the length of more than 3 kilometers, due to the loss of GNSS signals, the pose data of the vehicle-mounted mobile measurement system can only be provided through the inertial navigation system and the odometer, so that sequential error accumulation is caused, the spatial position of the point cloud data in the acquired tunnel is misaligned, and the requirement of high-precision point cloud data acquisition cannot be met.

Therefore, the applicant applies for a geometric correction method of vehicle-mounted scanning point cloud under a tunnel GNSS rejection environment and an invention patent with the patent number of 2021108201564, wherein measurement targets are reasonably arranged in the tunnel environment, the center points of the targets are controlled and measured in advance, the tunnel targets are used as known control points, and geometric correction is performed on continuous point cloud data obtained by vehicle-mounted laser scanning, so that the problem of overall measurement of the highway tunnel in the mountainous area is solved efficiently and precisely.

However, during the actual movement measurement work, the following problems are faced:

most of vehicle-mounted laser scanning belongs to a linear array scanning mode, in order to guarantee high-efficiency operation of vehicle-mounted mobile measurement, the vehicle speed is generally not lower than 60km/h, the point cloud data density of a target obtained by mobile measurement is small, and the target area is partially lost, so that high-precision extraction of a target center in the later period is not facilitated.

And further, under the condition of GNSS signal loss of the long-distance tunnel, effective geometric correction cannot be performed after the point cloud of the vehicle-mounted laser is deviated, and further the high-precision measurement of the whole tunnel is influenced.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the target identification and center point extraction method of the tunnel vehicle-mounted scanning point cloud is beneficial to effectively identifying the target in the vehicle-mounted laser point cloud, realizes high-precision extraction of the center coordinate of the target, and provides important technical support for geometric correction of subsequent point cloud data.

The technical scheme adopted by the invention for solving the technical problems is as follows:

on one hand, the invention provides a target identification method of tunnel vehicle-mounted scanning point cloud, and the target adopted by the identification method comprises a main body and strips distributed along the edge of the main body, wherein the main body is made of a material with the laser reflectivity higher than that of a tunnel side wall, the strips are made of a material with the laser reflectivity lower than that of the tunnel side wall, and the main body is in a regular shape with a central point; the identification method comprises the following steps:

s1, aiming at the current identification target, baseSetting a height condition at the set height of the target, dividing the scanning point cloud data, only keeping the point cloud data of the points meeting the set height condition, and constructing an initial point cloud data set corresponding to the current identification targetQ；

S2, extracting the track time of the vehicle-mounted scanning system corresponding to the target based on the set position of the current identification target along the track direction, and calculating the cross line corresponding to the track position of the track time;

s3, in the initial point cloud data setQIn the method, based on each cross-sectional line corresponding to the current identification target, point cloud data of points meeting distance conditions are screened along the track direction, and an area point cloud data set corresponding to the target is constructedR(ii) a The distance condition is a screening distance set based on the scanning characteristic of the vehicle-mounted scanning system, the track positioning error and the target positioning error along the track direction so as to ensure that the distance condition is the screening distanceRThe point cloud data of the corresponding target is contained;

s4, aiming at the area point cloud data set corresponding to the current identification targetRBased on the intensity of the reflectionI _pSorting the point cloud data from big to small, and extracting the top sorted point cloud datan _maxPoint cloud data of points to form a seed point cloud data set corresponding to the targetB _p；

S5, aiming at the seed point cloud data set corresponding to the current identification targetB _pGo through the points it includesP _BiRespectively at theB _pCorresponding regional point cloud data setRExtract and point ofP _BiPoint cloud data of points whose spatial distances meet set constraint conditions are formed corresponding to the pointsP _BiOf the image spot point cloud data setB _tiThe constraint condition is a shape constraint condition set based on the shape size of the target so as to ensure each image spot point cloud data setB _tiThe shape of the region formed by the points in (1) is a shape adapted to the shape of the target body;

s6, aiming at each image spot point cloud data setB _tiFor respectively included points thereofThe point cloud data is subjected to reflection intensity judgment, and point cloud data of points with reflection intensity higher than a set threshold value is extracted to form a point cloud data corresponding to the point cloud dataB _tiSubject candidate point cloud dataset ofB _wi；

S7, setting geometric constraint conditions based on the shape and size of the target, and aiming at the current identification target, respectively collecting the corresponding main body alternative point cloud data setsB _wiScreening according to the set geometric constraint condition, merging the targets corresponding to eachB _wiConstructing a main point cloud data set of the current identification target according to the point cloud data of the screened points meeting the geometric constraint conditionsB _w-final。

Further, in step S1, the height conditions are set as follows:

H _min－σ≤Z _point ≤H _max＋σ

wherein the content of the first and second substances,Z _pointis the height coordinate of the point cloud data,H _maxthe highest height of the height tolerance range is set for the target,H _minsetting a minimum height of a height tolerance range for the target, saidσA tolerance distance set based on the mounting error.

Further, in step S2, extracting a track time of the vehicle-mounted scanning system corresponding to the target based on the set position of the current identification target along the track direction includes:

if the current identification target is the first target, determining the track position of the vehicle-mounted scanning system corresponding to the current identification target as the moment along the track direction of the vehicle-mounted scanning system based on the setting position of the first targett ₀；

Otherwise, setting a distance interval based on the target along the track directionDAccording to the time when the previous target corresponds to the track position of the vehicle-mounted scanning systemt _iDetermining the time of the track position of the vehicle-mounted scanning system corresponding to the current identification targett _i+1：

t _i+1 －t _i=D/V

Wherein, the first and the second end of the pipe are connected with each other,Dthe distance along the track direction from the previous target to the currently identified target,Vis composed oft _iTot _i+1The average vehicle speed of the vehicle-mounted scanning system at the moment.

Further, in step S2, the cross-sectional line corresponding to the trajectory position at the trajectory time is determinedL _cThe point above, satisfies the following constraint:

wherein the content of the first and second substances,x、yis a transverse broken lineL _cGo up and down at each pointxyCoordinates in a plane, saidxyThe plane is a projection plane of the point cloud data;P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the position of the track of the current targetyThe coordinates of the axes are set to be,P _{z_c}for the spatial elevation of the currently identified target's corresponding track location,zis a transverse broken lineL _cThe spatial elevation of (a);P _kfor the current identification target corresponding to the track positionxySlope of tangent in plane.

Further, in step S3, in the initial point cloud data setQScreening point cloud data of points meeting distance conditions along the track direction based on a cross-sectional line corresponding to the current identification target, and constructing an area point cloud data set corresponding to the targetRThe method specifically comprises the following steps:

s31, extracting the track position corresponding to the current identification target and the cross-sectional line corresponding to the track positionL _c；

S32, setting screening conditions based on the transverse width of the tunnel, and traversing the initial point cloud data setQSearching points meeting the screening condition, and constructing an alternative point cloud data set of the current identification targetT；

S33, traversing the alternative point cloud data set of the current identification targetTAll points in (1), calculatingEach point to corresponding transverse lineL _cThe spatial sag of (c);

s34 based on the point cloud data setTTo the corresponding cross-sectional lineL _cPoint of minimum spatial sagP _T-minVdFrom an initial point cloud data setQExtracting points meeting the following constraint conditions, and constructing a local area data set of the current targetR：

x(P _T-minVd )－d _L ＜x _{Q_p} ＜ x(P _T-minVd ) ＋d _L

y(P _T-minVd )－d _L ＜y _{Q_p} ＜ y(P _T-minVd ) ＋d _L

Wherein, the first and the second end of the pipe are connected with each other,x _{Q_p}、y _{Q_p}as a data setQPoint of (5)PIsxShaft andythe coordinates of the axes are set to be,x(P _T-minVd )、y(P _T-minVd )is a pointP _T-minVdIs/are as followsxShaft andythe coordinates of the axes are set to be,d _Lis a distance condition set based on the trajectory positioning error and the target positioning error.

Further, in step S32, the screening conditions set based on the tunnel lateral width are:

P _{x_c} －n _c ×W _c ＜x _p ＜ P _{x_c} ＋n _c ×W _c

P _{y_c} －n _c ×W _c ＜y _p ＜ P _{y_c} ＋n _c ×W _c

wherein the content of the first and second substances,x _p 、y _pfor an initial point cloud data setQInPOf dotsxShaft andythe coordinates of the axes are set to be,P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,n _cas to the number of lanes of the vehicle,W _ca single lane width.

Further, in step S33, the points to the corresponding cross-sectional lines are calculated as followsL _cThe space vertical distance of (2):

wherein the content of the first and second substances,V _difor alternative point cloud data setsTTo middleiFrom point to transverse lineL _cThe space vertical distance of the air inlet pipe is less than the air outlet pipe,x _piandy _pirespectively alternative point cloud data setsTTo middleiOf dotsxAxis coordinate sumyAxis coordinates;P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,P _kfor current identificationThe target corresponding to the track positionxySlope of a tangent in a plane, saidxyThe plane is a projection plane of the point cloud data.

Further, in step S6, the setting of the reflection intensity threshold value is:

wherein the content of the first and second substances,I _threshold-iis as followsiIndividual image spot point cloud data setB _tiThe determination threshold value of the reflection intensity of (b),max I{B _ti }is as followsiIndividual image spot point cloud data setB _tiThe maximum reflected intensity value of all the points in (c),I _p-minis as followsiIndividual image spot point cloud data setB _tiCorresponding local area data setRThe lowest reflected intensity value of.

Preferably, the target body is square.

Further, in step S5, the shape constraint conditions set based on the shape size of the target are:

wherein the content of the first and second substances,D _bis composed ofP _BiAndRthe spatial distance of the points in (a),L _wlis the side length of the body of the target,W _bwthe width of the strip that is the target;

the above-mentionedD _bThe calculation formula is as follows:

wherein the content of the first and second substances,x(P _Bi )、y(P _Bi )andz(P _Bi )are respectively a pointP _BiIs/are as followsxA shaft,yShaft andzaxis coordinates;x(P _Rj )、y(P _Rj )andz (P _Rj )are respectively asRMiddle removingP _BiExternal firstjOf dotsxA shaft,yShaft andzaxis coordinates.

Further, in step S7, the geometric constraint conditions are set based on the shape and size of the target as follows:

wherein the content of the first and second substances,max DeltaZ {p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiThe maximum height difference between all the point pairs;max D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiThe maximum spatial distance between all the point pairs;L _wlis the side length of the body of the target,W _bwis the width of the swath of the target,L _wdis the diagonal length of the body of the target;

subject candidate point cloud datasetB _wiThe spatial distance between the middle point pairs is calculated as follows:

subject candidate point cloud datasetB _wiThe height difference between the middle point pairs is calculated as follows:

wherein the content of the first and second substances,x(p _wk )、y(p _wk )andz(p _wk )are respectively asB _wiTo middlekDotp _wkIs/are as followsxA shaft,yShaft andzaxis coordinates;x(p _wj )、y(p _wj )andz(p _wj )are respectively asB _wiTo middlejDotp _wjIsxA shaft,yShaft andzaxis coordinates;D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe spatial distance of (a);DeltaZ {p _wk ,p _wj }candidate point cloud data set for subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe difference in height of (a).

On the other hand, the invention also provides a target center point extraction method of tunnel vehicle-mounted scanning point clouds, which is used for carrying out center point extraction on the target identified by the target identification method on the tunnel with the square target, and the extraction method comprises point cloud data processing, vector data preparation and data matching;

wherein the vector data preparation comprises the following steps:

a1, loading vector line frame data in accordance with the target measuring region size in the computer based on the target design sizeL _b(ii) a The target measuring region is a target body region within a wire frame enclosed by the strips of the target;

a2, line frame data of vectorL _bWill convert to a single edgeUThe divided point data and the four sides are generated in total 4U+1 dot data, dot-to-dot spacing of: (L _wl-W _bw)/UConstructing a vector line frame data point setB _lbWherein, in the step (A),L _wlis the side length of the body of the target,W _bwthe width of the strip that is the target;

the point cloud data processing comprises the following steps:

b1, extracting a main body point cloud data set of the current identification targetB _w-finalEach scanning linear array based on the vehicle-mounted scanning systemB _w-finalFirst recording spot ofp _ti-fAnd the last recording spotp _ti-eConstructing an initial body contour point cloud dataset of the targetB _wl；

B2 Point cloud dataset of initial body contours of the targetB _wlIn the method, data interpolation is carried out to construct a body contour point cloud data set of the targetB _wl-b(ii) a The subject contour point cloud datasetB _wl-bData point number and vector line frame data point setB _lbThe number of the data points in the data base is consistent;

the data matching comprises the following steps:

c1, using SVD least square method to identify the vector line frame data point set of the target currentlyB _lbWith subject contour point cloud data setB _wl-bMatching is carried out;

c2, based on the matching result in step C1, the vector line frame data point set is matchedB _lbRotating and translating to obtain the target vector line frame data point set after rotating and translatingB _lb-finalCalculatingB _lb-finalGeometric center ofB _P-centerAs the center point of the target.

Further, in step a2, the U equal score is calculated as follows:

wherein the content of the first and second substances,H _Lthe linear array scanning frequency of the laser equipment of the vehicle-mounted scanning system is obtained.

Further, in step B2, the point cloud data set is recorded in the initial body contour of the targetB _wlIn the method, data interpolation is carried out to construct a body contour point cloud data set of the targetB _wl-bThe method specifically comprises the following steps:

b21, calculating an initial subject contour point cloud data setB _wlWith vector line frame data point setB _lbDifference in number of data points indelta_n：

delta_n=4U+1-n(B _wl)

Wherein the content of the first and second substances,n(B _wl) Point cloud data set for initial subject profileB _wlThe number of data points in (a);

b22, collecting the initial subject contour point cloud data setB _wlThe points in the target are sorted based on the scanning time to respectively obtain a data point set of each scanning linear array entering a target main body areaFAnd a set of data points away from the target body regionE；

B23 set of data points to be entered into target body regionFAnd a set of data points away from the target body regionERespectively introducing interpolation modules, forming point pairs by the points of the adjacent scanning linear arrays corresponding to the data point sets, and inserting interpolation points between the point pairs to obtain the respectively corresponding data point setsFAnd a set of data pointsEThe interpolation point set of (2);

b24, collecting the interpolation point set obtained in the step B23 and the initial subject contour point cloud data setB _wlMerging to obtain a main body contour point cloud data setB _wl-b。

Further, in step B23, the interpolation module includes the following steps:

b231, sequentially calculating the spatial distance between the point pairs formed by the points of the adjacent scanning line arrays in the corresponding data point set:

wherein the content of the first and second substances,d _{delta_i}is a firstiThe strip scanning linear array and the adjacent secondi+1 scanning line array points to form the space distance between the point pairs;x(p _ScanL-t1 )、y(p _ScanL-t1 )andz(p _ScanL-t1 )for points at a point preceding the scanning timexA shaft,yShaft andzaxis coordinates;x(p _ScanL-t2 )、y(p _ScanL-t2 )andz(p _ScanL-t2 )for points subsequent to the scanning timexA shaft,yShaft andzaxis coordinates;

b232, sorting the space distances between the point pairs formed by the points of the adjacent scanning line arrays obtained in the step B231 from large to small;

if introduced as a set of data points into the target body regionFThen, determinedelta_nIf the parity number of (1) is even, the front one with the largest spatial distance is takendelta_n/2Point pairs; if the number is odd, the front (B) with the largest spatial distance is takendelta_n+ 1)/2Point pairs forming a data set of point pairs to be interpolatedC；

If introduced as a set of data points off the target body regionEThen, judgedelta_nIf the parity number of (1) is even, the front one with the largest spatial distance is takendelta_n/2Point pairs; if the number is odd, take the front with the largest space distance (delta_n- 1)/2Point pairs forming a data set of point pairs to be interpolatedC；

B233, generating interpolation points in sequence according to the following formula based on the point pairs obtained in step B232:

x _i=x(p _Cj1)+( x(p _Cj2)- x(p _Cj1))/2

y _i=y(p _Cj1)+( y(p _Cj2)- y(p _Cj1))/2

z _i=z(p _Cj1)+( z(p _Cj2)- z(p _Cj1))/2

wherein, the first and the second end of the pipe are connected with each other,x _i、y _iandz _iare respectively the firstiOf an interpolation pointxA shaft,yShaft andzaxis coordinates;x(p _Cj1)、y(p _Cj1) Andz(p _Cj1) Respectively for point pairs to be interpolatedCTo middlejOf points at a point pair whose scanning time is earlierxA shaft,yShaft andzaxis coordinates;x(p _Cj2)、y(p _Cj2) Andz(p _Cj2) Respectively for point pairs to be interpolatedCTo middlejOf points at a later scanning timexA shaft,yShaft andzaxis coordinates.

Further, in step C1, a vector wire frame data point set of the current recognition target is determined by SVD least squaresB _lbWith subject contour point cloud data setB _wl-bMatching, specifically comprising:

based on SVD least square method, calculating vector line frame data point setB _lbWith subject contour point cloud data setB _wl-bOptimal estimation transformation matrix of (1):

wherein the content of the first and second substances,P _{lb_i}for vector line frame data point setsB _lbTo middleiPoint;P _{wl_i}point cloud data set for subject profileB _wl-bTo (1) aiThe point of the light beam is the point,icalculating a sequence number for traversal;Ra rotation matrix that is a match between the two sets of data,Ta translation matrix that is a match between the two sets of data,Nfor vector line frame data point setsB _lbAnd subject contour point cloud data setB _wl-bThe number of dot data in (1).

Further, in step C2, the vector wireframe data point set is matched based on the matching result in step C1B _lbRotating and translating to obtain the target vector line frame data point set after rotating and translatingB _lb-finalCalculatingB _lb-finalGeometric center ofB _P-centerAs the center point of the target, specifically, the method includes:

based on the rotation matrix obtained in step C1RAnd translation matrixTSet vector line frame data pointsB _lbRotating and translating to obtain the target vector line frame data point set after rotating and translatingB _lb-finalAnd calculateB _lb-finalGeometric center ofB _P-center：

Wherein, the first and the second end of the pipe are connected with each other,x(P _{wl-final_i})、y(P _{wl-final_i}) Andz(P _{wl-final_i}) Respectively a set of target vector line frame data pointsB _lb-finalTo middleiOf point dataxA shaft,yShaft andzthe coordinates of the axes are set to be,iin order to compute the sequence number for the traversal,Nset of line-frame data points for target vectorB _lb-finalThe number of data points in (a).

The beneficial effects of the invention are:

through the target of supporting design, obtain the apparent intensity contrast between mark target main part and the tunnel side wall to constitute the frame of different reflection intensity through the strip, distinguish main part and tunnel side wall mutually, guaranteed that the later stage carries out effective discernment to the target in some cloud data.

Secondly, by utilizing multi-level spatial position constraint and target geometric constraint, point cloud data of a target core area can be accurately identified and extracted; then, the defect that the coordinate back calculation of the real center point of the target cannot be carried out due to the partial defects of the target caused by the sparse point cloud density is effectively overcome by combining a standard target template with a geometric centroid calculation method, and the high-precision calculation of the target center points on two sides of the tunnel roadside in the high-speed movement measurement process is realized.

The method has high extraction efficiency and flexible application mode, can process point cloud data after finishing point cloud scanning, and can also identify and extract in real time in the vehicle-mounted laser scanning motion process.

Drawings

FIG. 1 is a technical idea diagram of an embodiment of the present invention;

FIG. 2 is a schematic illustration of a target designed according to an embodiment of this invention;

fig. 3 is a schematic diagram of target body region contour point cloud extraction in an embodiment of the invention.

Detailed Description

The invention aims to provide a target identification and center point extraction method for tunnel vehicle-mounted scanning point cloud, which is beneficial to effectively identifying a target in vehicle-mounted laser point cloud, realizes high-precision extraction of target center coordinates, and provides important technical support for geometric correction of subsequent point cloud data.

The technical idea of the invention is as shown in fig. 1, firstly a target is designed in a matching way, the target comprises a main body and strips arranged along the edge of the main body, wherein the main body is made of a material with a laser reflectivity higher than that of a tunnel side wall, the strips are made of a material with a laser reflectivity lower than that of the tunnel side wall, and the main body is in a regular shape with a central point. The high laser reflectivity of the main body is used for better realizing laser scanning and avoiding data loss; the strip is used for forming a frame which is different from the main body and the side wall of the tunnel at the edge of the theme, so that the separation is realized, and the data identification and extraction at the later stage are convenient; the regular shape is used for facilitating the drawing and matching of the target template; having a center point is the goal of extracting the target center for geometric correction.

In the target point cloud identification process, the vehicle-mounted mobile laser scanning equipment is utilized, the equipment vehicle continuously advances to obtain tunnel cave wall three-dimensional laser scanning data, and multi-level spatial position constraint, reflection intensity constraint, target shape constraint and the like are carried out on the point cloud data obtained by scanning, so that the point cloud data of a target area are quickly and accurately identified; finally, in the process of extracting the target center point, considering the incomplete feature of the target point cloud, after the rough extraction of the target body contour point cloud data is carried out, the discretized standard template is utilized to carry out body contour point cloud data alignment, the aligned body contour point cloud data and the aligned template data are matched through an SVD least square method, according to the matching result, the template point cloud is subjected to translation conversion, and then the geometric center coordinate of the template is calculated to serve as the target center coordinate, so that the geometric center calculation error caused by the missing target is effectively compensated, and the target center point coordinate is accurately calculated.

Example (b):

the following three aspects of target design and installation, target point cloud identification and target center point extraction are specifically explained respectively:

firstly, designing and installing a target:

in this embodiment, a square target with significant intensity light contrast between the target body and the strip is used, the target is designed to have a white bottom and a black edge, that is, a white body and a black strip, and after the central point is marked, the target fixing work is completed on two sides of the wall of the highway tunnel. Specifically speaking:

step 1, designing a target to be a square, wherein the target mainly comprises a black strip and a white main body, and the design side length L of the main body of the target_wlWherein the black stripe is designed to have a width W_bwThe material of the target main body area is rough paper material, the black strips have no special material requirement, and the black strips are alignedThe slot is affixed to the target body edge. The larger the design side length of the target body is, the more points are obtained by scanning, but the larger the points are not suitable for being too large, so that the length is generally 50-60 cm; the wider the black stripe design width, the better the partitioning effect, but the wider the restriction on the theme, and therefore, is typically 5-10 cm.

And 2, finishing diagonal marking in a white main body measuring area within the strip by using a red neutral pen, or printing a picture by using a red diagonal of white matte paper by using an extremely thin red printing width such as 1mm, wherein the intersection point of the red diagonals is a target center point, and the marking of the center point is convenient for the control measurement of the target.

The target fabrication is completed, and the finished target is shown in fig. 2.

And 3, fixing the manufactured target on two sides of the tunnel wall, wherein the flatness of the fixed target needs to be ensured as much as possible. It is of course also possible to arrange the targets only on one side of the tunnel wall, for example when the tunnel width is small.

According to the size of the target and the general height of the human body, the target is not suitable to be installed more than 2m in consideration of the convenience of the human body in the target installation; secondly, considering the effective visual angle position when the vehicle-mounted equipment is moved and measured, when the target is installed too low, the information of the laser scanning target is easy to be incomplete, and the target distance and the ground height after the target is fixed are not less than 1.2 m.

Secondly, target point cloud identification:

step 1, combined filtering coarse extraction:

aiming at the current identification target, setting the following height conditions based on the set height of the target, dividing point cloud data of target adjacent height areas at two sides of a hole wall of vehicle-mounted mobile laser scanning data by utilizing the height conditions, completing point cloud data filtering processing of other non-target areas such as the ground, the hole top and the like, only reserving the point cloud data of points meeting the set height conditions, and constructing an initial point cloud data set corresponding to the current identification targetQ。

The set height conditions were:

H _min－σ≤Z _point ≤H _max＋σ

wherein the content of the first and second substances,Z _pointis the height coordinate of the point cloud data,H _maxthe highest height of the height tolerance range is set for the target,H _minsetting a minimum height of a height tolerance range for the target, saidσA tolerance distance set based on the mounting error. In the present embodiment, theH _maxAndH _minthe unit of (a) is a meter,σvalue of 0.3m。

Since the setting heights of the targets are uniform, point filtering or integral filtering can be performed for the step. For example, if the scan has been completed, the entire filter may be filtered. In this embodiment, the identification is performed in real time, that is, the identification is performed while scanning, and therefore, the filtering is performed as a single point, which may be performed for all the obtained point cloud data, or may be performed according to a set length that should meet the requirements of the subsequent steps, for example, the point cloud data obtained by the latest scanning that has not been identified yet in the obtained data.

And 2, extracting the track time of the vehicle-mounted scanning system corresponding to the target based on the set position of the current identification target along the track direction, and calculating a cross line corresponding to the track position of the track time. The extraction of the track time can adopt a manual confirmation mode, such as: in the point cloud scanning process, each time one target passes through, the target is manually confirmed.

However, in order to perform the measurement more conveniently, in this embodiment, the method specifically includes:

if the current identification target is the first target, determining the track position of the vehicle-mounted scanning system corresponding to the current identification target as the moment along the track direction of the vehicle-mounted scanning system based on the setting position of the first targett ₀。t ₀The confirmation of (2) may be a manual confirmation, or may be a setting of the point cloud scanning start point at the first target position. In this embodiment, the first target position is manually confirmed to correspond tot ₀The time of day mode is implemented.

If the current identification target is not topEach target is spaced along the track based on the set distance of the targetDAccording to the time when the previous target corresponds to the track position of the vehicle-mounted scanning systemt _iDetermining the time of the track position of the vehicle-mounted scanning system corresponding to the current identification targett _i+1：

t _i+1 －t _i=D/V

Wherein the content of the first and second substances,Dthe distance along the track direction from the previous target to the currently identified target,Vis composed oft _iTot _i+1The average vehicle speed of the vehicle-mounted scanning system at the moment.

The embodiment is real-time identification, that is, identification while scanning, and if the point cloud scanning is completed first and then identification is performed, a point-by-point circulation manner same as that in real time may be adopted, or circulation may be performed in steps, for example: and circularly executing the steps until the track time of the vehicle-mounted scanning system corresponding to all the targets is calculated.

The line of intersection is a term used in the traffic design art to denote the location of the trajectory where the plane perpendicular to the trajectory line isxyA projected line in a plane. Therefore, the cross-sectional line corresponding to the trajectory position at the trajectory time corresponding to the current recognition targetL _cThe point above, satisfies the following constraint:

wherein the content of the first and second substances,x、yis a transverse lineL _cGo up and down at each pointxyCoordinates in a plane of saidxyThe plane is a projection plane of the point cloud data;P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,P _{z_c}for the spatial elevation of the currently identified target's corresponding track location,zis a transverse broken lineL _cThe spatial elevation of (a);P _ktargeting for current identificationAt the position of the trackxySlope of tangent in plane.

In general, in the projection plane coordinate system, the x-axis points to the north direction, the y-axis points to the east direction, and the origin of the projection plane coordinate system is the origin of the gaussian banded projection coordinate system.

Step 3, regional space constraint: at the initial point cloud data setQIn the method, based on a cross-sectional line corresponding to a current identification target, point cloud data of points meeting distance conditions are screened along the track direction, and an area point cloud data set corresponding to the target is constructedR. The distance condition is a screening distance set based on the scanning characteristic of the vehicle-mounted scanning system, the track positioning error and the target positioning error along the track direction so as to ensure that the distance condition is the screening distanceRIncluding point cloud data of the corresponding target.

The track positioning error is based on inertial navigation positioning mainly in a GNSS rejection environment in a tunnel due to point cloud scanning, the positioning accuracy is low, and the accumulated position error reaches the meter level; the target positioning error is an error of target positioning caused by target installation error and target corresponding track moment extraction error; the scanning characteristic of the vehicle-mounted scanning system indicates that at the track moment, the current laser emission angle based on the scanning system has uncertainty, and the scanning point corresponding to the moment may be located on the transverse line, behind the transverse line or in front of the transverse line. Therefore, it is ensured thatRThe method comprises the steps of including point cloud data of corresponding targets, firstly considering dislocation caused by target positioning errors, then considering drift caused by track positioning errors, and finally considering intervals introduced by scanning characteristics.

The step is to complete the data screening along the track direction on the basis of completing the data screening along the height direction in the step 1, and the data screening can be directly intercepted according to the set distance condition. However, as mentioned above, direct interception is used to ensureRThe method comprises the steps of including point cloud data of a corresponding target, wherein the screening distance needs to consider a plurality of factors such as scanning characteristics of a vehicle-mounted scanning system along the track direction, track positioning errors, target positioning errors and the like, the numerical value is large, more noise points are introduced, and the calculation burden is increased.Therefore, in this embodiment, a step-by-step screening method is adopted, which specifically includes:

step 31, extracting the track position corresponding to the current identification target and the cross-sectional line corresponding to the track positionL _c。

Step 32, setting screening conditions based on the transverse width of the tunnel, and traversing the initial point cloud data setQSearching points meeting the screening condition, and constructing an alternative point cloud data set of the current identification targetT(ii) a Wherein, the screening condition that the tunnel transverse width set for is:

P _{x_c} －n _c ×W _c ＜x _p ＜ P _{x_c} ＋n _c ×W _c

P _{y_c} －n _c ×W _c ＜y _p ＜ P _{y_c} ＋n _c ×W _c

Step 33, traversing the currently identified targetsAlternative point cloud data setTCalculating all points to corresponding cross-sectional linesL _cThe space vertical distance of (2):

wherein the content of the first and second substances,V _difor alternative point cloud data setsTTo middleiFrom point to transverse lineL _cThe space vertical distance of the air inlet pipe is less than the air outlet pipe,x _piandy _pirespectively alternative point cloud data setsTTo middleiOf dotsxAxis coordinate sumyAxis coordinates;P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,P _ktarget corresponding trajectory position for current identificationxySlope of a tangent in a plane, saidxyThe plane is a projection plane of the point cloud data.

Step 34, based on the point cloud data setTTo the corresponding cross-sectional lineL _cPoint of minimum spatial sagP _T-minVdFrom an initial point cloud data setQExtracting points meeting the following constraint conditions, and constructing a local area data set of the current identification targetR：

x(P _T-minVd )－d _L ＜x _{Q_p} ＜ x(P _T-minVd ) ＋d _L

y(P _T-minVd )－d _L ＜y _{Q_p} ＜ y(P _T-minVd ) ＋d _L

Wherein the content of the first and second substances,x _{Q_p}、y _{Q_p}as a data setQPoint of (5)PIs/are as followsxShaft andythe coordinates of the axes are set to be,x(P _T-minVd )、y(P _T-minVd )is a pointP _T-minVdIs/are as followsxShaft andythe coordinates of the axes are set to be,d _Lthe distance condition is set based on the positioning error of the trajectory position and the mounting error of the target. In the present embodiment, it is preferred that,d _Lset to 5 m. The embodiment is real-time identification, that is, identification while scanning, and if the point cloud scanning is completed first and then identification is performed, a point-by-point circulation manner same as that in real time may be adopted, or circulation may be performed in steps, for example: the steps 31 to 34 are executed in a loop until the local area data sets of all the targets are completedRAnd (4) constructing. Other steps are similar and are not described in detail.

The advantages of this approach are: firstly, considering the scanning characteristics, defining a larger range and constructing an alternative point cloud data setT(ii) a Then, atTTo the corresponding cross-sectional lineL _cPoint of minimum spatial sagP _T-minVdAs the scanning reference point corresponding to the track time; finally, according toP _T-minVdData interception is carried out based on the track positioning error and the target positioning error, the intercepted data volume is smaller, and the calculation burden can be effectively reduced.

Step 4, aiming at the regional point cloud data set corresponding to the current identification targetRBased on the intensity of the reflectionI _pSorting the point cloud data from big to small, and extracting the top sorted point cloud datan _maxPoint cloud data of points constituting a seed point cloud data set corresponding to the targetB _p。

n _maxThe larger the point location, the more the point locations can be identified, the more accurate the matching in the central point extraction process is, but the larger the point location is, the more complicated the measurement in the tunnel isn _maxThe more likely noise points are introduced, increasing the difficulty of post-processing. In this embodiment, based on the size of the target and the scanning parameters, the lower limit of the number of scanning points corresponding to the target is estimated,n _maxthe value is 30.

And 5, further restricting the local range: seed point cloud data set corresponding to current identification targetB _pGo through the points it includesP _BiRespectively atB _pCorresponding regional point cloud data setRExtract and point ofP _BiPoint cloud data of points whose spatial distances meet set constraint conditions are formed corresponding to the pointsP _BiOf the image spot point cloud data setB _ti(ii) a The constraint condition is a shape constraint condition set based on the shape and size of the target to ensure each image spot point cloud data setB _tiThe shape of the region formed by the dots in (a) is a shape corresponding to the shape of the target body.

Wherein, the shape constraint condition set by the shape and the size of the target is as follows:

wherein the content of the first and second substances,D _bis composed ofP _BiAndRthe spatial distance of the points in (a),L _wlthe length of the sides of the body of the target, i.e. the length of the sides of the white matte paper in figure 2,W _bwis the width of the band of the target, i.e., the width of the black band in fig. 2.

The above-mentionedD _bThe calculation formula is as follows:

Step 6, reflection intensity constraint: point cloud data sets for each image spotB _tiRespectively judging the reflection intensity of the point cloud data of the points included in the point cloud data, extracting the point cloud data of the points with the reflection intensity higher than a set threshold value to form point cloud data corresponding to the pointsB _tiSubject candidate point cloud dataset ofB _wi。

The setting mode of the reflection intensity setting threshold is as follows:

wherein the content of the first and second substances,I _threshold-iis as followsiIndividual image spot point cloud data setB _tiThe determination threshold value of the reflection intensity of (2),max I{B _ti }is as followsiIndividual image spot point cloud data setB _tiThe maximum reflected intensity value of all the points in (c),I _p-minis as followsiIndividual image spot point cloud data setB _tiCorresponding local area data setRThe lowest reflected intensity value of.

Step 7, shape constraint: setting geometric constraint conditions based on the shape and size of the target, and aiming at the current identification target, respectively collecting the corresponding main body alternative point cloud data setsB _wiScreening according to the set geometric constraint condition, merging the targets corresponding to eachB _wiConstructing a main point cloud data set of the current identification target according to the point cloud data of the screened points meeting the geometric constraint conditionsB _w-final。

The geometric constraints are set based on the shape and size of the target as:

wherein the content of the first and second substances,max DeltaZ {p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiThe maximum height difference between all the point pairs;max D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiThe maximum spatial distance between all the point pairs;L _wlis the side length of the body of the target,W _bwis the width of the band of the target,L _wdis the diagonal length of the body of the target.

wherein the content of the first and second substances,x(p _wk )、y(p _wk )andz(p _wk )are respectively asB _wiTo middlekDotp _wkIs/are as followsxA shaft,yShaft andzaxis coordinates;x(p _wj )、y(p _wj )andz(p _wj )are respectively asB _wiTo middlejDotp _wjIs/are as followsxA shaft,yShaft andzaxis coordinates;D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe spatial distance of (a);DeltaZ {p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe difference in height of (a).

Thirdly, extracting a target center point:

the extraction method comprises three processes of point cloud data processing, vector data preparation and data matching; the concrete description is as follows:

the vector data preparation comprises the following steps:

a1, loading vector line frame data in accordance with the target measuring region size in the computer based on the target design sizeL _b(ii) a The target measuring region is a target main body region within a wire frame enclosed by the strips of the target, and the side length L of the region_wl-W_bw. The vector line frame data can be created by software such as Arcgis/CAD/Microstation.

A2, line frame data of vectorL _bWill convert to a single edgeUThe divided point data and the four sides are generated in total 4U+1 dot data, dot-to-dot spacing of: (L _wl-W _bw)/UConstructing a vector line frame data point setB _lbWherein, in the step (A),L _wlis the side length of the body of the target,W _bwthe width of the band of the target.

Wherein, U scans frequency H according to linear array of laser equipment_LAnd calculating to obtain:

the point cloud data processing comprises the following steps:

b1, extracting the main point cloud data set of the current identification targetB _w-finalEach scanning linear array based on the vehicle-mounted scanning systemB _w-finalFirst recording spot ofp _ti-fAnd the last recording spotp _ti-eConstructing an initial body contour point cloud dataset of the targetB _wl；

Specifically, a point cloud data set B is extracted_w-finalEach scanning line array inB _w-final1 st recording spot ofp _ti-fAnd the last 1 recording pointp _ti-eAs a set of contour point cloud data of target body regionsB _wlAs schematically shown in fig. 3, the first,p _t1-f、p _t1-ethe entry point and the exit point corresponding to the 1 st scanning line in the laser point cloud of the target body area are respectively shown, and the like.

B2 point cloud data set of initial body contour of the targetB _wlIn the method, data interpolation is carried out to construct a body contour point cloud data set of the targetB _wl-b(ii) a The subject contour point cloud datasetB _wl-bData point number and vector line frame data point setB _lbThe number of the data points in the data base is consistent; the method specifically comprises the following steps:

b21, calculating an initial subject contour point cloud data setB _wlWith vector line frame data point setB _lbDifference in number of data points in (1)delta_n：

delta_n=4U+1-n(B _wl)

b22, point cloud data set of initial body contourB _wlThe points in the target are sorted based on the scanning time to respectively obtain the entering targets of the scanning linear arraysSet of data points for a subject regionFAnd a set of data points away from the target body regionE；

Specifically, a contour point cloud data setB _wlTime sequencing is carried out on different scanning line arrays to obtain a scanning line data set { ScanL_t1, ScanL_t2, ScanL_t3 ,... ScanL_tnIn which ScanL_t1ScanL for the first scan line array point data set, entering the target body region_t2Expressed as ScanL_t1The next scan line array point data set into the target body region, and so on, ScanL_tnAnd n is the number of scanning linear arrays covered by the target body region.

For the scan line data set { ScanL_t1, ScanL_t2, ScanL_t3 ,... ScanL_tnTime sequencing is carried out on the point data in each scanning linear array in the sequence to obtain

、

、

、...、

Wherein

Is ScanL_t1Scanning the line array data for point data entering the target body region,

is ScanL_t1And scanning the point data which is separated from the target body area in the linear array data, and the like.

Set ScanL_t1, ScanL_t2, ScanL_t3 ,... ScanL_tnAll the point data entering the target main body area are sorted according to time to form a point data set F entering the target main body area,

set ScanL_t1, ScanL_t2, ScanL_t3 ,... ScanL_tnAll the point data leaving the target body area are sorted according to time to form a point data set E leaving the target body area,

。

b23 set of data points to be entered into target body regionFAnd a set of data points away from the target body regionERespectively introducing interpolation modules, forming point pairs by the points of the adjacent scanning linear arrays corresponding to the data point sets, and inserting interpolation points between the point pairs to obtain respectively corresponding data point setsFAnd a set of data pointsEThe set of interpolation points of (1).

The interpolation introducing module is used for forming a point pair by points of adjacent scanning linear arrays corresponding to the data point set, and specifically comprises the following steps:

wherein the content of the first and second substances,d _{delta_i}is a firstiThe strip scanning linear array and the adjacent secondi+1 scanning line array points to form the space distance between the point pairs;x(p _ScanL-t1 )、y(p _ScanL-t1 )andz(p _ScanL-t1 )for points at a point preceding the scanning timexA shaft,yShaft andzaxis coordinates;x(p _ScanL-t2 )、y(p _ScanL-t2 )andz(p _ScanL-t2 )for points subsequent to the scanning timexA shaft,yShaft andzaxis coordinates.

If introduced as a set of data points off the target body regionEThen, judgedelta_nIf the parity number of (1) is even, the front one with the largest spatial distance is takendelta_n/2Point pairs; if the number is odd, the front (B) with the largest spatial distance is takendelta_n- 1)/2Point pairs forming a data set of point pairs to be interpolatedC。

x _i=x(p _Cj1)+( x(p _Cj2)- x(p _Cj1))/2

y _i=y(p _Cj1)+( y(p _Cj2)- y(p _Cj1))/2

z _i=z(p _Cj1)+( z(p _Cj2)- z(p _Cj1))/2

wherein the content of the first and second substances,x _i、y _iandz _iare respectively the firstiOf an interpolation pointxA shaft,yShaft andzaxis coordinates;x(p _Cj1)、y(p _Cj1) Andz(p _Cj1) Respectively as the point pair data set to be interpolatedCTo middlejOf points at a point pair whose scanning time is earlierxA shaft,yShaft andzaxis coordinates;x(p _Cj2)、y(p _Cj2) Andz(p _Cj2) Respectively for point pairs to be interpolatedCTo middlejOf points at a later scanning timexA shaft,yShaft andzaxis coordinates.

The data matching comprises the following steps:

specifically, based on the SVD least square method, the vector wireframe data point set of the current identification target is calculatedB _lbWith subject contour point cloud data setB _wl-bOptimal estimation transformation matrix of (1):

wherein the content of the first and second substances,P _{lb_i}for vector line frame data point setsB _lbTo middleiCounting;P _{wl_i}point cloud data set for subject profileB _wl-bTo (1) aiThe point of the light beam is the point,icalculating a sequence number for traversal;Ra rotation matrix that is a match between the two sets of data,Ta translation matrix that is a match between the two sets of data,Nfor vector line frame data point setsB _lbAnd subject contour point cloud datasetB _wl-bThe number of dot data in (1).

C2 based on matching in step C1Matching results, to vector line frame data point setsB _lbRotating and translating to obtain the target vector line frame data point set after rotating and translatingB _lb-finalCalculatingB _lb-finalGeometric center ofB _P-centerAs the center point of the target.

In particular, based on the rotation matrix obtained in step C1RAnd translation matrixTSet vector line frame data pointsB _lbRotating and translating to obtain the target vector line frame data point set after rotating and translatingB _lb-finalAnd calculateB _lb-finalGeometric center ofB _P-center：

Wherein the content of the first and second substances,x(P _{wl-final_i})、y(P _{wl-final_i}) Andz(P _{wl-final_i}) Respectively a set of target vector line frame data pointsB _lb-finalTo middleiOf point dataxA shaft,yShaft andzthe coordinates of the axes are set to be,iin order to compute the sequence number for the traversal,Nset of line-frame data points for target vectorB _lb-finalThe number of data points in (a).

TABLE 1 target center extraction accuracy test

Table 1 shows the accuracy test of extracting the target center by using the visual method, the centroid method, and the target center point extraction method of the present invention, respectively, under the condition of different degrees of integrity of point cloud data. From the data in table 1:

the target contour point centroid method automatic extraction and the manual visual method manual extraction are adopted, and due to the sparsity of point cloud data, target boundary points obtained through scanning are sparse even under the condition that the point cloud data are complete, so that the accuracy of calculating the centroid through the contour points is directly influenced. When the target point cloud data is incomplete, namely point cloud data is obviously missing, due to insufficient geometrical characteristics, the extraction error of the target center is large, and the measurement precision is seriously influenced. Although the visual method can select the most appropriate point as the central point by empirical estimation and judgment, compared with the automatic extraction of the centroid method, the accuracy is improved, but the extraction efficiency is low. Therefore, no matter the centroid method is used for automatic extraction, or the visual method is used for manual extraction, obvious dependence is provided for the integrity of the target point cloud and the point cloud scanning density.

Compared with automatic extraction by a centroid method and manual extraction by a visual method, the method can better avoid target point cloud loss and obvious errors caused by the fact that an actual central point has no scanning point due to sparse point cloud density, obviously improves the precision, and is more obvious in precision improvement and accurate in measurement precision particularly aiming at the condition of target point cloud loss. As a key step in overall geometric correction of laser scanning data of tunnel mobile measurement, the problem of serious accumulated errors in a long-distance tunnel under the condition of multiple target arrangement environments can be obviously improved by improving the accuracy of a single target center point in a cm level, tests show that the extraction accuracy of each single-point target center is improved by 1cm averagely, the error of a control network constructed by 10 targets is reduced by more than 10cm overall, and the accuracy improvement of high-accuracy laser mapping data of a long-distance tunnel, particularly an extra-long tunnel with a mountain area exceeding 3km, and the realization of m-level accuracy crossing becomes possible.

It should be noted that the above-mentioned embodiments are only preferred embodiments and are not intended to limit the present invention. It should be noted that those skilled in the art can make various changes, substitutions and alterations herein without departing from the spirit of the invention and the scope of the appended claims.

Claims

1. A target identification method of tunnel vehicle-mounted scanning point cloud is characterized in that,

the target adopted by the identification method comprises a main body and strips distributed along the edge of the main body, wherein the main body is made of a material with the laser reflectivity higher than that of the side wall of the tunnel, the strips are made of a material with the laser reflectivity lower than that of the side wall of the tunnel, and the main body is in a regular shape with a central point;

the identification method comprises the following steps:

s1, setting a height condition based on the target setting height for the current identification target, dividing the scanning point cloud data, only reserving the point cloud data of the points meeting the set height condition, and constructing an initial point cloud data set corresponding to the current identification targetQ；

s3, in the initial point cloud data setQIn the method, based on a cross-sectional line corresponding to a current identification target, point cloud data of points meeting distance conditions are screened along the track direction, and an area point cloud data set corresponding to the target is constructedR(ii) a The distance condition is a screening distance set based on the scanning characteristic of the vehicle-mounted scanning system, the track positioning error and the target positioning error along the track direction so as to ensure that the distance condition is the screening distanceRThe system comprises point cloud data of corresponding targets;

s4, aiming at the area point cloud data set corresponding to the current identification targetRBased on the intensity of the reflectionI _pSorting the point cloud data from big to small, and extracting the top sorted point cloud datan _maxPoint cloud data of points constituting a seed point cloud data set corresponding to the targetB _p；

S5, aiming at the seed point cloud corresponding to the current identification targetData setB _pGo through the points it includesP _BiRespectively at theB _pCorresponding regional point cloud data setRExtract and point ofP _BiPoint cloud data of points whose spatial distances meet set constraint conditions are formed corresponding to the pointsP _BiOf the image spot point cloud data setB _tiThe constraint condition is a shape constraint condition set based on the shape size of the target so as to ensure each image spot point cloud data setB _tiThe shape of the region formed by the points in (1) is a shape adapted to the shape of the target body;

s6, aiming at each image spot point cloud data setB _tiRespectively judging the reflection intensity of the point cloud data of the points included in the point cloud data, extracting the point cloud data of the points with the reflection intensity higher than a set threshold value to form point cloud data corresponding to the pointsB _tiSubject candidate point cloud dataset ofB _wi；

2. The method as claimed in claim 1, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud is characterized in that,

in step S1, the height condition is set as follows:

H _min－σ≤Z _point ≤H _max＋σ

wherein the content of the first and second substances,Z _pointis the height coordinate of the point cloud data,H _maxthe highest height of the height tolerance range is set for the target,H _minis a targetSetting a minimum height of a height tolerance range, saidσA tolerance distance set based on the mounting error.

3. The method as claimed in claim 1, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud is characterized in that,

in step S2, extracting a trajectory time of the vehicle-mounted scanning system corresponding to the current identification target based on the set position of the target along the trajectory direction includes:

t _i+1 －t _i=D/V

4. The method for target identification of the cloud of scanning points carried on the vehicle in the tunnel according to any one of claims 1 to 3,

in step S2, the cross-sectional line corresponding to the trajectory position at the trajectory timeL _cThe point above, satisfies the following constraint:

wherein the content of the first and second substances,x、yis a transverse broken lineL _cGo up and down at each pointxyCoordinates in a plane, saidxyThe plane is a projection plane of the point cloud data;P _{x_c}for the current position of the track of the identified targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,P _{z_c}for the spatial elevation of the currently identified target's corresponding track location,zto correspond to a transverse lineL _cThe spatial elevation of (a);P _ktarget corresponding trajectory position for current identificationxySlope of tangent in plane.

5. The method as claimed in claim 1, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud is characterized in that,

in step S3, an initial point cloud data setQIn the method, based on a cross-sectional line corresponding to a current identification target, point cloud data of points meeting distance conditions are screened along the track direction, and an area point cloud data set corresponding to the target is constructedRThe method specifically comprises the following steps:

S33, traversing the alternative point cloud data set of the current identification targetTCalculating all points to corresponding cross-sectional linesL _cThe spatial sag of (c);

s34 based on the point cloud data setTTo the corresponding cross-sectional lineL _cPoint of minimum spatial sagP _T-minVdFrom an initial point cloud data setQExtracting points meeting the following constraint conditions, and constructing a local area data set of the current identification targetR：

x(P _T-minVd )－d _L ＜x _{Q_p} ＜ x(P _T-minVd ) ＋d _L

y(P _T-minVd )－d _L ＜y _{Q_p} ＜ y(P _T-minVd ) ＋d _L

Wherein the content of the first and second substances,x _{Q_p}、y _{Q_p}as a data setQPoint of (5)PIs/are as followsxShaft andythe coordinates of the axes are set to be,x(P _T-minVd )、y(P _T-minVd )is a pointP _T-minVdIsxShaft andythe coordinates of the axes are set to be,d _Lis a distance condition set based on the trajectory positioning error and the target positioning error.

6. The method as claimed in claim 5, wherein the target identification method comprises the steps of,

in step S32, the screening conditions set based on the tunnel lateral width are:

P _{x_c} －n _c ×W _c ＜x _p ＜ P _{x_c} ＋n _c ×W _c

P _{y_c} －n _c ×W _c ＜y _p ＜ P _{y_c} ＋n _c ×W _c

wherein the content of the first and second substances,x _p 、y _pfor an initial point cloud data setQIn (1)POf dotsxShaft andythe coordinates of the axes are set to be,P _{x_c}for the current identification of the corresponding track position of the targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,n _cas to the number of lanes of the vehicle,W _cis a single lane width.

7. The method as claimed in claim 5, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud comprises the steps of,

in step S33, each point is calculated to correspond to a cross-sectional line according to the following formulaL _cThe space vertical distance of (2):

wherein the content of the first and second substances,V _difor alternative point cloud data setsTTo middleiFrom point to transverse lineL _cThe space vertical distance of the air inlet pipe is less than the air outlet pipe,x _piandy _pirespectively alternative point cloud data setsTTo middleiOf dotsxAxis coordinate sumyAxis coordinates;P _{x_c}for the current position of the track of the identified targetxThe coordinates of the axes are set to be,P _{y_c}for the current identification of the corresponding track position of the targetyThe coordinates of the axes are set to be,P _ktargeting the current identification target to the corresponding track locationIs located inxySlope of a tangent in a plane, saidxyThe plane is a projection plane of the point cloud data.

8. The method as claimed in claim 1, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud is characterized in that,

in step S6, the setting method of the reflection intensity setting threshold value is:

wherein, the first and the second end of the pipe are connected with each other,I _threshold-iis as followsiIndividual image spot point cloud data setB _tiThe determination threshold value of the reflection intensity of (b),max I{B _ti }is as followsiIndividual image spot point cloud data setB _tiThe maximum reflected intensity value of all the points in (c),I _p-minis as followsiIndividual image spot point cloud data setB _tiCorresponding local area data setRThe lowest reflected intensity value of.

9. The method as claimed in claim 1, wherein the target body is square.

10. The method as claimed in claim 9, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud comprises the steps of,

in step S5, the shape constraint conditions set based on the shape size of the target are:

the above-mentionedD _bThe calculation formula is as follows:

wherein the content of the first and second substances,x(P _Bi )、y(P _Bi )andz(P _Bi )are respectively a pointP _BiIs/are as followsxA shaft,yShaft andzaxis coordinates;x(P _Rj )、y(P _Rj )andz(P _Rj )are respectively asRMiddle removingP _BiExternal firstjOf dotsxA shaft,yShaft andzaxis coordinates.

11. The method as claimed in claim 9, wherein the target identification method of the tunnel vehicle-mounted scanning point cloud comprises the steps of,

in step S7, geometric constraints are set based on the shape and size of the target as:

wherein the content of the first and second substances,max DeltaZ {p _wk ,p _wj }candidate point cloud data set for subjectB _wiThe maximum height difference between all the point pairs;max D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiThe maximum spatial distance between all the point pairs;L _wlis the side length of the body of the target,W _bwis the width of the band of the target,L _wdis a body of a targetThe length of the diagonal line of (a);

wherein the content of the first and second substances,x(p _wk )、y(p _wk )andz(p _wk )are respectively asB _wiTo middlekDotp _wkIsxA shaft,yShaft andzaxis coordinates;x(p _wj )、y (p _wj )andz(p _wj )are respectively asB _wiTo middlejDotp _wjIs/are as followsxA shaft,yShaft andzaxis coordinates;D{p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe spatial distance of (a);DeltaZ {p _wk ,p _wj }candidate point cloud dataset for a subjectB _wiTo middlekDotp _wkAnd a firstjDotp _wjThe difference in height of (a).

12. A target center point extraction method of tunnel vehicle-mounted scanning point clouds is used for extracting a center point of a target identified by the target identification method of the tunnel vehicle-mounted scanning point clouds according to any one of claims 9-11, and is characterized in that the extraction method comprises point cloud data processing, vector data preparation and data matching;

wherein the vector data preparation comprises the following steps:

a2, line frame data of vectorL _bWill convert to a single edgeUThe divided point data and the four sides are generated in total 4U+1 dot data, dot-to-dot spacing of: (L _wl-W _bw)/UConstructing a vector line frame data point setB _lbWherein, in the process,L _wlis the side length of the body of the target,W _bwthe width of the strip that is the target;

the point cloud data processing comprises the following steps:

b1, extracting a main body point cloud data set of the current identification targetB _w-finalEach scanning line array based on the vehicle-mounted scanning systemB _w-finalFirst recording spot ofp _ti-fAnd the last recording spotp _ti-eConstructing an initial body contour point cloud dataset of the targetB _wl；

B2 point cloud data set of initial body contour of the targetB _wlIn the method, data interpolation is carried out to construct a body contour point cloud data set of the targetB _wl-b(ii) a The subject contour point cloud datasetB _wl-bData point number and vector line frame data point setB _lbThe number of the data points in the data base is consistent;

the data matching comprises the following steps:

13. The method as claimed in claim 12, wherein the target center point of the point cloud of the tunnel scan point is extracted from the point cloud of the tunnel scan point,

in step a2, the U score is calculated as follows:

14. The method as claimed in claim 12, wherein the target center point of the point cloud of the tunnel scan point is extracted from the point cloud of the tunnel scan point,

step B2, generating a point cloud data set of the initial body contour of the targetB _wlIn the method, data interpolation is carried out to construct a body contour point cloud data set of the targetB _wl-bThe method specifically comprises the following steps:

b21, calculating an initial subject contour point cloud data setB _wlData point set with vector line frameB _lbDifference in number of data points indelta_n：

delta_n=4U+1-n(B _wl)

Wherein, the first and the second end of the pipe are connected with each other,n(B _wl) Point cloud data set for initial subject profileB _wlThe number of data points in (a);

b22, point cloud data set of initial body contourB _wlThe points in the table are sorted based on the scanning time to respectively obtain the entry of each scanning linear arraySet of data points for target body regionsFAnd a set of data points away from the target body regionE；

B23 set of data points to be entered into target body regionFAnd a set of data points away from the target body regionERespectively introducing interpolation modules, forming point pairs by the points of the adjacent scanning linear arrays corresponding to the data point sets, and inserting interpolation points between the point pairs to obtain respectively corresponding data point setsFAnd a set of data pointsEThe interpolation point set of (2);

15. The method as claimed in claim 14, wherein the target center point of the point cloud of the tunnel scan point is extracted from the point cloud of the tunnel scan point,

step B23, the interpolation module includes the following steps:

and B231, sequentially calculating the spatial distance between the point pairs formed by the points of the adjacent scanning line arrays in the corresponding data point set:

wherein, the first and the second end of the pipe are connected with each other,d _{delta_i}is as followsiThe strip scanning linear array and the adjacent secondi+1 scanning line array points to form the space distance between the point pairs;x(p _ScanL-t1 )、y(p _ScanL-t1 )andz(p _ScanL-t1 )for points at a point preceding the scanning timexA shaft,yShaft andzaxis coordinates;x(p _ScanL-t2 )、y(p _ScanL-t2 )andz(p _ScanL-t2 )for points subsequent to the scanning timexA shaft,yShaft andzaxis coordinates;

if introduced as a set of data points into the target body regionFThen, determinedelta_nIf the parity number of (1) is even, the front one with the largest spatial distance is takendelta_n/2Point pairs; if the number is odd, take the front with the largest space distance (delta_n+1)/2Point pairs forming a data set of point pairs to be interpolatedC；

If introduced as a set of data points off the target body regionEThen, judgedelta_nIf the parity number of (1) is even, the front one with the largest spatial distance is takendelta_n/2Point pairs; if the number is odd, the front (B) with the largest spatial distance is takendelta_n-1)/2Point pairs forming a data set of point pairs to be interpolatedC；

x _i=x(p _Cj1)+( x(p _Cj2)- x(p _Cj1))/2

y _i=y(p _Cj1)+( y(p _Cj2)- y(p _Cj1))/2

z _i=z(p _Cj1)+( z(p _Cj2)- z(p _Cj1))/2

wherein the content of the first and second substances,x _i、y _iandz _iare respectively the firstiOf an interpolation pointxA shaft,yShaft andzaxis coordinates;x(p _Cj1)、y(p _Cj1) Andz(p _Cj1) Respectively for point pairs to be interpolatedCTo middlejOf points at a point pair whose scanning time is earlierxA shaft,yShaft andzaxis coordinates;x(p _Cj2)、y(p _Cj2) Andz(p _Cj2) Respectively for point pairs to be interpolatedCTo middlejOf points at a later scanning timexA shaft,yShaft andzaxis coordinates.

16. The method as claimed in claim 12, wherein the target center point of the point cloud of the tunnel scan point is extracted from the point cloud of the tunnel scan point,

step C1, using SVD least square method to the vector line frame data point set of the current identification targetB _lbWith subject contour point cloud data setB _wl-bMatching, specifically comprising:

17. The method as claimed in claim 16, wherein the target center point of the point cloud of the tunnel vehicle-mounted scanning point is extracted,

in step C2, the vector wireframe data point set is matched based on the matching result in step C1B _lbRotating and translating to obtain the target vector wireframe data point set after rotating and translatingB _lb-finalCalculatingB _lb-finalGeometric center ofB _P-centerAs the center point of the target, specifically, the method includes:

based on the rotation matrix obtained in step C1RAnd translation matrixTSet vector line frame data pointsB _lbRotating and translating to obtain the target vector wireframe data point set after rotating and translatingB _lb-finalAnd calculateB _lb-finalGeometric center ofB _P-center：

Wherein the content of the first and second substances,x(P _{wl-final_i})、y(P _{wl-final_i}) Andz(P _{wl-final_i}) Respectively a target vector line frame data point setB _lb-finalTo middleiOf point dataxA shaft,yShaft andzthe coordinates of the axes are set to be,iin order to compute the sequence number for the traversal,Nset of line-frame data points for target vectorB _lb-finalNumber of data points in (c).