CN102136155B - Object elevation vectorization method and system based on three dimensional laser scanning - Google Patents
Object elevation vectorization method and system based on three dimensional laser scanning Download PDFInfo
- Publication number
- CN102136155B CN102136155B CN201010102541A CN201010102541A CN102136155B CN 102136155 B CN102136155 B CN 102136155B CN 201010102541 A CN201010102541 A CN 201010102541A CN 201010102541 A CN201010102541 A CN 201010102541A CN 102136155 B CN102136155 B CN 102136155B
- Authority
- CN
- China
- Prior art keywords
- point
- boundary
- border
- dimensional
- plane
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention relates to an object elevation vectorization method and system based on three dimensional laser scanning. The method comprises the steps of regional dividing: functional regions are extracted from the cloud data of three dimensional laser points on the surface of the object through a method of small plane dividing-polymerizing; boundary extracting: geometric mutation is detected through a directional image, and then regional boundaries are used for supplementing to extract the geometric edge information of the functional regions included in the three dimensional laser points; and boundary correcting: a convex decomposition method is used for correcting the boundaries of the functional regions to complete the reconfiguration of the object elevation. The system comprises a laser scanning unit, a regional dividing unit, a boundary extracting unit, a boundary correcting unit and a three dimensional model mapping unit. The regional dividing unit extracts the functional regions from the obtained cloud data of the three dimensional laser points; the boundary extracting unit extracts the boundary information of the functional regions included in the cloud data of the three dimensional laser points; and the boundary correcting unit fits the functional regions again to correct the boundaries.
Description
Technical field
The present invention relates to three-dimensional spatial information and obtain and the expression technology field, particularly a kind of in remote sensing, mapping, computer vision, use based on the 3 D laser scanning data to object facade vector quantization, carry out the method and system of three-dimensional reconstruction.
Background technology
Three-dimensional laser scanning technique is a kind of new and high technology of obtaining three-dimensional space data that middle nineteen nineties in last century begins to occur, and it obtains to large tracts of land high resolving power the three-dimensional coordinate data on measurand surface fast through the method for high-rate laser scanning survey.After the volume coordinate of obtaining each sampled point of body surface, obtain set a little, it is a large amount of spatial spreading dot matrix data, is referred to as " some cloud (Points Cloud) " usually.Because the three-dimensional data itself obtained comprises geometric properties, can directly write down the geometric attribute of body surface, therefore can be in detail, the shape of representation surface exactly, be used to set up the three-dimensional model of object.
In the prior art; Applicant of the present invention is at its paper " increment type three-dimensional rebuilding method " (Capital Normal University's master thesis; On June 13rd, 2007, postgraduate: Gong Lida, instructor: the real scene three-dimensional rebuilding method that has proposed a kind of increment type Zhang Aiwu); With the scene three dimensional point cloud that obtains based on the three-dimensional laser scanner system is research object, accomplishes the three-dimensional reconstruction of indoor and outdoor scene according to the characteristics of range image.Its method is divided into two stages; Phase one is that the data at single station are done vector quantization; Earlier the single site plane is cut apart in this stage, extracted the border of plane point set then, accomplish the vector quantization of single site range image; Subordinate phase is that the vector data that the phase one extracts is processed into final mask; Owing to have a large amount of non-convex polygons in the data boundary that extracts in the phase one, all be based on convex polygon in a lot of work of the drafting in later stage and processing, so the non-convex polygon that needs will obtain is processed into polygon.
The pressure that classic method is calculated is the data processing of multistation registration after merging, and has a large amount of redundant datas on the one hand, the calculating pressure when being the big and modeling that brings of data volume after being merged by website, abbreviation on the other hand.For overcoming the above problems, in the above-mentioned paper pressure that calculates is shared each website, earlier to the website independent vectorization, the result with vector quantization does processing such as protruding decomposition again, adds final mask one by one.Because the data volume of single site is little, it is little that single site is done the data processing calculating pressure, secondly; In the website merging process, merging be the data after the vector quantization, data volume reduces greatly; Therefore, calculating pressure is far smaller than classic method, has reduced the complexity of three-dimensional modeling simultaneously.
But single station data of handling in the phase one of above-mentioned paper are not upset the order of former three-dimensional point cloud.Usually, the cloud data that obtains by laser scanner, the influence from acquisition method, external environment to instrument itself, cloud data is always incomplete to the description of building surface, contains emptyly, and contains noise.The three dimensional point cloud that measured target object Direct Sampling is obtained is through after the pre-service such as denoising and filling-up hole; Relation does not exist between the point that cloud data is had when preliminary sweep; Become unordered mixed and disorderly sampling point set, lost the original ranks index information of each point on range image in the data file.In this case, the method for the single station of processing data and inapplicable in the phase one in the above-mentioned paper.
Summary of the invention
The objective of the invention is to, solve existing laser scanning data vector technology because the cloud data amount is big, data volume near or surpass the memory size vector quantization and be difficult to carry out, and therefrom extract characteristic and border ten minutes difficult technologies problem.
For achieving the above object, the present invention provides a kind of object facade vectorization method based on the 3 D laser scanning data, comprising:
Step a, Region Segmentation utilizes facet to cut apart-method of polymerization, abstraction function zone from the body surface three-dimensional laser point cloud data;
Step b, Boundary Extraction, how much sudden changes of direction of passage image detection replenish through the zone boundary again, extract the frontier point of the said functional area that three-dimensional laser point cloud comprises;
Step c, the border is revised, and the border that uses the said functional area of protruding decomposition method correction is accomplished the object facade and is rebuild.
The present invention further provides a kind of object facade vectored system based on the 3 D laser scanning data, comprising: laser scan unit, Region Segmentation unit, Boundary Extraction unit, border amending unit and three-dimensional model drawing unit;
Described laser scan unit obtains the three dimensional point cloud of object facade through transmitting and receiving laser;
The three dimensional point cloud abstraction function zone of described Region Segmentation unit from obtaining;
The boundary information of the said functional area that described Boundary Extraction unit detection three dimensional point cloud comprises;
Described border amending unit is the said functional area of match again, revises the boundary.
Beneficial effect of the present invention is, can detect the geometry marginal information that three-dimensional laser point cloud comprises, such as buildings corner angle etc.The notion of a directional image is proposed, and how much sudden changes of direction of passage image detection (comprising fold and step border), replenish through the zone boundary again, thereby fully extract the geometry marginal information that three-dimensional laser point cloud comprises.Region segmentation method of the present invention both had been applicable to that orderly three-dimensional point cloud also was applicable to unordered rambling three-dimensional point cloud.The present invention is applicable to all 3 D laser scanning data, handles through the present invention and has not only effectively extracted three-dimensional feature but also reduced data volume.
Description of drawings
Fig. 1 is the step of the cloud data vector quantization of the embodiment of the invention one;
Fig. 2 is the process flow diagram of the Region Segmentation of the embodiment of the invention one;
Fig. 3 is the process flow diagram of the Boundary Extraction of the embodiment of the invention one;
Fig. 4 is the process flow diagram that the border of the embodiment of the invention one is revised;
Fig. 5 is the step of the cloud data vector quantization of the embodiment of the invention two;
Fig. 6 is the process flow diagram that merge on the border of the embodiment of the invention two;
Fig. 7 is the system construction drawing of object facade vector quantization of the 3 D laser scanning data of the embodiment of the invention three;
Fig. 8 is the system construction drawing of object facade vector quantization of the 3 D laser scanning data of the embodiment of the invention four;
Fig. 9 a and Fig. 9 b be for being example with Capital Normal University's school gate, the result of the Region Segmentation of two different websites;
Figure 10 a is the x-image of a certain website;
Figure 10 b is the y-image of a certain website;
Figure 10 c is the z-image of a certain website;
Figure 10 d is the R-image of a certain website;
Figure 11 is the synoptic diagram of frontier point in the method for neighbours territory;
Figure 12 a is the zone boundary of different websites and the extraction result of how much sudden changes with Figure 12 b;
Figure 13 is the synoptic diagram that different site zone merge;
Figure 14 a is the temporary pattern according to the three-dimensional laser data extract of website 1 collection;
Figure 14 b is the temporary pattern after website 1 adds website 2;
Figure 14 c is the temporary pattern after website 1 adds website 2 and website 3;
Figure 14 d and Figure 14 e are according to the model observations from different perspectives after the three-dimensional laser data fusion of a plurality of websites collections;
Figure 15 is the synoptic diagram of little trapezoidal and protruding trapezoidal merging;
Figure 16 a to Figure 16 c is the observations with the not ipsilateral after protruding little trapezoidal merging;
Figure 16 d is the border correction result.
Embodiment
Below in conjunction with specific embodiment feature and advantage of the present invention are done further description.
The present invention is based on the buildings facade vectorization method of 3 D laser scanning data, can detect the geometry marginal information that three-dimensional laser point cloud comprises, such as buildings corner angle etc.The present invention proposes the notion of a directional image, and how much sudden changes of direction of passage image detection (comprising fold and step border), replenishes through the zone boundary, thereby fully extracts the geometry marginal information that three-dimensional laser point cloud comprises.Region segmentation method of the present invention both had been applicable to that orderly three-dimensional point cloud also was applicable to unordered rambling three-dimensional point cloud.The present invention is applicable to all 3 D laser scanning data, handles through the present invention and has not only effectively extracted three-dimensional feature but also reduced data volume.
Embodiment one
Embodiment one is used to explain the implementation step when method of the present invention is applied to the cloud data vector quantization with single website.
Please with reference to Fig. 1, the method based on the buildings facade vector quantization of three-dimensional laser data that the embodiment of the invention one provides comprises step:
Step 11, Region Segmentation extracts functional areas such as wall, door, eaves from the body surface three-dimensional laser point cloud data.
Step 12, Boundary Extraction, the border of detecting said functional area suddenlys change with how much.
Step 13, the border is revised, and uses protruding decomposition method to revise the boundary, and fairing processing, accomplishes body surface and rebuilds.
Region Segmentation in the said step 11 is to utilize facet to cut apart-the method abstraction function of polymerization zone.Image is divided into different zones, and lets different zones have different implications respectively, promptly the point in the same area has identical geometric properties.Please with reference to Fig. 2, the step that comprises is:
Step 1101, facet is cut apart.Cloud data is set up the spatial index based on linear Octree, distribute cloud data to get into cubic units, accomplish facet and cut apart.
Facet dividing method of the present invention is the Octree algorithm to the space dispersion point cloud, and the relative position of treating the 3 D laser scanning point cloud data does not require, and can be on the optional position in space.
Concrete steps are following:
At first, set minimum laser data that the Octree unit the comprises n that counts
Min, with tropism's angle threshold α
Max, coplanarity distance threshold d
MinThree threshold values.
Here, judge whether coplane according to PCA.Suppose that i small plane unit comprises n
iIndividual three-dimensional point, cog
iBe the center of gravity that this unit comprises three-dimensional point, so, covariance matrix arranged:
v
kBe a p
kPhasor coordinate, m is center of gravity cog
iPhasor coordinate.Ask the eigenvalue of covariance matrix A
0≤λ
1≤λ
2If, λ
0＞λ
Min, i three-dimensional point that the unit comprises coplane not then; Otherwise coplane, and the normal direction on plane is the minimal eigenvalue λ of A
0The characteristic of correspondence vector.
Octree is a kind of rule-based eight minutes principles, the segmenting structure that adopts the recurrence is olation to form.Its basic thought is that 3D region is divided into three-dimensional cubic units, comprises a plurality of cloud datas in each cubic units.Each cubic units is divided into little unit, is to divide with one minute eight principle.Cubic units is divided into eight identical small cubes of size, corresponding, all comprise a plurality of three-dimensional laser point cloud datas in each small cubes.
Through after the above-mentioned processing, three-dimensional point cloud just is dispersed in the different stereo-units, and participates in the contained three-dimensional point coplane of each cubic units of calculating, on same facet, is called first plane.
So-called merger travels through the point in all stereo-units exactly, and same tropism and coplanarity are judged in pointwise at 2.If satisfy two conditions simultaneously, then first plane with 2 places merges, and forms described second plane, otherwise will set up new plane set.
Two conditions that judge whether merger are respectively with tropism and coplanarity.
Be meant two planar processes to conforming judgement with the tropism, adjacent two facet i and j, it defines as follows with the tropism:
|n
i·n
j|＜α
max
Wherein, n
iAnd n
jBe respectively the normal vector of facet i and j, α
MaxBe preset same tropism's angle threshold.
Coplanarity is meant the judgement of two facet distances, adjacent two facet i and j, and its coplanarity defines as follows:
d
IjAnd d
JiThe center of gravity cog on a plane of expression is to its distance in another plane projection, d
MinBe preset coplanarity distance threshold.
Boundary Extraction in the said step 12 is that the direction of passage image carries out geometry sudden change (comprising fold and step border) detection, replenishes through the zone boundary, thereby fully extracts the geometry marginal information that three-dimensional laser point cloud comprises.Thereby give geometric meaning more specifically with the point set on second plane that splits in the step 11, the polygon that uses boundary profile to constitute replaces cutting apart the point set on second plane that obtains.
Please with reference to Fig. 3, the step that comprises is:
Travel through regional point set, with spot projection to the plane;
Travel through regional point set,, search the neighborhood point by neighbours territory direction.If a point is arranged in the neighborhood not at regional point set then for frontier point.Concrete condition such as accompanying drawing 2.
Step 12033, closure are not sealed segmental arc: if the distance that segmental arc is put end to end directly will be end-to-end less than fdisThrsh.If do not find, then find the segmental arc nearest with this segmental arc.If the distance between two segmental arcs then links up two segmental arcs less than fdisThrsh.Distance is defined as the two segmental arcs point-to-point transmission smallest point of distance in twos end to end between segmental arc.
It is the grid model that constructs functional area earlier through frontier point that border in the said step 13 is revised, and match functional area again calculates the friendship of adjacent area, thereby revises the boundary on this basis, improves the border precision.Divide the polygon that constitutes by frontier point and be actually one, and the frontier point area surrounded is a polygonal region for little trapezoidal.Please with reference to Fig. 4, concrete steps are following:
Implementation step by embodiment one can find out that the present invention proposes the notion of a directional image, and how much sudden changes of direction of passage image detection, comprises fold and step border.
Embodiment two
Embodiment two is used to explain the implementation step when method of the present invention is applied to the cloud data vector quantization of a plurality of websites.Different with embodiment one is to need to increase the step that merge on the border among the embodiment two.Please, comprise step with reference to Fig. 5:
Step 21, Region Segmentation extracts functional areas such as wall, door, eaves from the body surface three-dimensional laser point cloud data.
Step 22, Boundary Extraction, the border of detecting said functional area suddenlys change with how much.
Step 23, merge on the border, and merge fast on the multi-site border, constitutes the whole description of body surface vector quantization.
Step 24, the border is revised, and uses protruding decomposition method to revise the boundary, and fairing processing, accomplishes body surface and rebuilds.
Similar with embodiment one, the Region Segmentation in the said step 21 is to utilize facet to cut apart-the method abstraction function of polymerization zone.Image is divided into different zones, and lets different zones have different implications respectively, promptly the point in the same area has identical geometric properties.The step that comprises is:
Step 2101, facet is cut apart.Cloud data is set up the spatial index based on linear Octree, distribute cloud data to get into cubic units, accomplish facet and cut apart.
Facet dividing method of the present invention is the Octree algorithm to the space dispersion point cloud, and the relative position of treating the 3 D laser scanning point cloud data does not require, and can be on the optional position in space.
Concrete steps are following:
At first, set minimum laser data that the Octree unit the comprises n that counts
Min, with tropism's angle threshold α
Max, coplanarity distance threshold d
MinThree threshold values.
Step 21011 surrounds as outsourcing with a minimum cubic units pending laser point cloud data is surrounded into.
Step 21012 is judged the three-dimensional point coplane whether in the said cubic units.If coplane stops subdivision, execution in step 2102; If coplane not, then execution in step 21013.
Here, judge whether coplane according to PCA.Suppose that i small plane unit comprises n
iIndividual three-dimensional point, cog
iBe the center of gravity that this unit comprises three-dimensional point, so, covariance matrix arranged:
v
kBe a p
kPhasor coordinate, m is center of gravity cog
iPhasor coordinate.Ask the eigenvalue of covariance matrix A
0≤λ
1≤λ
2If, λ
0＞λ
Min, i three-dimensional point that the unit comprises coplane not then; Otherwise coplane, and the normal direction on plane is the minimal eigenvalue λ of A
0The characteristic of correspondence vector.
Step 21013 is judged counting whether greater than given threshold value n in the said minimum cubic units
MinIf then execution in step 21014, continue subdivision.If then execution in step 21015 not.
Step 21014, according to the Octree algorithm with the cubic units subdivision.Execution in step 21012 again.
Octree is a kind of rule-based eight minutes principles, the segmenting structure that adopts the recurrence is olation to form.Its basic thought is that 3D region is divided into three-dimensional cubic units, comprises a plurality of cloud datas in each cubic units.Each cubic units is divided into little unit, is to divide with one minute eight principle.Cubic units is divided into eight identical small cubes of size, corresponding, all comprise a plurality of three-dimensional laser point cloud datas in each small cubes.
Step 21015 is judged counting in the said cubic units, if counting less than given threshold value n of comprising in the cubic units
Min, and these put not coplane, and so, these points are temporarily abandoned, and this stereo-unit is illegal stereo-unit.
Step 21016, if be legal stereo-unit around the cubic units abandoned, then the point that comprises of this stereo-unit is a Null Spot, deletion; Otherwise keep, but do not participate in calculating.
Through after the above-mentioned processing, three-dimensional point cloud just is dispersed in the different stereo-units, and participates in the contained three-dimensional point coplane of each cubic units of calculating, on same facet, is called first plane.
Step 2102, the facet polymerization.The first plane merger with tropism and coplanarity of satisfying with in the adjacent cubic units is merged into a new bigger flat unit, is the functional area that extracts, and is called second plane.
Two conditions that judge whether merger are respectively with tropism and coplanarity.
Be meant two planar processes to conforming judgement with the tropism, adjacent two facet i and j, it defines as follows with the tropism:
|n
i·n
j|＜α
max
Wherein, n
iAnd n
jBe respectively the normal vector of facet i and j, α
MaxBe preset threshold value.
Coplanarity is meant the judgement of two facet distances, adjacent two facet i and j, and its coplanarity defines as follows:
d
IjAnd d
JiThe center of gravity cog on a plane of expression is to its distance in another plane projection, d
MinBe preset threshold value.
With Capital Normal University's school gate is example, and Fig. 9 a and Fig. 9 b are depicted as the result of the Region Segmentation of two different websites.
Similar with embodiment one, the Boundary Extraction in the said step 22 is that the direction of passage image carries out geometry sudden change (comprising fold and step border) detection, replenishes through the zone boundary, thereby fully extracts the geometry marginal information that three-dimensional laser point cloud comprises.
The step that comprises is:
Step 2201, how much sudden changes detect.The 2D grid that utilizes the laser scanning point arrangement is as the projected image plane, and can construct 4 width of cloth directional image (even more): x-image, y-image, z-image, R-image (range image), edge of image has reflected the uncontinuity of view data.With structure x-image is example, and putting in order of image pixel is the laser scanning sampling order.The putting in order of laser sampling point during laser sampling is a two-dimensional grid just.And with x value interpolation between 0-255, the data after the interpolation are the gray-scale value of respective pixel, thereby form a width of cloth x-image.In like manner, can form y-image, z-image, R-image.Four width of cloth directional image of one of them website of Figure 10 a to Figure 10 d explanation.Adopt the Wavelet Edge operator that 4 width of cloth images are carried out rim detection respectively.In fact, this 4 width of cloth image has reflected 4 variations along direction of scene, locates on the edge of, depicts different directions spacing step.The edge that again 4 directions is detected synthesizes, and constitutes the synthesis result of how much catastrophe points detections.
Step 2202, the zone boundary is detected.The point set on said second plane is projected on the plane, utilize neighbours territory method to extract the zone boundary.Crucial the choosing of projection plane that be extracted in the zone boundary.Projection plane is chosen the profile information that must not lose original point set.According to the characteristics of laser scanning sampling, the two-dimensional grid plane that utilizes laser sampling dot matrix sequence to constitute is the projecting plane.Concrete grammar is following:
Travel through regional point set, with spot projection to the plane;
Travel through regional point set,, search the neighborhood point by neighbours territory direction.If a point is arranged in the neighborhood not at regional point set then for frontier point.Concrete condition such as Figure 11.
Step 2203, frontier point connects.Concrete step is:
Step 22031 travels through regional point set and finds wherein minimax ranks iMaxRow, iMaxColumn, iMinRowm, iMinColumn.
Step 22032 begins by counterclockwise searching frontier point (zone boundary point and geometry catastrophe points) from the point of the row of minimum number, and record simultaneously is as a segmental arc.Lookup method is that the point of searching its direction has or not frontier point, has then to continue, and the point before then not noting constitutes segmental arc.And then number be criterion with minimum row again, select the NEW BEGINNING point, search again, go down successively, till all frontier points of traversal.If the next one point that finds in addition is the segmental arc starting point just, also look for starting point again, be recorded as new segmental arc.
Step 22033, closure are not sealed segmental arc: if the distance that segmental arc is put end to end directly will be end-to-end less than fdisThrsh.If do not find, then find the segmental arc nearest with this segmental arc.If the distance between two segmental arcs then links up two segmental arcs less than fdisThrsh.Distance is defined as the two segmental arcs point-to-point transmission smallest point of distance in twos end to end between segmental arc.
Figure 12 a is the zone boundary of different websites and the extraction result of how much sudden changes with Figure 12 b.
Different with embodiment one is to increase step 23.Merge on border in the said step 23 is that merge on the border of multi-site, is about to the border that each website extracts and is fused into an integral body, constitutes the full geometry of survey object is explained.Reduced the pressure of big data quantity so effectively to computing machine.Please with reference to Fig. 6, concrete steps are following:
Wherein, black color dots among Figure 13 and grey color dot are represented the projection of frontier point on the projecting plane that adjacent two websites extract, and white point representes to calculate the merging point that obtains behind the arithmetic mean of two points.
Figure 14 a representes the temporary pattern according to the three-dimensional laser data extract of website 1 collection.Website 1 is initial website.Figure 14 b representes the temporary pattern after website 1 adds website 2.Figure 14 c representes the temporary pattern after website 1 adds website 2 and website 3.Figure 14 d and Figure 14 e are the model observations from different perspectives after merging according to the three-dimensional laser data that a plurality of websites are gathered.
It is the grid model that constructs functional area earlier through frontier point that border in the said step 24 is revised, and match functional area again calculates the friendship of adjacent area, thereby revises the boundary on this basis, improves the border precision.The polygon that divide to constitute by frontier point is actually one and is little trapezoidal (Figure 15), and the frontier point area surrounded is a polygonal region.Concrete steps are following:
Step 2401 is carried out protruding decomposition, judges the concavity of frontier point.The frontier point area surrounded is a polygonal region, establishes polygon vertex p
1, p
2, p
3P
nBy counterclockwise arranging, press right-hand rule and judge concavity.To any 1 p
i, the some p that it is adjacent
I-1And p
I+1, p
I-1At p
iBefore the order, p
I+1After order, directed quantity then
And vector
Then be salient point if the cross product direction of two vectors meets inverse time pin mark preface closure, otherwise be concave point.If little four trapezoidal summits all are salient points, claim that so this is little trapezoidal protruding little trapezoidal.
Step 2402 is with protruding little trapezoidal merging.If adjacent little trapezoidal all be protruding little trapezoidal, then these protruding trapezoidal merging are become new polygon, merging the polygon that obtains is convex polygon (please with reference to Figure 15).Non-protruding little trapezoidal remaining.
Step 2403, the match functional area calculates adjacent area and hands over again, revises the boundary, and rebuilds the polygonal region of object facade.The result is referring to Figure 16 a to Figure 16 d.
Wherein Figure 16 a to Figure 16 c is the observations with the not ipsilateral after protruding little trapezoidal merging.Figure 16 d is the border correction result.
Implementation step by embodiment two can find out that the present invention proposes the notion of a directional image, and how much sudden changes of direction of passage image detection, comprises fold and step border; Vectorization method of the present invention is after the data of all websites are carried out Region Segmentation and Boundary Extraction, to carry out the border again to merge; In said step 23; According to multi-site registration transformed matrix other station data boundary treatment result is added the result of initial website one by one, so data processing amount reduces.
Embodiment three
Embodiment three provides a kind of system of object facade vector quantization of 3 D laser scanning data, in order to realize embodiment one described method.
Please with reference to Fig. 7, said system comprises laser scan unit, Region Segmentation unit, Boundary Extraction unit, border amending unit and three-dimensional model drawing unit.
Described laser scan unit obtains the three dimensional point cloud of object facade through transmitting and receiving laser.
The three dimensional point cloud abstraction function zone of described Region Segmentation unit from obtaining, for example wall, door, eaves etc.
Described Region Segmentation unit comprises that facet cuts apart module and facet polymerization module.
Described facet is cut apart module to the spatial index of cloud data foundation based on linear Octree, distributes cloud data to get into cubic units, accomplishes facet and cuts apart.Make all interior three-dimensional point of cubic units in same facet, be called first plane.Described facet polymerization module is with the first plane merger with tropism and coplanarity of satisfying in the adjacent cubic units, is merged into a new bigger flat unit, is called second plane, is the functional area that extracts.
Described Boundary Extraction unit detects the geometry marginal information that three-dimensional laser point cloud data comprises, for example the corner angle of buildings etc.How much sudden changes of direction of passage image detection (comprising fold and step border) replenish through the zone boundary, thereby fully extract the geometry marginal information that three-dimensional laser point cloud comprises.
Described Boundary Extraction unit comprises that how much sudden changes detect that mould is determined, zone boundary detection module and frontier point link block.
The 2D grid that described how much sudden change detection modules utilize the laser scanning point arrangement carries out rim detection as several directional image of projected image planar configuration to described directional image, and is again that the edge that detects is synthetic, constitutes the synthesis result of sudden change detection.
Described zone boundary detection module utilizes neighbours' method to extract the zone boundary.The two-dimensional grid plane that utilizes laser sampling dot matrix sequence to constitute is the projecting plane, travels through the point set on said second plane, projects on the said projecting plane.By neighbours territory direction, search the neighborhood point.If a point is arranged in the neighborhood not at regional point set then for frontier point.
Described frontier point link block constitutes the sealing segmental arc with the frontier point on the said how much sudden change testing results and second plane.
Described border amending unit constructs the grid model of functional area earlier through frontier point, match functional area again calculates the friendship of adjacent area, thereby revises the boundary on this basis, improves the border precision.
Described border amending unit comprises concavity judge module and protruding little trapezoidal merging module.
Said concavity judge module carries out protruding decomposition to the polygonal region that frontier point surrounds, and judges the concavity of frontier point.The polygon that divide to constitute by frontier point be actually one little trapezoidal.
Described protruding little trapezoidal merging module becomes new polygon with protruding little trapezoidal merging, non-protruding little trapezoidal remaining.Be specially,, claim that so this is little trapezoidal protruding little trapezoidal if little four trapezoidal summits all are salient points.If adjacent little trapezoidal all be protruding little trapezoidal, then these protruding trapezoidal merging are become new polygon, merging the polygon that obtains is convex polygon.
Described three-dimensional model drawing unit constructs the facade model that functional area is drawn becomes three-dimensional body with said border amending unit.
Embodiment four
Embodiment four provides a kind of system of object facade vector quantization of 3 D laser scanning data, in order to realize embodiment two described methods.
Please with reference to Fig. 8, said system comprises laser scan unit, Region Segmentation unit, Boundary Extraction unit, border amending unit, border integrated unit and three-dimensional model drawing unit.
Different with embodiment three is, has increased the border integrated unit in the present embodiment, is used to merge the cloud data that different websites are gathered.
The border that described border integrated unit extracts an a plurality of websites integral body that permeates constitutes the full geometry description to the survey object.
Described border integrated unit comprises multistation registration transformed matrix and frontier point update module.
Described multistation registration transformed matrix is that the frontier point that other websites extract is converted under the coordinate system of current stop.
The frontier point of described frontier point update module traversal block mold is sought the point with the minor increment that adds the website frontier point, replaces the frontier point in the original model with the arithmetic mean of two points.
Embodiment four carries out the data of all websites to carry out the border fusion again after Region Segmentation and the Boundary Extraction, therefore data processing amount is reduced.
The above description of this invention is illustrative, and nonrestrictive, and those skilled in the art is understood, and within spirit that claim limits and scope, can carry out many modifications, variation or equivalence to it, but they will fall in protection scope of the present invention all.
Claims (11)
1. the object facade vectorization method based on the 3 D laser scanning data is characterized in that, comprising:
Step a, Region Segmentation utilizes facet to cut apart-method of polymerization, abstraction function zone from the body surface three-dimensional laser point cloud data;
Step b, Boundary Extraction, how much sudden changes of direction of passage image detection; Replenish through the zone boundary again; Extract the frontier point of the said functional area that three-dimensional laser point cloud comprises, wherein, described directional image has reflected the variation of scene along direction; Locate on the edge of, depict different directions spacing step;
Step c, the border is revised, and the border that uses the said functional area of protruding decomposition method correction is accomplished the object facade and is rebuild, wherein,
Described step a Region Segmentation comprises:
Step a1, facet is cut apart, and cloud data is set up the spatial index based on linear Octree, distributes cloud data to get into cubic units, makes the interior three-dimensional point of said cubic units in one first plane;
Described step a Region Segmentation also comprises:
Step a2, the facet polymerization, the first plane merger with tropism and coplanarity of satisfying with in the adjacent cubic units is merged into one second plane and is described functional area.
2. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 1 is characterized in that described step b Boundary Extraction comprises:
Step b1; How much sudden changes detect; The 2D grid that utilizes the laser scanning point arrangement is constructed several directional image as the projected image plane, adopts the Wavelet Edge algorithm that said directional image is carried out rim detection respectively; Again that the detected edge of said several directional image is synthetic, constitute the synthesis result that the geometry catastrophe point detects.
3. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 2 is characterized in that described step b Boundary Extraction also comprises:
Step b2, the zone boundary is detected, and the two-dimensional grid plane that utilizes laser sampling dot matrix sequence to constitute is the projecting plane, and the point set on said second plane is projected on the described projecting plane, utilizes neighbours territory method to extract the zone boundary.
4. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 3 is characterized in that described step b Boundary Extraction also comprises:
Step b3, frontier point connects, and travels through the zone boundary point and how much catastrophe points on said second plane, constitutes closed segmental arc.
5. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 1 is characterized in that, described step c border is revised and to be comprised: little trapezoidally carry out protruding decomposition to what said frontier point surrounded, judge the concavity of said frontier point; With protruding little trapezoidal merging, with non-protruding little trapezoidal reservation; Again the said functional area of match revises the boundary.
6. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 1 is characterized in that, also comprises the step that merge on the border between said step b and the step c:
The frontier point that a plurality of websites are extracted is fused into an integral body.
7. the object facade vectorization method based on the 3 D laser scanning data as claimed in claim 6 is characterized in that, the step that merge on described border is:
According to multi-site registration transformed matrix the result of other station data frontier point is added initial website; The frontier point of traversal block mold is sought the point with the minor increment that adds the website frontier point, replaces the frontier point in the original model with the arithmetic mean of two points.
8. the object facade vectored system based on the 3 D laser scanning data is characterized in that, comprising: laser scan unit, Region Segmentation unit, Boundary Extraction unit, border amending unit and three-dimensional model drawing unit;
Described laser scan unit obtains the three dimensional point cloud of object facade through transmitting and receiving laser;
The three dimensional point cloud abstraction function zone of described Region Segmentation unit from obtaining;
The boundary information of the said functional area that described Boundary Extraction unit detection three dimensional point cloud comprises;
Described border amending unit is the said functional area of match again, revises the boundary;
Said three-dimensional model drawing unit is drawn the three-dimensional facade model that becomes object with the said functional area that said border amending unit simulates, wherein,
Described Region Segmentation unit comprises: facet is cut apart module and facet polymerization module;
Described facet is cut apart module to the spatial index of cloud data foundation based on linear Octree, distributes cloud data to get into cubic units, accomplishes facet and cuts apart, and makes all interior three-dimensional point of cubic units in one first plane;
Described facet polymerization module is with the first plane merger with tropism and coplanarity of satisfying in the adjacent cubic units, is merged into one second plane, is described functional area;
Described Boundary Extraction unit comprises: how much sudden changes detection module, zone boundary detection module and frontier point link blocks;
The 2D grid that described how much sudden change detection modules utilize the laser scanning point arrangement is as several directional image of projected image planar configuration; Described directional image is carried out rim detection, again that the edge that detects is synthetic, constitute the synthesis result that sudden change detects; Wherein, Described directional image has reflected the variation of scene along direction, locates on the edge of, depicts different directions spacing step;
Described zone boundary detection module utilizes neighbours territory method to extract the border on said second plane;
Described frontier point link block constitutes the sealing segmental arc with the result of said how much sudden change detection modules and the frontier point on second plane.
9. the object facade vectored system based on the 3 D laser scanning data as claimed in claim 8 is characterized in that, described border amending unit comprises concavity judge module and protruding little trapezoidal merging module;
Little trapezoidal protruding decomposition, the concavity of judgement frontier point of carrying out that described concavity judge module surrounds frontier point;
Described protruding little trapezoidal merging module becomes new polygon with protruding little trapezoidal merging, non-protruding little trapezoidal remaining.
10. the object facade vectored system based on the 3 D laser scanning data as claimed in claim 8; It is characterized in that; Further comprise the border integrated unit; Be used to merge the cloud data that different websites are gathered, the border that a plurality of websites an are extracted integral body that permeates constitutes the full geometry description to the survey object.
11. the object facade vectored system based on the 3 D laser scanning data as claimed in claim 10 is characterized in that described border integrated unit comprises multistation registration transformed matrix module and frontier point update module;
Described multistation registration transformed matrix module is that the frontier point that other websites extract is converted under the coordinate system of current stop;
The frontier point of described frontier point update module traversal block mold is sought the point with the minor increment that adds the website frontier point, replaces the frontier point in the original model with the arithmetic mean of two points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010102541A CN102136155B (en) | 2010-01-27 | 2010-01-27 | Object elevation vectorization method and system based on three dimensional laser scanning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010102541A CN102136155B (en) | 2010-01-27 | 2010-01-27 | Object elevation vectorization method and system based on three dimensional laser scanning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102136155A CN102136155A (en) | 2011-07-27 |
| CN102136155B true CN102136155B (en) | 2012-10-03 |
Family
ID=44295928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010102541A Active CN102136155B (en) | 2010-01-27 | 2010-01-27 | Object elevation vectorization method and system based on three dimensional laser scanning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102136155B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257222A (en) * | 2018-01-31 | 2018-07-06 | 杭州中科天维科技有限公司 | The automatic blending algorithm of steel stove converter three-dimensional laser point cloud |
Families Citing this family (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102446354A (en) * | 2011-08-29 | 2012-05-09 | 北京建筑工程学院 | Integral registration method of high-precision multisource ground laser point clouds |
| CN102436654B (en) * | 2011-09-02 | 2013-07-10 | 清华大学 | Adaptive segmentation method of building point cloud |
| CN103489222B (en) * | 2013-09-06 | 2016-06-22 | 电子科技大学 | Target body surface reconstruction method in 3-D view |
| CN104574282B (en) * | 2013-10-22 | 2019-06-07 | 鸿富锦精密工业（深圳）有限公司 | Point cloud noise spot removes system and method |
| CN104765915B (en) * | 2015-03-30 | 2017-08-04 | 中南大学 | Laser scanning data modeling method and system |
| CN105184854B (en) * | 2015-08-24 | 2018-01-16 | 北京麦格天宝科技股份有限公司 | For the fast modeling method of underground space scanning element cloud performance data |
| CN105184855B (en) * | 2015-08-25 | 2018-09-11 | 广州市城市规划勘测设计研究院 | Characteristic face construction method based on three-dimensional point cloud and device |
| CN105571571B (en) * | 2016-01-26 | 2017-11-17 | 中国科学院水利部成都山地灾害与环境研究所 | Accumulation section spatial structural form analysis method based on 3 D laser scanning |
| CN107784691A (en) * | 2016-08-26 | 2018-03-09 | 上海宝冶集团有限公司 | Curved boundary approximating method based on steel member feature |
| CN108541322B (en) * | 2016-08-29 | 2021-09-14 | 北京清影机器视觉技术有限公司 | Method and device for processing three-dimensional vision measurement data |
| CN108986024B (en) * | 2017-06-03 | 2024-01-23 | 西南大学 | Grid-based laser point cloud rule arrangement processing method |
| CN109598757A (en) * | 2017-09-30 | 2019-04-09 | 爱唯秀股份有限公司 | A method of for capturing the 3D model of object in space |
| CN109325953B (en) * | 2018-02-05 | 2021-09-21 | 黑龙江科技大学 | Method for determining large-scale dense point cloud normal |
| CN108645339A (en) * | 2018-05-14 | 2018-10-12 | 国能生物发电集团有限公司 | A kind of acquisition of bio-power plant material buttress point cloud data and calculation method of physical volume |
| CN108875804B (en) * | 2018-05-31 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Data processing method based on laser point cloud data and related device |
| CN108876744B (en) * | 2018-06-27 | 2020-01-17 | 大连理工大学 | Large-scale point cloud noise denoising method based on region segmentation |
| CN110132168A (en) * | 2019-05-13 | 2019-08-16 | 苏州嘉奕晟中小企业科技咨询有限公司 | A kind of three-dimensional laser point cloud data processing system |
| CN110310322B (en) * | 2019-07-06 | 2021-08-10 | 北方工业大学 | Method for detecting assembly surface of 10-micron-level high-precision device |
| CN110533778B (en) * | 2019-08-09 | 2021-01-12 | 中国科学院自动化研究所 | Large-scale image point cloud parallel distributed gridding reconstruction method, system and device |
| WO2021253429A1 (en) * | 2020-06-19 | 2021-12-23 | 深圳市大疆创新科技有限公司 | Data processing method and apparatus, and laser radar and storage medium |
| CN111708022B (en) * | 2020-07-15 | 2022-02-08 | 四川长虹电器股份有限公司 | Method and device for calculating scanning area boundary of millimeter wave radar |
| CN112017199B (en) * | 2020-10-26 | 2021-02-12 | 广东博智林机器人有限公司 | Floor boundary detection method, device, equipment and storage medium |
| CN113220018B (en) * | 2021-04-23 | 2023-03-28 | 上海发电设备成套设计研究院有限责任公司 | Unmanned aerial vehicle path planning method and device, storage medium and electronic equipment |
| CN113436335B (en) * | 2021-06-18 | 2023-06-30 | 招远市国有资产经营有限公司 | Incremental multi-view three-dimensional reconstruction method |
| CN113313761B (en) * | 2021-07-28 | 2022-04-01 | 盎锐(常州)信息科技有限公司 | Site acquisition method, scanning device and system for actual measurement |
| CN114609591B (en) * | 2022-03-18 | 2022-12-20 | 湖南星晟智控科技有限公司 | Data processing method based on laser point cloud data |
| CN115077437B (en) * | 2022-05-13 | 2023-06-20 | 东北大学 | Rock hydraulic fracturing crack morphology characterization method based on acoustic emission positioning constraint |
| CN114791270B (en) * | 2022-06-23 | 2022-10-25 | 成都飞机工业（集团）有限责任公司 | PCA-based aircraft surface key topography feature envelope measurement field construction method |
| CN117308821B (en) * | 2023-11-28 | 2024-02-06 | 江苏华辉建筑装饰工程有限公司 | Building decoration modeling precision inspection method and system based on scanner |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266694A (en) * | 2008-04-28 | 2008-09-17 | 武汉大学 | A network construction method for single-station territorial laser scanning point cloud |
| CN101581575A (en) * | 2009-06-19 | 2009-11-18 | 南昌航空大学 | Three-dimensional rebuilding method based on laser and camera data fusion |
-
2010
- 2010-01-27 CN CN201010102541A patent/CN102136155B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266694A (en) * | 2008-04-28 | 2008-09-17 | 武汉大学 | A network construction method for single-station territorial laser scanning point cloud |
| CN101581575A (en) * | 2009-06-19 | 2009-11-18 | 南昌航空大学 | Three-dimensional rebuilding method based on laser and camera data fusion |
Non-Patent Citations (3)
| Title |
|---|
| Hugues Hoppe et al.Surface Reconstruction from Unorganized Points.《ACM SIGGRAPH Computer Graphics》.1992,第26卷(第2期),第5.2节第2段. * |
| 路兴昌 等.基于激光扫描数据的三维可视化建模.《系统仿真学报》.2007,第19卷(第7期),全文. * |
| 龚俐达.增量式真三维重建方法.《中国优秀硕士学位论文全文数据库 基础科学辑》.2007,(第2期),摘要，第3.1节、第3.2节、第4.1节、第4.2节. * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257222A (en) * | 2018-01-31 | 2018-07-06 | 杭州中科天维科技有限公司 | The automatic blending algorithm of steel stove converter three-dimensional laser point cloud |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102136155A (en) | 2011-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102136155B (en) | Object elevation vectorization method and system based on three dimensional laser scanning | |
| Yu et al. | Automatic 3D building reconstruction from multi-view aerial images with deep learning | |
| Chen et al. | Topologically aware building rooftop reconstruction from airborne laser scanning point clouds | |
| Bazazian et al. | Fast and robust edge extraction in unorganized point clouds | |
| Suveg et al. | Reconstruction of 3D building models from aerial images and maps | |
| CN102938066B (en) | A kind of based on the polygonal method of multivariate data reconstruction buildings outline | |
| Xu et al. | Reconstruction of scaffolds from a photogrammetric point cloud of construction sites using a novel 3D local feature descriptor | |
| Cheng et al. | Integration of LiDAR data and optical multi-view images for 3D reconstruction of building roofs | |
| Bulatov et al. | Context-based automatic reconstruction and texturing of 3D urban terrain for quick-response tasks | |
| Becker | Generation and application of rules for quality dependent façade reconstruction | |
| CN106874580A (en) | A kind of bend pipe model reconstruction method based on cloud data | |
| CN103703490A (en) | Device for generating three-dimensional feature data, method for generating three-dimensional feature data, and recording medium on which program for generating three-dimensional feature data is recorded | |
| CN102750449B (en) | Point cloud linear feature extraction method based on substep three-dimensional space and feature dimension mapping | |
| CN103258203A (en) | Method for automatically extracting road centerline of remote-sensing image | |
| CN111652241B (en) | Building contour extraction method integrating image features and densely matched point cloud features | |
| CN115564926B (en) | Three-dimensional patch model construction method based on image building structure learning | |
| CN102622753A (en) | Semi-supervised spectral clustering synthetic aperture radar (SAR) image segmentation method based on density reachable measure | |
| CN111932669A (en) | Deformation monitoring method based on slope rock mass characteristic object | |
| Hu et al. | Geometric feature enhanced line segment extraction from large-scale point clouds with hierarchical topological optimization | |
| Taillandier | Automatic building reconstruction from cadastral maps and aerial images | |
| CN101383046A (en) | Three-dimensional reconstruction method on basis of image | |
| Zhou et al. | 3D building change detection between current VHR images and past lidar data | |
| Xu et al. | A method of 3d building boundary extraction from airborne lidar points cloud | |
| Bulatov et al. | Vectorization of road data extracted from aerial and uav imagery | |
| Shahzad et al. | Reconstruction of building footprints using spaceborne tomosar point clouds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |