CN105354880B - A kind of sand blasting machine people's automatic path generation method based on line laser structured light - Google Patents

A kind of sand blasting machine people's automatic path generation method based on line laser structured light Download PDF

Info

Publication number
CN105354880B
CN105354880B CN201510666446.2A CN201510666446A CN105354880B CN 105354880 B CN105354880 B CN 105354880B CN 201510666446 A CN201510666446 A CN 201510666446A CN 105354880 B CN105354880 B CN 105354880B
Authority
CN
China
Prior art keywords
point
workpiece
sandblasting
point cloud
stl
Prior art date
Application number
CN201510666446.2A
Other languages
Chinese (zh)
Other versions
CN105354880A (en
Inventor
周波
吴宝举
孟正大
戴先中
韩明瑞
徐龙
Original Assignee
东南大学
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 东南大学 filed Critical 东南大学
Priority to CN201510666446.2A priority Critical patent/CN105354880B/en
Publication of CN105354880A publication Critical patent/CN105354880A/en
Application granted granted Critical
Publication of CN105354880B publication Critical patent/CN105354880B/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G06T17/30Polynomial surface description
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2200/00Indexing scheme for image data processing or generation, in general
    • G06T2200/04Indexing scheme for image data processing or generation, in general involving 3D image data

Abstract

The invention discloses sand blasting machine people's automatic path generation method based on line laser structured light, comprise the following steps:The first step:Obtain workpiece data information:The scanning sample point data information for treating sandblasting workpiece surface is obtained using 2D laser scanning methodses;Second step:Establish workpiece surface STL models:By point cloud pretreatment, sandblasting workpiece surface STL models are treated in foundation;3rd step:Divide curved surface:Extract and utilize the geometric properties and topological characteristic of workpiece surface STL models, curve surface of workpiece is divided into some planes without cavity;4th step:Generate sandblasting track:According to workpiece surface STL model construction feature frames, using cutting planes projecting method, generation sandblasting track.

Description

A kind of sand blasting machine people's automatic path generation method based on line laser structured light

Technical field

The present invention relates to a kind of sand blasting machine people automatic path generation method, it particularly relates to which a kind of be based on line laser Sand blasting machine people's automatic path generation method of scanning, particularly sand blasting machine people carry out the automatic of sandblasting to Unknown Model workpiece Path generating method.

Background technology

Large-scale steel structure weldment, because product is bordering on customed product, so the workpiece shapes processed every time are different from; Need to carry out surface pretreatment, clear up surface oxide layer.Combined at present using steel sand and lead sand, using 0.8MPa compressed air Steel sand is blown, throws to workpiece surface at a high speed, realizes the sandblasting work to surface.Because the kinetic energy of sand is higher, man-hour manually hand-held at present The injury to personnel is easily caused during sand rifle.Simultaneously because producing more dust in sandblasting procedures, worker is also easily caused Respiratory system sense of discomfort.The workpiece of processing is larger, handles a workpiece at present and takes around 2 classes of times of 2 people, is badly in need of using Equipment frees people from heavy dangerous work.

Common robot system is directed to standardization, repeated production by the way of teaching and off-line programing The processing of product.For some clients, belong to customization and nonstandard manufacture product, batch very little.Due to the know-how of operator With the limitation of job specification, client is also not intended to provide the operating path of robot by way of teaching or off-line programing.This Operating path planning for robot proposes high requirement, to meet the requirement of real-time of work, can not have again too many Preparation.From it is currently understood that situation in terms of, can be used for reference without similar Engineering Projects, be current robot application compared with For the problem in forward position, therefore there are more actual reference and demonstration meaning.

Current NI Vision Builder for Automated Inspection generally use visible light source, available for the image procossing for pre-entering fixed form Comparison, for non-standard, template can not be pre-entered, therefore, realizing the detection of workpiece according to this can not also realize, machining path Generation be even more do not know where to begin.

The content of the invention

Technical problem:The technical problems to be solved by the invention are:A kind of sand blasting machine based on line laser structured light is provided People's automatic path generation method, the generation method are modeled by line laser structured light to unknown work piece in sandblasting project, and then Automatically generating for path is carried out according to model.

Technical scheme:In order to solve the above technical problems, the technical scheme that uses of the embodiment of the present invention for:

A kind of sand blasting machine people's automatic path generation method based on line laser structured light, the generation method include following step Suddenly:

The first step:Obtain workpiece data information:Obtained using 2D laser scanning methodses and treat that the scanning of sandblasting workpiece surface is adopted Sampling point data message;

Second step:Establish workpiece surface STL models:By point cloud pretreatment, sandblasting workpiece surface STL models are treated in foundation;

3rd step:Divide curved surface:Extract and utilize the geometric properties and topological characteristic of workpiece surface STL models, by workpiece Curved surface is divided into some planes without cavity;

4th step:Generate sandblasting track:According to workpiece surface STL model construction feature frames, using cutting planes projection side Method, generation sandblasting track.

As preference, the described first step comprises the following steps:

Step 101):Arrange sand-blast shop:AGV dollies are placed in the middle part of workshop, by 3 industrial 2D laser scanners point It is not fixedly mounted on support, and 3 industrial 2D laser scanners are located at the top and two sides of AGV dollies respectively, will Treat that sandblasting workpiece is placed on AGV dollies, treat that sandblasting workpiece and AGV dollies are synchronized with the movement with guide rail;

Step 102):System calibrating is calibrated:By three-coordinates measuring machine to sensor pose and servomechanism installation zero point system Parameter is calibrated, and completes system calibrating;

Step 103):Workpiece is scanned, obtains data message:Driving guide rail is moved with known speed, while 3 2D laser are swept To retouch instrument sandblasting workpiece is treated with setpoint frequency and be scanned, sample, guarantee treats sandblasting workpiece completely by laser sampling plane, and The laser measurement values data message using the time as label that 3 2D laser scanners of record collect.

As preference, described second step comprises the following steps:

Step 201) point cloud is changed and splicing:The reference frame of sand blasting machine people is established, is swashed what step 103) obtained Light measurement value data message, passage time label and sensor measurement polar coordinate system and the relative pose relation of reference frame, Be converted to the cloud data under sand blasting machine people's reference frame, and by all Point-clouds Registrations to together, obtaining with reference to sitting The part model being made up of cloud data under mark system;

Step 202) point cloud filters:The part model obtained for step 201), processing is filtered, obtains accurate work Part point cloud model:

Step 203) point cloud smoothing processing:Using moving least squares surfaces approximating method, obtained for step 202) Accurate workpiece point cloud model, is smoothed, obtains smooth point cloud model;

Step 204) builds STL models:The smooth point cloud mould obtained using greedy projected triangle method to step 203) Type is handled, and generates the STL models of curve surface of workpiece.

As preference, described step 202) specifically includes:

The position and range of movement that step 2021) is installed according to AGV dollies, reject the sample region unrelated with treating sandblasting workpiece Domain, obtain the noisy workpiece point cloud model of sand blasting machine people's reference frame lower band;

Step 2022) rejects the noise spot in the workpiece point cloud model of step 2021) by median filtering algorithm:To step 2021) each data point p in the noisy workpiece point cloud model of band obtainedi, with piFor median window center, intermediate value window is taken The mouth length of side is w, then asks for intermediate value to w × w scan data in median windowWhenWhen,WhenWhen, piKeep constant;Wherein, δ represents median filter filtering threshold;So as to obtain robot reference coordinate system Under accurate workpiece point cloud model.

As preference, described step 203) specifically includes:

The point cloud model region gridding that step 2031) obtains step 202);

Step 2032) corrects the coordinate of each mesh point:The influence area size of mesh point is determined first, it is then determined that position In the point (x in mesh point influence areaI,yI,zI), I=1,2 ..., n, n represent the sum of point in mesh point influence area;With The fitting function according to formula (1) draws revised grid point coordinates afterwards:

Z=f (x, y)=pT(x, y) α (x, y) formula (1)

In formula, p (x, y)=[1, x, y]TRepresenting linear base, subscript T represents transposition, and α (x, y) represents coefficient vector, α (x, Y) according to optimization methodMinimum is obtained to try to achieve;

Step 2033) connects mesh point, forms fitting surface:Mesh point, shape are connected according to revised grid point coordinates Into fitting surface;

Step 2034) projects to a cloud on the fitting surface of step 2033), obtains smooth point cloud model.

As preference, described step 204) specifically includes:

Step 2041), which is appointed, takes a plane, using the point cloud that step 203) obtains as original point cloud, projects to the plane, Obtain and the one-to-one planar point cloud of point in original point cloud;

The planar point cloud that step 2042) obtains for step 2041), randomly select three points and be connected as a sample triangle Piece uses algorithm of region growing constantly to merge to triangular plate to expand surface boundary, forms complete triangle as initial surface Grid surface;

Step 2043) determines opening up between each initial three-dimensional point according to the annexation of point cloud projection, using back projection method Flutter connection, the surface mesh model that gained triangle gridding obtains for reconstruction;

Step 2044) is stored the surface mesh model that step 2043) generates according to stl file format standard, is obtained To the STL digital-to-analogue models of workpiece.

As preference, the 3rd described step comprises the following steps:

Step 301) areal model point cloud segmentation:The smooth point cloud model obtained to step 203), using based on adopting at random The point cloud segmentation algorithm of sample uniformity is split, and sets parted pattern as areal model, and segmentation obtains one group by workpiece surface The plane point set set of Different Plane point set composition and the fitting parameter of each plane point set;

Step 302) detects inner and outer boundary key point;

Step 303) simplifies boundary point:The crucial point sequence obtained to step 302) simplifies, and extraction angle point is as pass Key point, reject non-angle point element in sequence, the boundary point sequence set being simplified;

Step 304) carries out STL semantic labels:The boundary point sequence set of the simplification obtained according to step 303) is to step 2044) the STL digital-to-analogues model of generation carries out region division, and the inner and outer boundary in sand region to be painted is marked in STL digital-to-analogue models, real The semantic label of existing STL digital-to-analogue models.

As preference, described step 302) specifically includes:Each plane point set S that step 301) is obtainediAccording to Its corresponding plane fitting parameters is handled as follows:

Step 3021) point cloud projects:By plane point set SiIn the plane that is represented to fitting parameter of all spot projections in, obtain To the agonic plane point cloud model of the plane point set;

Packet boundary inside and outside step 3022) detection:The inner and outer boundary key point arrangement set of extraction point cloud:For each plane Cloud data in point cloud model, plane minimum outsourcing polygon vertex is chosen respectively as key point:It is more for key point Side shape external boundary summit situation, using planar process vector direction as top, with one group clockwise crucial point sequence form border key Point;It is polygon inner boundary summit situation for key point, using planar process vector direction as top, with one group of key point counterclockwise Sequence forms border key point.

As preference, the 4th described step comprises the following steps:

Step 401) extraction model information:The STL models generated according to the 3rd step, extract triangular plate grid information;

Step 402) builds bounding box:Structure one can make the cuboid that the STL grid surfaces of workpiece are included For bounding box, and create corresponding bounding box coordinate system, the origin of bounding box coordinate system bounding box cuboid geometric center, Each reference axis is parallel with each side of cuboid;

Step 403) chooses sandblasting feature frame:STL grid surfaces are carried out to perpendicular to the face of each axle positive direction of bounding box Projection, calculates the area of each projection, and chooses feature frame of the maximum face of projected area as sandblasting;

Step 404) generates sandblasting characteristic curve:If sandblasting covering radius is r, it is by the spacing selection of each stroke in sandblasting path R/2, as the distance between adjacent feature line, so as to generate the sandblasting characteristic curve being parallel to each other on characteristic face;

Step 405) generation sandblasting track:Projected with characteristic curve along normal direction in workpiece surface, the curve intercepted As sandblasting stroke, adjacent sandblasting stroke is attached, obtains final sandblasting track.

Beneficial effect:Compared with prior art, sand blasting machine people's automatic path generation method of the embodiment of the present invention passes through Inexpensive 2D laser sensors are scanned online to workpiece surface, automatic raw on this basis to establish accurate 3D digital-to-analogues Into the movement locus of robot, traditional artificial teaching programming mode has been abandoned.The method is relative to traditional reverse-engineering cost Low and speed block, can be widely used for the surface sand-blasting process of large-scale steel structure weldment (such as large-scale engineering machinery), particularly Small lot, multi items, the surface sand-blasting process of non-standard product.It can apply simultaneously and the processing of such as oil derrick, metal surface The fields such as processing, have broad application prospects and economic benefit.

Brief description of the drawings

Fig. 1 is the FB(flow block) of the embodiment of the present invention;

Fig. 2 is the layout of sand-blast shop in the embodiment of the present invention;

Fig. 3 is the FB(flow block) of the first step in the embodiment of the present invention;

Fig. 4 is the FB(flow block) of second step in the embodiment of the present invention;

Fig. 5 is greedy sciagraphy schematic diagram in the embodiment of the present invention;

Fig. 6 is the FB(flow block) of the 3rd step in the embodiment of the present invention;

Fig. 7 a are the organigram of bounding box in the embodiment of the present invention;

Fig. 7 b are intermediate cam mesh flake schematic diagram of the embodiment of the present invention;

Fig. 8 is feature frame and characteristic curve schematic diagram in the embodiment of the present invention;

Fig. 9 is the FB(flow block) of the 4th step in the embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawings, the technical scheme of the embodiment of the present invention is described further.

A kind of as shown in figure 1, sand blasting machine people automatic path generation side based on line laser structured light of the embodiment of the present invention Method, the generation method comprise the following steps:

The first step:Obtain workpiece data information:Obtained using 2D laser scanning methodses and treat that the scanning of sandblasting workpiece surface is adopted Sampling point data message;

Second step:Establish workpiece surface STL models:By point cloud pretreatment, sandblasting workpiece surface STL models are treated in foundation;

3rd step:Divide curved surface:Extract and utilize the geometric properties and topological characteristic of workpiece surface STL models, by workpiece Curved surface is divided into some planes without cavity;

4th step:Generate sandblasting track:According to workpiece surface STL model construction feature frames, using cutting planes projection side Method, generation sandblasting track.

As shown in figure 3, the described first step comprises the following steps:

Step 101):Arrange sand-blast shop:By automatic guided vehicle (in text referred to as:AGV) dolly is placed in the middle part of workshop, will 3 industrial 2D laser scanners are respectively and fixedly installed on support, and 3 industrial 2D laser scanners are located at AGV respectively The top of dolly and two sides, will treat that sandblasting workpiece is placed on AGV dollies, treat that sandblasting workpiece and AGV dollies are synchronous with guide rail Motion;As shown in Figure 2, it is contemplated that 2D laser scanning surfaces can only obtain two-dimensional distance information, in order to obtain the complete sample of workpiece letter Breath, 3 2D laser scanners are respectively placed in the top and side of AGV dollies, and ensure that 3 2D laser scannings regions are coplanar, Sandblasting workpiece is placed on AGV dollies and moved with guide rail with known speed, while 3 2D laser scanners are swept with setpoint frequency Retouch sampling and record data.

Step 102):System calibrating is calibrated:By three-coordinates measuring machine to sensor pose and servomechanism installation zero point system Parameter is calibrated, and completes system calibrating;

Step 103):Workpiece is scanned, obtains data message:Driving guide rail is moved with known speed, while 3 2D laser are swept To retouch instrument sandblasting workpiece is treated with setpoint frequency and be scanned, sample, guarantee treats sandblasting workpiece completely by laser sampling plane, and The laser measurement values data message using the time as label that 3 2D laser scanners of record collect.

As shown in figure 4, described second step comprises the following steps:

Step 201) point cloud is changed and splicing:The reference frame of sand blasting machine people is established, is swashed what step 103) obtained Light measurement value data message, passage time label and sensor measurement polar coordinate system and the relative pose relation of reference frame, Be converted to the cloud data under sand blasting machine people's reference frame, and by all Point-clouds Registrations to together, obtaining with reference to sitting The part model being made up of cloud data under mark system;

Step 202) point cloud filters:The part model obtained for step 201), processing is filtered, obtains accurate work Part point cloud model;Step 202) specifically includes:

The position and range of movement that step 2021) is installed according to AGV dollies, reject the sample region unrelated with treating sandblasting workpiece Domain, obtain the noisy workpiece point cloud model of sand blasting machine people's reference frame lower band;

Step 2022) rejects the noise spot in the workpiece point cloud model of step 2021) by median filtering algorithm:To step 2021) each data point p in the noisy workpiece point cloud model of band obtainedi, with piFor median window center, intermediate value window is taken The mouth length of side is w, then asks for intermediate value to w × w scan data in median windowWhenWhen,WhenWhen, piKeep constant;Wherein, δ represents median filter filtering threshold;So as to obtain robot reference coordinate system Under accurate workpiece point cloud model.

Step 203) point cloud smoothing processing:The accurate point cloud model obtained for step 202), due to work in sampling process Certain do not know be present in the uncertainty and the uncertainty of sensor that part moves with guide rail, obtained curved surface point cloud model Property, it is exactly that point cloud is not smooth enough that it, which is showed,.On the other hand, the present embodiment uses the curved surface fitting method pair based on Moving Least Point cloud carries out smoothly, to remove the scanning error problem that guide rail movement, sensor accuracy etc. are brought, obtaining accurate curve surface of workpiece Point cloud model.

Using moving least squares surfaces approximating method, the accurate workpiece point cloud model obtained for step 202), enter Row smoothing processing, obtain smooth point cloud model.Step 203) specifically includes:

The point cloud model region gridding that step 2031) obtains step 202);

Step 2032) corrects the coordinate of each mesh point:The influence area size of mesh point is determined first, it is then determined that position In the point (x in mesh point influence areaI,yI,zI), I=1,2 ..., n, n represent the sum of point in mesh point influence area;With The fitting function according to formula (1) draws revised grid point coordinates afterwards:

Z=f (x, y)=pT(x, y) α (x, y) formula (1)

In formula, p (x, y)=[1, x, y]TRepresenting linear base, subscript T represents transposition, and α (x, y) represents coefficient vector, α (x, Y) according to optimization methodMinimum is obtained to try to achieve;

Step 2033) connects mesh point, forms fitting surface:Mesh point, shape are connected according to revised grid point coordinates Into fitting surface;

Step 2034) projects to a cloud on the fitting surface of step 2033), obtains smooth point cloud model.

Step 204) builds STL models:The smooth point cloud mould obtained using greedy projected triangle method to step 203) Type is handled, and generates the STL models of curve surface of workpiece.As shown in figure 5, step 204) specifically includes:

Step 2041), which is appointed, takes a plane, using the point cloud that step 203) obtains as original point cloud, projects to the plane, Obtain and the one-to-one planar point cloud of point in original point cloud;In Fig. 5, original point cloud p1、p2、p3、p4、p5、p6And p7, it is corresponding Planar point cloud be p '1、p′2、p′3、p′4、p′5、p′6With p '7

The planar point cloud that step 2042) obtains for step 2041), randomly select three points and be connected as a sample triangle Piece uses algorithm of region growing constantly to merge to triangular plate to expand surface boundary, forms complete triangle as initial surface Grid surface;

Step 2043) determines opening up between each initial three-dimensional point according to the annexation of point cloud projection, using back projection method Flutter connection, the surface mesh model that gained triangle gridding obtains for reconstruction;Triangle gridding is that space original point first is projected into one Individual plane, plane projection triangular mesh is obtained, back projection is then carried out according to the one-to-one relationship of projection, obtain space original The triangle gridding of initial point;

Step 2044) is stored the surface mesh model that step 2043) generates according to stl file format standard, is obtained To the STL digital-to-analogue models of workpiece.

As shown in fig. 6, the 3rd described step comprises the following steps:

Step 301) areal model point cloud segmentation:The smooth point cloud model obtained to step 203), using based on adopting at random The point cloud segmentation algorithm of sample uniformity is split, set parted pattern as areal model, model intra-office point minimal amount with And algorithm tolerance obtains one group of plane point set set being made up of workpiece surface Different Plane point set apart from absolute value, segmentation And the fitting parameter of each plane point set.As preference, the minimal amount of model intra-office point is 20, algorithm tolerance away from It is 1cm from absolute value.

The smooth point cloud model obtained for step 203), form are as follows:

Ax+by+cz+d=0

Wherein, (x, y, z) represents the three-dimensional coordinate of point, and a, b, c, d are plane fitting parameters;Using based on stochastical sampling one The point cloud segmentation algorithm of cause property is split, and sets parted pattern as areal model, and algorithm operational factor sets as follows:Model office Interior minimal amount is 20, and algorithm tolerance is 1cm apart from absolute value.Final segmentation obtains some fit Plane point set Si(i =1,2 ..., N) and corresponding fitted model parameters ai,bi,ci,di

It is prior art to sample consistency algorithm.The described point cloud segmentation algorithm based on stochastical sampling uniformity is divided The process cut is:

If 3011) concentrate random selection to do from the cloud data of input, and Calculation Plane model parameter.Will be current maximum Interior number is set to dm=0;

3012) distance threshold is set to the cloud data collection, the institute of test data concentration a little, puts down if point arrives fitting one by one The distance in face is less than threshold value, then the point is classified as into interior point, is otherwise exterior point;

3013) the number d1 of this interior point of statistics, judges whether the number threshold value d0 more than setting, if it is, performing step 3014), step 3015) is performed if it is not, directly redirecting;

3014) with interior point again fit Plane, exported as model, all interior points of storage are used as segmentation result, segmentation knot Beam;

3015) by d1 compared with dm, if d1>Dm then stores the areal model of this step fitting, and sets dm=d1;

3015) contrast with current maximum intra-office point number, if greater than the current maximum intra-office point number of then substitution, and deposit Store up current model coefficient, return to step 3011), until segmentation terminates.

Step 302) detects inner and outer boundary key point:Step 302) specifically includes:Each plane that step 301) is obtained Point set SiIt is handled as follows according to its corresponding plane fitting parameters:

Step 3021) point cloud projects:By plane point set SiIn the plane that is represented to fitting parameter of all spot projections in, obtain To the agonic plane point cloud model of the plane point set;

Packet boundary inside and outside step 3022) detection:The inner and outer boundary key point arrangement set of extraction point cloud:For each plane Cloud data in point cloud model, plane minimum outsourcing polygon vertex is chosen respectively as key point:It is more for key point Side shape external boundary summit situation, using planar process vector direction as top, with one group clockwise crucial point sequence form border key Point;It is polygon inner boundary summit situation for key point, using planar process vector direction as top, with one group of key point counterclockwise Sequence forms border key point.Whether key point is polygon angle point, collinearly obtained by the point and front and rear each 1 flanking sequence point .

Step 303) simplifies boundary point:The crucial point sequence obtained to step 302) simplifies, and extraction angle point is as pass Key point, reject non-angle point element in sequence, the boundary point sequence set being simplified;

Step 304) carries out STL semantic labels:The boundary point sequence set of the simplification obtained according to step 303) is to step 2044) the STL digital-to-analogues model of generation carries out region division, and the inner and outer boundary in sand region to be painted is marked in STL digital-to-analogue models, real The semantic label of existing STL digital-to-analogue models.

As shown in figure 9, the 4th step comprises the following steps:

Step 401) extraction model information:The STL models generated according to the 3rd step, extract triangular plate grid information.

Step 402) builds bounding box:Structure one can make the cuboid that the STL grid surfaces of workpiece are included For bounding box, and create corresponding bounding box coordinate system, the origin of bounding box coordinate system bounding box cuboid geometric center, Each reference axis is parallel with each side of cuboid.

The construction process of bounding box as shown in Figure 7a, by the origin O of bounding box coordinate systemBOXIt is set in bounding box cuboid Geometric center, the reference axis of bounding box coordinate system is parallel with each side of cuboid, and sets and axle xBOXThe parallel length of side is A, with axle yBOXThe parallel length of side is B, with axle zBOXThe parallel length of side is C.As long as obtain point OBOX, unit vector xBOX、yBOX、zBOX, and The length of side A, B, C, so that it may determine the pose and size of bounding box.

As shown in Figure 7b, if N number of triangle gridding piece is shared, wherein three summits of i-th of triangle gridding piece are respectively pi、 qi、ri, area Ai, triangle barycenter is Oi(i=1,2 ..., N).If the gross area of N number of triangle gridding piece is AH, then bounding box The geometric center point O of cuboidBOXCoordinate value in workpiece coordinate system is sat for each triangle gridding piece barycenter in workpiece coordinate system The average μ of the Area-weighted of scale value, can be drawn by following formula:

If the covariance matrix on each summit of triangle gridding piece is C, 3 × 3 Matrix C jth rows, the Elements C of kth rowjkBy following formula Try to achieve:

Wherein,μ= (μ123)T

Because covariance matrix C is real symmetric matrix, the characteristic vector corresponding to different characteristic value is orthogonal, obtains square Battle array tri- mutually orthogonal characteristic vectors of C, the direction vector of bounding box coordinate system reference axis is obtained after unitization.Then to institute There is triangle gridding piece summit to be traveled through, obtain projection maximum and minimum value of each summit in three reference axis, be respectively xmax、xmin、ymax、ymin、zmax、zmin, so that it is determined that the length of side of bounding box:A=xmax-xmin, b=ymax-ymin, c=zmax- zmin

Step 403) chooses sandblasting feature frame:STL grid surfaces are carried out to perpendicular to the face of each axle positive direction of bounding box Projection, calculates the area of each projection, and chooses feature frame of the maximum face of projected area as sandblasting.In bounding box coordinate In system, if projected area of the sprayed surface along x-axis positive direction is Ax+, the projected area along y-axis positive direction is Ay+, it is square along z-axis To projected area be Az+。Ax+、Ay+、Az+Middle numerical value the maximum, select spraying of the bounding box surface as system corresponding to it Feature frame.

Step 404) generates sandblasting characteristic curve:As shown in figure 8, sandblasting covering radius is set as r, by each stroke in sandblasting path Spacing selection is r/2, as the distance between adjacent feature line, so as to generate the sandblasting feature being parallel to each other on characteristic face Line.

Step 405) generation sandblasting track:Projected with characteristic curve along normal direction in workpiece surface, the curve intercepted As sandblasting stroke, adjacent sandblasting stroke is attached, obtains final sandblasting track.

2D laser scanners can gather two-dimensional distance information, and insensitive to light intensity, be adapted to enter depth information Row measurement, but problem is become complicated because it can not obtain 3 D stereo depth information.In view of sandblasting engineering only needs to obtain The characteristics of part model, the workpiece advanced by special exercise rule is scanned by using multiple fixed 2D laser scanners, The 3-dimensional point cloud model of workpiece can be calculated, and then can solve sand blasting machine people's automatic path generation method.

Sand blasting machine people's automatic path generation method based on line laser structured light of the embodiment of the present invention, passes through 3 first It is fixed on the workpiece that the industrial 2D laser scanners of ad-hoc location are opposite on AGV dollies and is scanned and realizes line laser structured light Data acquisition;Secondly workpiece surface modeling is carried out, is filtered by a cloud conversion with splicing, point cloud, point cloud smoothing processing, STL numbers The steps such as mould structure, it is final to obtain STL digital-to-analogues;Then extraction model geometry and topological characteristic, refer mainly to plane characteristic and hole Feature, and then curved surface is divided into by some planes without cavity according to the characteristic information extracted;Finally according to STL model structures Feature frame is built, automatic sand blasting Track Pick-up is realized using cutting planes projecting method.The sand blasting machine people of the embodiment of the present invention is certainly Dynamic path generating method is scanned online by inexpensive 2D laser sensors to workpiece surface, to establish accurate 3D Digital-to-analogue, on this basis by the way of online programming automatic plane-generating robot optimization movement locus, abandoned tradition Artificial teaching programming mode, surface tracks are automatically generated in a manner of online programming, and reach all standing processing purpose, letter It is single efficient.

The method of the embodiment of the present invention realizes that mainly including 2D line laser structured lights carries out workpiece modeling and based on built workpiece The automatic sand blasting coordinates measurement two large divisions of model, its advantage major embodiment is in the following areas:

First, the first step and second step can quickly generate workpiece STL models by the 2D line laser structured lights of low cost.Compare In the 3D laser scanners that reverse modeling is commonly used, the 2D lines scanning laser that the embodiment of the present invention uses is ensureing specific nozzleman On the premise of skill precision, hardware cost can be greatly reduced, and the data handling procedure of 2D line laser structured lights is made than 3D laser Type process is more quick, the real-time online modeling of workpiece can be realized, so as to substantially increase the applicable surface of this method and easy-to-use Property.

Secondly, the embodiment of the present invention proposed by the method in the STL auto-building model sandblastings path generated in real time, Compared to traditional artificial teaching operation mode, lengthy and tedious artificial teaching process can be avoided, substantially reduces the overall production time, And improve the automated job level and production efficiency of sandblasting industry.And compared with the off-line programing mode of routine, the present invention Realize without workpiece CAD model can solve the defects of off-line programing depends on accurate CAD digital-to-analogues (for most of domestic processing It is difficult often to obtain accurate digital-to-analogue in advance for process), and can realize and be carried out online according to real time laser scan data The optimization process such as adjustment and automatic deviation correction, wide adaptation range, flexibility is strong, is particularly suitable for small lot, multi items, non-standard work The occasion of part.

On the whole, the embodiment of the present invention passes through the online sandblasting coordinates measurement side based on 2D low cost line scanning lasers Method, cost is cheap, simple to operate efficient, without carrying out cumbersome teaching process, overcomes Traditional Man teaching mode and produces week The defects of phase length, very flexible, while also compensate for conventional offline and program the deficiency for needing the original CAD diagram paper of workpiece, it is especially suitable Close small lot, multi items, the processing of nonstandard product and manufacturing process.

Claims (7)

  1. A kind of 1. sand blasting machine people's automatic path generation method based on line laser structured light, it is characterised in that:The generation method bag Include following steps:
    The first step:Obtain workpiece data information:The scanning sample point for treating sandblasting workpiece surface is obtained using 2D laser scanning methodses Data message;The described first step comprises the following steps:
    Step 101):Arrange sand-blast shop:AGV dollies are placed in the middle part of workshop, 3 industrial 2D laser scanners are solid respectively It is fixed rack-mount, and 3 industrial 2D laser scanners are located at the top and two sides of AGV dollies respectively, will be to be painted Sand workpiece is placed on AGV dollies, treats that sandblasting workpiece and AGV dollies are synchronized with the movement with guide rail;
    Step 102):System calibrating is calibrated:By three-coordinates measuring machine to sensor pose and servomechanism installation zero point systematic parameter Calibrated, complete system calibrating;
    Step 103):Workpiece is scanned, obtains data message:Driving guide rail is moved with known speed, while 3 2D laser scanners Sandblasting workpiece is treated with setpoint frequency to be scanned, sample, guarantee treats sandblasting workpiece completely by laser sampling plane, and record The laser measurement values data message using the time as label that 3 2D laser scanners collect;
    Second step:Establish workpiece surface STL models:By point cloud pretreatment, sandblasting workpiece surface STL models are treated in foundation;It is described Second step comprise the following steps:
    Step 201) point cloud is changed and splicing:The reference frame of sand blasting machine people is established, the Laser Measuring that step 103) is obtained Magnitude data information, passage time label and sensor measurement polar coordinate system and the relative pose relation of reference frame, conversion For the cloud data under sand blasting machine people's reference frame, and all Point-clouds Registrations are obtained into reference frame to together Under the part model being made up of cloud data;
    Step 202) point cloud filters:The part model obtained for step 201), processing is filtered, obtains accurate workpiece point Cloud model:
    Step 203) point cloud smoothing processing:Using moving least squares surfaces approximating method, obtained for step 202) accurate Workpiece point cloud model, be smoothed, obtain smooth point cloud model;
    Step 204) builds STL models:The smooth point cloud model obtained using greedy projected triangle method to step 203) is entered Row processing, generate the STL models of curve surface of workpiece;
    3rd step:Divide curved surface:Extract and utilize the geometric properties and topological characteristic of workpiece surface STL models, by curve surface of workpiece It is divided into some planes without cavity;
    4th step:Generate sandblasting track:It is raw using cutting planes projecting method according to workpiece surface STL model construction feature frames Into sandblasting track.
  2. 2. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 1, its feature exists In:Described step 202) specifically includes:
    The position and range of movement that step 2021) is installed according to AGV dollies, the sample area unrelated with treating sandblasting workpiece is rejected, Obtain the noisy workpiece point cloud model of sand blasting machine people's reference frame lower band;
    Step 2022) rejects the noise spot in the workpiece point cloud model of step 2021) by median filtering algorithm:To step 2021) each data point p in the noisy workpiece point cloud model of band obtainedi, with piFor median window center, intermediate value window is taken The mouth length of side is w, then asks for intermediate value to w × w scan data in median windowWhenWhen,WhenWhen, piKeep constant;Wherein, δ represents median filter filtering threshold;So as to obtain robot reference coordinate system Under accurate workpiece point cloud model.
  3. 3. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 1, its feature exists In:Described step 203) specifically includes:
    The point cloud model region gridding that step 2031) obtains step 202);
    Step 2032) corrects the coordinate of each mesh point:The influence area size of mesh point is determined first, it is then determined that being located at net Point (x in lattice point influence areaI,yI,zI), I=1,2 ..., n, n represent the sum of point in mesh point influence area;Subsequent root Revised grid point coordinates is drawn according to the fitting function shown in formula (1):
    Z=f (x, y)=pT(x, y) α (x, y) formula (1)
    In formula, p (x, y)=[1, x, y]TLinear base is represented, subscript T represents transposition, and α (x, y) represents coefficient vector, α (x, y) root According to optimization methodMinimum is obtained to try to achieve;
    Step 2033) connects mesh point, forms fitting surface:Mesh point is connected according to revised grid point coordinates, is formed and intended Close curved surface;
    Step 2034) projects to a cloud on the fitting surface of step 2033), obtains smooth point cloud model.
  4. 4. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 1, its feature exists In:Described step 204) specifically includes:
    Step 2041), which is appointed, takes a plane, using the point cloud that step 203) obtains as original point cloud, projects to the plane, obtains With the one-to-one planar point cloud of point in original point cloud;
    The planar point cloud that step 2042) obtains for step 2041), randomly select three points and be connected as a sample triangular plate work For initial surface, use algorithm of region growing constantly to merge to triangular plate to expand surface boundary, form complete triangle gridding Curved surface;
    Step 2043) determines that the topology between each initial three-dimensional point connects using back projection method according to the annexation of point cloud projection Connect, the surface mesh model that gained triangle gridding obtains for reconstruction;
    Step 2044) is stored the surface mesh model that step 2043) generates according to stl file format standard, obtains work The STL digital-to-analogue models of part.
  5. 5. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 4, its feature exists In:The 3rd described step comprises the following steps:
    Step 301) areal model point cloud segmentation:The smooth point cloud model obtained to step 203), using based on stochastical sampling one The point cloud segmentation algorithm of cause property is split, and sets parted pattern as areal model, and segmentation obtains one group by workpiece surface difference The plane point set set of plane point set composition and the fitting parameter of each plane point set;
    Step 302) detects inner and outer boundary key point;
    Step 303) simplifies boundary point:The crucial point sequence obtained to step 302) simplifies, extraction angle point as key point, Reject non-angle point element in sequence, the boundary point sequence set being simplified;
    Step 304) carries out STL semantic labels:The boundary point sequence set of the simplification obtained according to step 303) is to step 2044) The STL digital-to-analogues model of generation carries out region division, and the inner and outer boundary in sand region to be painted is marked in STL digital-to-analogue models, realizes STL The semantic label of digital-to-analogue model.
  6. 6. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 5, its feature exists In:Described step 302) specifically includes:Each plane point set S that step 301) is obtainediAccording to its corresponding plane fitting Parameter is handled as follows:
    Step 3021) point cloud projects:By plane point set SiIn the plane that is represented to fitting parameter of all spot projections in, be somebody's turn to do The agonic plane point cloud model of plane point set;
    Packet boundary inside and outside step 3022) detection:The inner and outer boundary key point arrangement set of extraction point cloud:For each planar point cloud Cloud data in model, plane minimum outsourcing polygon vertex is chosen respectively as key point:It is polygon for key point External boundary summit situation, using planar process vector direction as top, with one group clockwise crucial point sequence form border key point;It is right In key point be polygon inner boundary summit situation, using planar process vector direction as top, with one group of crucial point sequence counterclockwise Form border key point.
  7. 7. according to sand blasting machine people's automatic path generation method based on line laser structured light described in claim 1, its feature exists In:The 4th described step comprises the following steps:
    Step 401) extraction model information:The STL models generated according to the 3rd step, extract triangular plate grid information;
    Step 402) builds bounding box:Structure one can be using the cuboid that the STL grid surfaces of workpiece are included as bag Box is enclosed, and creates corresponding bounding box coordinate system, the origin of bounding box coordinate system is in the geometric center of bounding box cuboid, each seat Parameter is parallel with each side of cuboid;
    Step 403) chooses sandblasting feature frame:STL grid surfaces are projected to perpendicular to the face of each axle positive direction of bounding box, The area of each projection is calculated, and chooses feature frame of the maximum face of projected area as sandblasting;
    Step 404) generates sandblasting characteristic curve:It is r/2 by the spacing selection of each stroke in sandblasting path if sandblasting covering radius is r, As the distance between adjacent feature line, so as to generate the sandblasting characteristic curve being parallel to each other on characteristic face;
    Step 405) generation sandblasting track:Projected with characteristic curve along normal direction in workpiece surface, the curve intercepted is Sandblasting stroke, adjacent sandblasting stroke is attached, obtains final sandblasting track.
CN201510666446.2A 2015-10-15 2015-10-15 A kind of sand blasting machine people's automatic path generation method based on line laser structured light CN105354880B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510666446.2A CN105354880B (en) 2015-10-15 2015-10-15 A kind of sand blasting machine people's automatic path generation method based on line laser structured light

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510666446.2A CN105354880B (en) 2015-10-15 2015-10-15 A kind of sand blasting machine people's automatic path generation method based on line laser structured light

Publications (2)

Publication Number Publication Date
CN105354880A CN105354880A (en) 2016-02-24
CN105354880B true CN105354880B (en) 2018-02-06

Family

ID=55330847

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510666446.2A CN105354880B (en) 2015-10-15 2015-10-15 A kind of sand blasting machine people's automatic path generation method based on line laser structured light

Country Status (1)

Country Link
CN (1) CN105354880B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109405767A (en) * 2018-12-25 2019-03-01 威海筑丰五金有限公司 Method is determined based on the cast member dead head cutting track of laser profile tester
CN110434671A (en) * 2019-07-25 2019-11-12 王东 A kind of cast member surface machining track calibration method based on pattern measurement

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106327561A (en) * 2016-08-31 2017-01-11 武汉湾流科技股份有限公司 Intelligent spraying method and system based on machine vision technology
CN106500617A (en) * 2016-09-30 2017-03-15 长春理工大学 Complex-curved trajectory planning and off-line programing method
CN107066654B (en) * 2016-12-15 2020-03-10 广东工业大学 Box surface containing track control point picking method and device facing point cloud model
CN106525116A (en) * 2016-12-26 2017-03-22 海克斯康测量技术(青岛)有限公司 Real-time online measurement device and method thereof
CN106960470A (en) * 2017-04-05 2017-07-18 未来科技(襄阳)有限公司 Three-dimensional point cloud curve reestablishing method and device
CN107127121A (en) * 2017-04-13 2017-09-05 天长市金陵电子有限责任公司 One kind finishing electrostatic powder spraying method
CN108227696A (en) * 2017-09-04 2018-06-29 河南森源电气股份有限公司 The mathematical model discrimination method and system of a kind of AGV mobile robots
CN107803598A (en) * 2017-09-28 2018-03-16 惠州市洛玛科技有限公司 Laser cutting machine and its cutting method
CN107767457A (en) * 2017-10-09 2018-03-06 东南大学 A kind of STL digital-to-analogue generation methods quickly rebuild based on a cloud
CN107728619B (en) * 2017-10-16 2020-06-23 邦鼓思电子科技(上海)有限公司 Planning method for robot teaching path
CN107999967B (en) * 2017-11-30 2019-05-31 华中科技大学 A kind of large-scale three dimensional piece surface parallel laser lithography method and apparatus
CN108153244B (en) * 2017-12-31 2020-05-12 芜湖哈特机器人产业技术研究院有限公司 Robot spraying track generation method based on ROS platform
CN109541997B (en) * 2018-11-08 2020-06-02 东南大学 Spraying robot rapid intelligent programming method for plane/approximate plane workpiece
CN109465753A (en) * 2018-11-21 2019-03-15 无锡荣恩科技有限公司 The full-automatic sandblasting paint removal method of aviation components
CN109530935A (en) * 2018-12-12 2019-03-29 上海新时达机器人有限公司 The method for processing regular hole using mechanical arm is cut by laser
CN109696127A (en) * 2019-03-05 2019-04-30 浙江国自机器人技术有限公司 A kind of method and apparatus of laser identification
CN110288706B (en) * 2019-07-02 2020-08-07 哈尔滨工业大学 Fitting method of small-curvature irregular point cloud curved surface
CN111055293A (en) * 2019-11-21 2020-04-24 东南大学 High-precision constant-force grinding method for industrial robot based on curved surface self-adaption
CN111098055B (en) * 2020-01-06 2020-11-20 成都卡诺普自动化控制技术有限公司 Intelligent welding method for pipe plate
CN111390777A (en) * 2020-03-23 2020-07-10 宁波江丰电子材料股份有限公司 Method for automatically attaching sand blasting shielding layer

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102831306A (en) * 2012-08-13 2012-12-19 东南大学 Layered spraying track planning method used for spraying robot
CN102909148A (en) * 2012-08-13 2013-02-06 东南大学 Spraying path automatic generating method for adaptive modeling of multiple spray guns
CN103394430A (en) * 2013-08-15 2013-11-20 东南大学 Inter-sheet dead area optimization process based uniform-spraying manufacturing method for complex curved surface

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10086564B2 (en) * 2013-10-04 2018-10-02 Stratsys, Inc. Additive manufacturing process with dynamic heat flow control

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102831306A (en) * 2012-08-13 2012-12-19 东南大学 Layered spraying track planning method used for spraying robot
CN102909148A (en) * 2012-08-13 2013-02-06 东南大学 Spraying path automatic generating method for adaptive modeling of multiple spray guns
CN103394430A (en) * 2013-08-15 2013-11-20 东南大学 Inter-sheet dead area optimization process based uniform-spraying manufacturing method for complex curved surface

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
喷涂机器人轨迹优化关键技术研究;陈伟;《万方学位论文数据库》;20131231;第23-25、37、48页 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109405767A (en) * 2018-12-25 2019-03-01 威海筑丰五金有限公司 Method is determined based on the cast member dead head cutting track of laser profile tester
CN110434671A (en) * 2019-07-25 2019-11-12 王东 A kind of cast member surface machining track calibration method based on pattern measurement

Also Published As

Publication number Publication date
CN105354880A (en) 2016-02-24

Similar Documents

Publication Publication Date Title
Kasper et al. The KIT object models database: An object model database for object recognition, localization and manipulation in service robotics
Ding et al. Process planning for 8-axis robotized laser-based direct metal deposition system: A case on building revolved part
CN104759379B (en) Intelligent full-process closed-loop spray painting robot based on spray painting target three-dimensional imaging technology
Castagnetti et al. The domain of admissible orientation concept: a new method for five-axis tool path optimisation
CN102151984B (en) Laser machining method and device applicable for complicated curved surface
Ainsworth et al. CAD-based measurement path planning for free-form shapes using contact probes
JP5981143B2 (en) Robot tool control method
CN101726255B (en) Method for extracting interesting buildings from three-dimensional laser point cloud data
Harik et al. 5-axis flank milling: a state-of-the-art review
CN101786200B (en) Method for projection-type laser etching on free curved surface
CN108496124A (en) The automatic detection and robot assisted processing of surface defect
CN102905848B (en) Automated determination of jet orientation parameters in three-dimensional fluid jet cutting
CN100418027C (en) A helix guidance surfaces numerical control processing method
CN103678754B (en) Information processor and information processing method
CN100435055C (en) Method for planning smooth and non-interference tool route of 5-axis numerical control machining
CN105149794B (en) A kind of Intelligent Laser method for cutting edge based on binocular vision
ES2399883T3 (en) Procedure and system for displaying surface errors
US20130060369A1 (en) Method and system for generating instructions for an automated machine
CN105215987B (en) Industrial robot technology cloud system and working method thereof
CN106423656A (en) Automatic spraying system and automatic spraying method based on point cloud and image matching
CN102962452A (en) Metal laser deposition manufactured scan route planning method based on infrared temperature measurement images
CN103336485B (en) Rapid generating method of milling path of web of airplane structural member
CN106002277A (en) Electric arc additive and milling composite machining method and product thereof
CN101497279B (en) Measuring and machining integrated laser three-dimensional marking method and device
Li et al. Dual sliding mode contouring control with high accuracy contour error estimation for five-axis CNC machine tools

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant