CN106096497B - A kind of house vectorization method for polynary remotely-sensed data - Google Patents
A kind of house vectorization method for polynary remotely-sensed data Download PDFInfo
- Publication number
- CN106096497B CN106096497B CN201610360575.3A CN201610360575A CN106096497B CN 106096497 B CN106096497 B CN 106096497B CN 201610360575 A CN201610360575 A CN 201610360575A CN 106096497 B CN106096497 B CN 106096497B
- Authority
- CN
- China
- Prior art keywords
- point
- house
- edge line
- line segment
- edge
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
Abstract
The invention discloses a kind of house vectorization methods for polynary remotely-sensed data, comprising the following steps: acquisition is to be treated can to represent house location, the information of shape;Orderly marginal point is obtained using eight neighborhood Edge Following, while generating house edge binary images, marginal point is " 1 ", and non-edge point is " 0 ";Principal Axes Analysis is combined to obtain two mutually orthogonal Main ways in house using Radon transformation;The edge line in house is divided into three classes edge line segment: the first principal direction class, the second principal direction class, non-directional class;Accurate positioning to the edge line segment of the first principal direction class and the second principal direction class;Calculate the corner point of neighboring edge line segment;Generate the house boundary line of vector quantization.The present invention can avoid trouble and adverse effect that initial point and processing sequence are selected in house regularisation procedure, can also weaken the influence of buildings extraction resultant error, can obtain more coincide with house shape, effectively, the outline of house information of regularization.
Description
Technical field
The present invention relates to a kind of house vectorization methods for polynary remotely-sensed data, belong to technical field of image processing.
Background technique
As a kind of main geographical space research information, house (also referred to as building) update to two-dimensional map, wisdom city
Building, the supervision of the architecture against regulations in city etc. are of great significance.The continuous development of remote sensing technology is so that Automatic Vector extraction is distant
Houseclearing is possibly realized in sense data.
Polynary remotely-sensed data includes: the data such as image data, LiDAR data, DEM, DSM.Due to the interference of many factors,
The house patch (edge line) extracted from polynary remotely-sensed data at present is not often inconsistent with house true form, needs to carry out it
Post-processing --- denoising, regularization etc..
Contour of building rule method disclosed in 104156988 A of Chinese patent literature CN mainly utilizes modern house
Adjacent profile and border and orthogonal characteristic carry out outline of house rule information based on iteration minimum outsourcing rectangle, mainly for
The house point set obtained in LiDAR data.
Borderline distilling method outside building disclosed in 104200212 A of Chinese patent literature CN is mainly drawn using Doug
This algorithm and least squares line fitting method are straight with the length in fitting a straight line direction, fitting in conjunction with two Main ways in house
The threshold parameters such as the difference between line and two Main ways complete the regularization of house boundary line, to initial buildings extraction knot
The precision dependence of fruit is higher.
The disclosed method for rebuilding building outer profile polygon of 102938066 B of Chinese patent literature CN, for DSM
Data and image data rebuild outline of house using the complementary characteristic of two kinds of data.
Either LiDAR data or DSM data all have accurate elevation information.Elevation is house difference and its week
The important feature of exclosure object (road, vehicle, vegetation) has ensured that the precision of buildings extraction result is unlikely to too low.However, accurate
Elevation information and be not easy to obtain, this also cause using polynary remotely-sensed data (especially only with image data the case where) extract
House zigzag deformation it is more serious, existing outline of house information post-processing approach applicability is not strong.Therefore, how to not
The house (patch, edge line, boundary line etc.) extracted in congener polynary remotely-sensed data carries out vector quantization, weakens buildings extraction
As a result influence is particularly important.
The prior art especially rule method needs to face the select permeability of optimized start point, optimization process sequence, no
Same starting point and processing sequence may bring entirely different regularization result.The prior art needs to be arranged too many threshold value,
The degree of automation is not high.In addition, the situation that the prior art is preferable mainly for buildings extraction result precision, directly to extraction result
It is post-processed, and buildings extraction result is often unsatisfactory in practical application, still lacks effectively for not of the same race at present
The post-processing approach in house in class image data.
Summary of the invention
The present invention is exactly in view of the deficienciess of the prior art, provide a kind of house vector quantization for polynary remotely-sensed data
Method can avoid the trouble and adverse effect that select initial point and processing sequence in house regularisation procedure, can also weaken house
The influence for extracting resultant error, can obtain more coincide with house shape, effectively, the outline of house information of regularization, can be to more
First remotely-sensed data (LiDAR data, DEM, DSM, aviation image data, satellite image data, unmanned plane image data etc.) mentions
Result is taken to be post-processed.
To solve the above problems, the technical solution used in the present invention is as follows:
A kind of house vectorization method for polynary remotely-sensed data, comprising the following steps:
Step 1, acquisition is to be treated can represent house location, the information of shape;
Step 2, it obtains house edge: orderly marginal point being obtained using eight neighborhood Edge Following, while generating house side
Edge bianry image, marginal point are " 1 ", and non-edge point is " 0 ";
Step 3, Principal Axes Analysis is combined to obtain two mutually orthogonal Main ways in house using Radon transformation;
Step 4, the edge line in house is divided into three classes edge line segment: the first principal direction class, the second principal direction class, indefinite
To class;
Step 5, template is carried out on its method direction to all marginal points in the first principal direction class and the second principal direction class
Matching finds optimal match point to complete the accurate positioning of the edge line segment to the first principal direction class and the second principal direction class;
Step 6, on the basis of accurately obtaining the edge line segment up contour point of the first principal direction class and the second principal direction class,
Calculate the corner point of neighboring edge line segment;
Step 7, the house boundary line of vector quantization is generated.
Compared with prior art, implementation result of the invention is as follows by the present invention:
A kind of house vectorization method for polynary remotely-sensed data of the present invention, with the house spot for detecting or extracting
Block (or house point set, edge line) is used as given data, is carrying out multiclass segmentation to house edge line segment based on α-extension theory
On the basis of, it is performed corresponding processing respectively for the buildings extraction result of different remotely-sensed datas: higher to buildings extraction precision
The case where, using this priori knowledge of house shape, corner point is sought in conjunction with straight line fitting;The feelings not high to buildings extraction precision
Condition has this feature of higher sample rate in edge using image data, constructs adaptive border template to each side
Edge line segment is accurately positioned, and on the basis of obtaining proper boundary line segment, seeks corner point in conjunction with straight line fitting;It can obtain and room
Room shape more coincide, effectively, the outline of house information of regularization.
A kind of house vectorization method for polynary remotely-sensed data of the present invention mainly have the characteristics that and into
Step place:
1, there is the feature of Main way mostly using house, combine Principal Axes Analysis to replace with Radon transformation traditional
Hough transform checks two vertical Main ways in house, and calculation amount is few, and testing result is quick, effectively.
2, it is directed to edge line segmentation problem, introduces the minimization problem that multiclass segmentation thought is translated into energy function,
The directional information for both having considered each marginal point in this way also utilizes neighboring edge point and tends to this of a sort priori and knows
Know, the classification results of approximate global optimum may be implemented, by the way that original corner point and zigzag deformed region are divided into indefinite side
To class, the trouble and adverse effect of selection initial point and processing sequence can avoid.In addition, the multiclass segmentation of entire edge line segment
Process be it is full automatic, do not need to carry out parameter too many adjustment.
3, for house edge line zigzag deform cause buildings extraction precision not high situation, by with image data
It combines, sorted edge line segment is accurately positioned respectively using the adaptive edge template of building, after record location
The center point coordinate of edge line segment and its Main way, can reduce the influence of buildings extraction resultant error.
Detailed description of the invention
Fig. 1 is a kind of Technology Roadmap of the house vectorization method for polynary remotely-sensed data of the present invention;
Fig. 2 is a kind of flow diagram of the house vectorization method for polynary remotely-sensed data of the present invention;
Fig. 3 is the schematic diagram that house is obtained in specific embodiment of the invention step 1;
Fig. 4 is the schematic diagram that house edge is obtained in specific embodiment of the invention step 2;
Fig. 5 is the schematic diagram that house Main way is obtained in specific embodiment of the invention step 3;
Fig. 6 is the schematic diagram of border template building in specific embodiment of the invention step 5;
Fig. 7 is the schematic diagram that corresponding topography is intercepted in specific embodiment of the invention step 5;
Fig. 8 is the schematic diagram in the specific embodiment of the invention after the segmentation of house edge line multiclass;
Fig. 9 is the schematic diagram of house vector quantization result in the specific embodiment of the invention;
Figure 10 is the schematic diagram that house vector quantization result is superimposed with actual video in the specific embodiment of the invention.
Specific embodiment
Illustrate the contents of the present invention below in conjunction with specific embodiments.
As depicted in figs. 1 and 2, a kind of house vectorization method for polynary remotely-sensed data described in the present embodiment specifically flows
Journey is as follows:
Step 1, house to be treated is obtained.
As shown in figure 3, house described herein can be the house candidate point obtained using various means, it is also possible to room
Room patch, house boundary line etc. can represent house location, the information of shape: for example utilize multi-scale division combination supporting vector
House patch that machine is divided, the house point set outer contour obtained using point cloud segmentation technology and α-shapes algorithm, benefit
The house edge line etc. obtained with active contour method.
Step 2, house edge is obtained.
As shown in figure 4, being substantially carried out two steps herein: first is that orderly marginal point is obtained using eight neighborhood Edge Following, it is main
It is used for subsequent step 4, step 5 and step 6;Second is that generating house edge binary images, marginal point is " 1 ", non-edge point
For " 0 ".
Step 3, as shown in figure 5, using Radon transformation combine Principal Axes Analysis obtain two of house it is mutually orthogonal main
Direction.
Specific step is as follows:
31) it carries out Radon to initial profile line image to convert to obtain a cumulative ordered series of numbers, projection angle θ is 0 °~179 °;
In cumulative ordered series of numbers, row indicates that projection angle θ, column indicate projection aggregate-value.
32) aggregate-value (including maximum value) in each column less than 2 is removed, to reduce calculation amount.
33) maximum value of each column is added to the end value as this column with second largest value, generates the number of a 1x180
Group.
34) array is divided into two groups: an one group angular ranges is 0 °~89 °, and another group of angular range is 90 °~179 °;It connects
By the projection aggregate-value of second part and corresponding first part projection aggregate-value be added, i.e., 0 °+90 °, 1 °+91 ° ... ...,
89 °+179 °, to obtain 0 °~89 ° new angle accumulation values of first part.
35) last to resequence according to angle accumulated value to angle, θ (0 °~89 °), it is maximized θmaxAs first
Principal direction, the second principal direction is θ at this timemax+90°。
Step 4, the multiclass for carrying out edge line segment to the edge line in house is divided.
Edge line segment multiclass segmentation described herein, which refers to, is divided into three classes edge line segment: the first main side for the edge line in house
To class, the second principal direction class, non-directional class.
Specific step is as follows:
41) it calculates the local direction of each marginal point: using centered on current point P, take and successively take adjacent thereto half
Profile point within the scope of diameter R, i.e., 2R+1 point carry out principal component analysis, normal vector using first principal component as current point P (a,
B), the local direction of current point is calculated.
(1)
42) it calculates and connects each marginal point weight with the T chain of two vertex (+90 ° of θ, θ) respectively.
(2)
(3)
Wherein,WithRespectively point p belongs to the first principal directionClass and the second principal direction
Class needs the cost paid.CoefficientFor a penalty coefficient for disconnecting T chain.ParameterFor constant.
43) weight of the T chain of tie point p and uncertain class is calculated are as follows:
(4)
Wherein, due to this kind of directions be it is uncertain, we are with a fixed valueAs cut-off point p and do not know
The cost that class connection needs to pay.
44) weight of N chain between adjoining border point is calculated.
Tend to this of a sort priori knowledge according to neighboring edge point, if neighboring edge point is divided into same class, this
N chain between two o'clock is not destroyed, then the weight of N chain is 0;If neighboring edge point is not divided into same class, must
It must be punished, penalty value and the local direction of the two points are closely related, and corresponding calculation formula is as follows:
(5)
Wherein, coefficientIt is not divided into of a sort penalty coefficient for neighboring edge point, and introduces auxiliary node
Penalty coefficient.Neighboring edge point is as divided into the different classes of cost for needing to pay.
45) the energy function form of edge line segment segmentation are as follows:
(6)
In formula,Indicate the set of all marginal points in house, N is the neighborhood system of marginal point, is all made of eight neighbours herein
Domain.It is a kind of segmentation result of edge line.For marginal point p generic in segmentation result, there are three mark values: " 1 ",
" 2 ", " 3 " respectively indicate the first principal direction class, the second principal direction class, indefinite direction class.
On the basis of building energy function, energy function is resolved with by house edge using α-expansion algorithm
Line is divided into three classes edge line segment.
Step 5, using template matching method, edge line segment is accurately positioned.
Edge line segment described herein, which refers to, is divided into the first principal direction class and the second principal direction in the segmentation of edge line multiclass
The edge line segment of class.Because indefinite method class is mainly made of zigzag noise, corner point and its surrounding consecutive points, only right
First principal direction class and the second principal direction class are accurately positioned.Since there is deviation situation in each marginal point in same edge line segment
Difference, using to all marginal points in each edge line segment (the first principal direction class and the second principal direction class) in its method side
Template matching is carried out upwards, finds optimal match point to complete the accurate positioning to the edge line segment.
Detailed step:
51) selection of edge line segment: according to clockwise, each edge line segment is successively chosen。For room
The sum of first principal direction class and the edge line number of segment mesh of the second principal direction class in room edge line, it is assumed that the direction of edge line segment is;
52) building of border template: as shown in fig. 6, constructing initial edge template Mask first;Then rotary template is to mould
Edge direction is in plate, and the region II corresponds to background, the region I corresponds to house.
53) since first marginal point, each marginal point is successively chosen;
54) as shown in fig. 7, the central point of border template Mask is directed at candidate marginal, corresponding topography is intercepted,
Calculate the related coefficient simultaneously between logging template and topographyAnd difference.In template shown in Fig. 6, differenceIt can table
It is shown as:
(7)
In formula,、Respectively indicate the number of pixels of local image region corresponding with the region template I and II.It is figure
As window.WithRespectively indicate pixel set corresponding with the region template I and II in image window.
55) template is directed at candidate marginal, along the method direction (+90 ° of θ) of edge line segment, radius R(R=, as SR=
0.13, R=8) it is translated in the range of, while the related coefficient between calculation template and corresponding topical imageAnd difference。
56) X of best match window center, Y-coordinate, related coefficient, difference, edge line direction, belonging to edge line are recorded
Classification.Best match window has maximum difference, and its related coefficientGreater than specified threshold(for example,
0.5).IfOr, then it is assumed that best match window is not found, deletes the information recorded at this time.
57) return step 53), repeat, until having handled all candidate marginals.
58) judge whether edge line segment should be deleted.During carrying out template matching, it is assumed that on one edge line segment
Point is all removed, that is, does not find qualified match point, if the line segment meets following three conditions simultaneously,
Then retain its interior original point coordinate:
The length of the edge line segment is greater than, (i.e. 1.5 meters);
The edge line segment and adjacent edge line segment belong to inhomogeneity;
Its two adjacent edge line segment belongs to same class.
59) return step 51), next edge line segment is chosen, repeats, is until having handled all edge line segments
Only.
Step 6, corner point is calculated.
On the basis of obtaining proper boundary line segment up contour point, the corner point of neighboring edge line segment is calculated.
61) all positioning back edge line segments are directed to, each edge line is successively chosen;
62) calculate positioning back edge line segment center of gravity ().With the affiliated direction of edge line segmentWith center of gravity () table
Show current edge line segment, record straight line parameter ();
63) length of edge line segment is calculated,,For the line segment length in image space, SR is the sky of image
Between resolution ratio;
If 64), then return step 61).Otherwise continue, calculate each point(x, y) to corresponding line segment away from
From:
(8)
If, d=[0.5/SR] then delete marginal point (x, y) at this time.
65) return step 62), until all marginal points meet;
66) return step 61), repeat, until having handled all edge line segments.
67) calculate adjacent boundary line segment between corner point (,), the straight line parameter difference of adjacent two edges boundary line section
For (,), (,):
(9)
(10)
Wherein, parameterIt is to guarantee that denominator is not zero.
Step 7, the house boundary line of vector quantization is generated.
71) be sequentially connected each corner point, obtain all boundary points (,).Point S1(,), S2(,) between
The calculation formula of straight line parameter is as follows:
() (11)
() (12)
Wherein, point () it is house marginal point between point S1 and S2.
72) coordinate is converted.Since experimentation is carried out in digital picture coordinate system, the house boundary of vector quantization is needed
Again coordinate is assigned.The upper left corner (X, Y) of known lab diagram, after the conversion of all house boundary lines coordinate () are as follows:;。
Schematic diagram, the Yi Jifang of schematic diagram, house vector quantization result in the present embodiment after the segmentation of house edge line multiclass
The visible attached drawing 8 to 10 of schematic diagram that room vector quantization result is superimposed with actual video.
The foregoing is a detailed description of the present invention in conjunction with specific embodiments, and it cannot be said that the present invention is specifically real
It applies and is only limitted to these explanations.For those skilled in the art to which the present invention belongs, before not departing from present inventive concept
It puts, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to the scope of protection of the invention.
Claims (4)
1. a kind of house vectorization method for polynary remotely-sensed data, characterized in that the following steps are included:
Step 1, acquisition is to be treated can represent house location, the information of shape;
Step 2, it obtains house edge: orderly marginal point being obtained using eight neighborhood Edge Following, while generating house edge two
It is worth image, marginal point is " 1 ", and non-edge point is " 0 ";
Step 3, Principal Axes Analysis is combined to obtain two mutually orthogonal Main ways in house using Radon transformation;
Step 4, the edge line in house is divided into three classes edge line segment: the first principal direction class, the second principal direction class, non-directional
Class;
Step 5, template matching is carried out on its method direction to all marginal points in the first principal direction class and the second principal direction class,
Optimal match point is found to complete the accurate positioning of the edge line segment to the first principal direction class and the second principal direction class;
Step 6, it on the basis of accurately obtaining the edge line segment up contour point of the first principal direction class and the second principal direction class, calculates
The corner point of neighboring edge line segment;
Step 7, the house boundary line of vector quantization is generated;
And step 3 the following steps are included:
31) it carries out Radon to initial profile line image to convert to obtain a cumulative ordered series of numbers, projection angle θ is 0 °~179 °;It is cumulative
In ordered series of numbers, row indicates that projection angle θ, column indicate projection aggregate-value;
32) aggregate-value comprising maximum value in each column less than 2 is removed, to reduce calculation amount;
33) maximum value of each column is added to the end value as this column with second largest value, generates the array of a 1x180;
34) array is divided into two groups: an one group angular ranges is 0 °~89 °, and another group of angular range is 90 °~179 °;Then will
The projection aggregate-value of second part is added with corresponding first part projection aggregate-value, i.e., and 0 °+90 °, 1 °+91 ° ... ..., 89 °+
179 °, to obtain 0 °~89 ° new angle accumulation values of first part;
35) last to resequence according to angle accumulated value to angle, θ, θ range is 0 °~89 °, is maximized θmaxAs
One principal direction, the second principal direction is θ at this timemax+90° ;
And step 4 the following steps are included:
41) it calculates the local direction of each marginal point: using centered on current point p, take and successively take radius R model adjacent thereto
Interior profile point is enclosed, i.e., 2R+1 point carries out principal component analysis, using first principal component as the normal vector (a, b) of current point p, meter
Calculate the local direction of current point:
;
42) it calculates and connects each marginal point weight with the T chain of two vertex (+90 ° of θ, θ) respectively:
,
,
Wherein, Dp(θ) and Dp(+90 ° of θ) respectively point p belongs to+90 ° of classes needs of the first principal direction θ class and the second principal direction θ and pays
Cost, coefficient lambda1For a penalty coefficient for disconnecting T chain, parameter ksif is constant;
43) weight of the T chain of tie point p and uncertain class is calculated are as follows:
,
Wherein, due to direction it is uncertain thus with fixed value λ2The cost for needing to pay is connect with uncertain class as cut-off point p;
44) it calculates the weight of N chain between adjoining border point: this of a sort priori knowledge being tended to according to neighboring edge point, such as
Fruit neighboring edge point is divided into same class, and the N chain between this two o'clock is not destroyed, then the weight of N chain is 0;If phase
Adjacent marginal point is not divided into same class, then must be punished, penalty value and the local direction of the two points are closely related, accordingly
Calculation formula it is as follows:
,
Wherein, coefficientIt is not divided into of a sort penalty coefficient for neighboring edge point, and introduces the punishment system of auxiliary node
Number, q are to be not divided into of a sort neighboring edge point, V (L with cut-off point pp,Lq) be neighboring edge point is divided into it is different classes of
The cost for needing to pay;
45) the energy function form of edge line segment segmentation are as follows:
,
In formula,Indicate the set of all marginal points in house, N is the neighborhood system of marginal point, is all made of eight neighborhood, f herein
It is a kind of segmentation result of edge line, fpFor marginal point p generic in segmentation result, there are three mark values: " 1 ", " 2 ",
" 3 " respectively indicate the first principal direction class, the second principal direction class, indefinite direction class;
On the basis of building energy function, energy function is resolved to divide house edge line using α-expansion algorithm
It is segmented into three classes edge line segment.
2. a kind of house vectorization method for polynary remotely-sensed data as described in claim 1, characterized in that step 5 packet
Include following steps:
51) selection of edge line segment: according to each edge line segment Li clockwise, is successively chosen, i ∈ N, N are house side
The sum of first principal direction class and the edge line number of segment mesh of the second principal direction class in edge line, it is assumed that the direction of edge line segment is θ;
52) building of border template: building initial edge template Mask first;Then rotary template edge direction into template is
θ, and the region II corresponds to background, the region I corresponds to house;
53) since first marginal point, each marginal point is successively chosen;
54) central point of border template Mask is directed at candidate marginal, intercepts corresponding topography, calculate simultaneously logging template
Correlation coefficient ρ and difference dif between topography, difference dif may be expressed as:
,
In formula, NI、NIIThe number of pixels of local image region corresponding with the region template I and II is respectively indicated, f is image window
Mouthful, fIAnd fIIRespectively indicate pixel set corresponding with the region template I and II in image window;
55) template is directed at candidate marginal, along the method direction (+90 ° of θ) of edge line segment, is carried out in the range of radius R flat
It moves, wherein, while correlation coefficient ρ and difference dif between calculation template and corresponding topical image;
56) X of record best match window center, Y-coordinate, related coefficient, difference, edge line direction, edge line generic,
Best match window has maximum difference difmax, and its correlation coefficient ρ is greater than specified threshold ρTIf ρ < ρTOr difmax
< difT, then it is assumed that best match window is not found, deletes the information recorded at this time;
57) return step 53), repeat, until having handled all candidate marginals;
58) judge whether edge line segment should be deleted: during carrying out template matching, it is assumed that the point on one edge line segment is complete
Portion is removed, that is, does not find qualified match point, if the line segment meets following three conditions simultaneously, is protected
Stay its interior original point coordinate:
1. the length of the edge line segment is greater than;
2. the edge line segment and adjacent edge line segment belong to inhomogeneity;
3. its two adjacent edge line segment belongs to same class;
59) return step 51), next edge line segment is chosen, is repeated, until having handled all edge line segments.
3. a kind of house vectorization method for polynary remotely-sensed data as claimed in claim 2, characterized in that step 6 packet
Include following steps:
61) all positioning back edge line segments are directed to, each edge line is successively chosen;
62) center of gravity (x of positioning back edge line segment is calculated0,y0), with the affiliated direction of edge line segmentWith center of gravity (x0,y0) indicate to work as
Leading edge line segment records straight line parameter (x0,y0,θ);
63) length of edge line segment is calculated,,For the line segment length in image space, SR is image
Spatial resolution;
If 64) L≤1.5, return step 61), otherwise continue, calculate each point p(x, y) arrive the distance p of corresponding line segmenti:
,
If pi> d, d=[0.5/SR] then delete marginal point (x, y) at this time;
65) return step 62), until all marginal points meet Dis≤d;
66) return step 61), repeat, until having handled all edge line segments;
67) corner point (xi, yi) between adjacent boundary line segment is calculated, the straight line parameter of adjacent two edges boundary line section is respectively
(xi,yi, θi), (xi+1,yi+1, θi+1):
,
,
Wherein, parameterIt is to guarantee that denominator is not zero.
4. a kind of house vectorization method for polynary remotely-sensed data as claimed in claim 3, characterized in that step 7 packet
Include following steps:
71) it is sequentially connected each corner point, obtains all boundary points (x, y), corner point S1(x1, y1), corner point S2(x2, y2)
Between straight line parameter calculation formula it is as follows:
WhenWhen,
,
WhenWhen,
,
Wherein, point (x, y) is the house marginal point between point S1 and S2;
72) coordinate is converted: since experimentation is carried out in digital picture coordinate system, the house boundary of vector quantization needs weight
It is new to assign coordinate, it is known that the upper left corner (X, Y) of lab diagram, the coordinate after all house boundary line conversionsAre as follows:
;。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610360575.3A CN106096497B (en) | 2016-05-28 | 2016-05-28 | A kind of house vectorization method for polynary remotely-sensed data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610360575.3A CN106096497B (en) | 2016-05-28 | 2016-05-28 | A kind of house vectorization method for polynary remotely-sensed data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106096497A CN106096497A (en) | 2016-11-09 |
CN106096497B true CN106096497B (en) | 2019-08-06 |
Family
ID=57229485
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610360575.3A Active CN106096497B (en) | 2016-05-28 | 2016-05-28 | A kind of house vectorization method for polynary remotely-sensed data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106096497B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108460776B (en) * | 2018-04-12 | 2022-03-25 | 广东工业大学 | Brake pad size detection method and device based on machine vision |
CN109815307B (en) * | 2019-02-13 | 2021-07-09 | 北京百度网讯科技有限公司 | Position determination method, apparatus, device, and medium |
CN112700464B (en) * | 2021-01-15 | 2022-03-29 | 腾讯科技(深圳)有限公司 | Map information processing method and device, electronic equipment and storage medium |
CN113239752A (en) * | 2021-04-27 | 2021-08-10 | 西安万飞控制科技有限公司 | Unmanned aerial vehicle aerial image automatic identification system |
CN114549569B (en) * | 2022-03-01 | 2024-04-09 | 中国电建集团中南勘测设计研究院有限公司 | House contour extraction method based on two-dimensional orthographic image |
CN114863288B (en) * | 2022-07-05 | 2022-09-23 | 航天宏图信息技术股份有限公司 | Building contour extraction and regularization method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7925085B2 (en) * | 2006-03-27 | 2011-04-12 | Brother Kogyo Kabushiki Kaisha | Image data processing device and image data processing method |
CN102938066A (en) * | 2012-12-07 | 2013-02-20 | 南京大学 | Method for reconstructing outer outline polygon of building based on multivariate data |
CN105574521A (en) * | 2016-02-25 | 2016-05-11 | 民政部国家减灾中心 | House contour extraction method and apparatus thereof |
-
2016
- 2016-05-28 CN CN201610360575.3A patent/CN106096497B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7925085B2 (en) * | 2006-03-27 | 2011-04-12 | Brother Kogyo Kabushiki Kaisha | Image data processing device and image data processing method |
CN102938066A (en) * | 2012-12-07 | 2013-02-20 | 南京大学 | Method for reconstructing outer outline polygon of building based on multivariate data |
CN105574521A (en) * | 2016-02-25 | 2016-05-11 | 民政部国家减灾中心 | House contour extraction method and apparatus thereof |
Non-Patent Citations (1)
Title |
---|
高分辨率遥感影像的建筑物轮廓信息提取方法;周绍光等;《国土资源遥感》;20150930;第27卷(第3期);第52-58页 |
Also Published As
Publication number | Publication date |
---|---|
CN106096497A (en) | 2016-11-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106096497B (en) | A kind of house vectorization method for polynary remotely-sensed data | |
CN107092877B (en) | Remote sensing image roof contour extraction method based on building base vector | |
CN106548173B (en) | A kind of improvement no-manned plane three-dimensional information acquisition method based on classification matching strategy | |
CN103703490B (en) | For generation of the equipment of three-dimensional feature data and the method for generation of three-dimensional feature data | |
CN109859226B (en) | Detection method of checkerboard corner sub-pixels for graph segmentation | |
CN104123730A (en) | Method and system for remote-sensing image and laser point cloud registration based on road features | |
CN105069843A (en) | Rapid extraction method for dense point cloud oriented toward city three-dimensional modeling | |
CN104574347A (en) | On-orbit satellite image geometric positioning accuracy evaluation method on basis of multi-source remote sensing data | |
WO2018061010A1 (en) | Point cloud transforming in large-scale urban modelling | |
CN104408689A (en) | Holographic-image-based streetscape image fragment optimization method | |
CN104536009A (en) | Laser infrared composite ground building recognition and navigation method | |
CN104754182B (en) | High-resolution based on adaptive motion filtering is taken photo by plane the steady phase method of video | |
CN105606123B (en) | A kind of method of the photogrammetric automatic correcting digital ground elevation model of low-altitude aerial | |
JP2009223220A (en) | Road surface marking map creating method | |
CN111652241B (en) | Building contour extraction method integrating image features and densely matched point cloud features | |
CN110084743B (en) | Image splicing and positioning method based on multi-flight-zone initial flight path constraint | |
CN110569861A (en) | Image matching positioning method based on point feature and contour feature fusion | |
CN102521597A (en) | Hierarchical strategy-based linear feature matching method for images | |
CN107220996A (en) | A kind of unmanned plane linear array consistent based on three-legged structure and face battle array image matching method | |
CN109492525A (en) | A method of measurement antenna for base station engineering parameter | |
CN104180794B (en) | The disposal route in digital orthoimage garland region | |
CN117029870A (en) | Laser odometer based on road surface point cloud | |
CN103473787B (en) | A kind of based on moving target detecting method on the bridge of space geometry relation | |
Boerner et al. | Brute force matching between camera shots and synthetic images from point clouds | |
CN114742876B (en) | Land vision stereo measurement method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 230001 Hongfeng Road, Hefei hi tech Zone, Anhui Province, No. 55 Applicant after: Anhui province (Ministry of water resources Huaihe Water Conservancy Committee) Water Conservancy Science Research Institute (Anhui water conservancy project quality inspection center station) Address before: No. 55, Hongfeng Road, Hefei Hi-tech Zone, Hefei, Anhui Province Applicant before: Anhui and Huai River Water Resources Research Institute |
|
GR01 | Patent grant | ||
GR01 | Patent grant |