CN103699900A - Automatic batch extraction method for horizontal vector contour of building in satellite image - Google Patents
Automatic batch extraction method for horizontal vector contour of building in satellite image Download PDFInfo
- Publication number
- CN103699900A CN103699900A CN201410003206.XA CN201410003206A CN103699900A CN 103699900 A CN103699900 A CN 103699900A CN 201410003206 A CN201410003206 A CN 201410003206A CN 103699900 A CN103699900 A CN 103699900A
- Authority
- CN
- China
- Prior art keywords
- image
- buildings
- pixel
- area
- region
- 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.)
- Granted
Links
Images
Abstract
The invention provides an automatic batch extraction method for horizontal vector contours of buildings in satellite images. The method comprises the steps of firstly using a K-means algorithm to class the images to obtain the backbone parts of the buildings, and solving the problem of selection of initial seed points by taking mass centers of all building areas; after the areas of all the seed points are grown, separating the building areas from surrounding areas by virtue of image edge information and removing non-building areas according to characteristic parameters such as rectangularity and strip index to realize the automatic extraction of the horizontal pixel contours of the buildings; then using techniques such has Hough transformation and block processing to perform linear vector processing to the horizontal pixel contours, and finally obtaining the horizontal vector contours of all buildings in a batch. The automatic batch extraction method for the horizontal vector contours of the buildings in the satellite images is applicable to the batch and quick extraction of the horizontal vector contours of common polygonal buildings with top views which are of straight-line segment structures in the satellite images.
Description
Technical field
The present invention relates to a kind of from single width groups of building satellite image automatic batch extract the method for buildings horizontal vector profile, the automatic batch that is particularly the horizontal vector profile of the polygonal buildings of linear structure to vertical view extracts.
Background technology
Utilizing Mono-satellite image to realize three-dimensional scenic Virtual Reconstruction is a very active research topic, and it is mainly used in the aspects such as physical construction planning, military scene simulation, resource management, earthquake relief work simulation.In the three-dimensional of real scene is rebuild, the overwhelming majority for simple in structure, vertical view be the polygonal common building thing of linear section structure, rapid modeling how to realize a large amount of common building things that exist of this class is the key of efficient reconstruction of three-dimensional groups of building, and how to realize the horizontal profile that extracts buildings from groups of building image, be the basis of efficient reconstruction of three-dimensional groups of building automatic batch, and determining the Distribution Pattern of groups of building and the matching degree of real scene in the following virtual scene of rebuilding.
In current research, for different application purposes, people have proposed various buildings Edge extraction algorithms, relatively common are snake modelling, level set curve evolvement method, region-growing method etc.The research object of above method is gray level image, is not suitable for coloured image, therefore cannot utilize colouring information abundant in coloured image.In addition, snake modelling is responsive to initial position, need to rely on other mechanism is placed on initial profile near interested characteristics of image, otherwise profile extracts can be failed, mostly adopt at present the manual way of choosing that initial boundary is set, not only very loaded down with trivial details, and automatically generate and caused difficulty to outline line.And comparatively conventional region-growing method exists 2 deficiencies: the first, be the On The Choice of initial seed point.Current method is manually to choose to the great majority of choosing of Seed Points, needs a large amount of artificial interferences, time and effort consuming, and efficiency is very low.The second, be the On The Choice of growing threshold.Growing threshold is excessive will there will be over-segmentation, and the target area area generating is often large than real area; And growing threshold is too small, can cause cutting apart deficiency, i.e. target area growth is imperfect.Therefore different buildingss need to be chosen different growing thresholds, and this need of work manually rule of thumb completes one by one.In addition, snake modelling and region-growing method for be single target region, also can only generate a target area at every turn, caused the poor efficiency of method.
The edge contour that above method is extracted, forms by pixel, is referred to as pixel profile here.The object that the horizontal profile of this patent research buildings extracts, to rebuild modeling data is provided for the three-dimensional of follow-up groups of building, need to further make Line vectorization to the buildings pixel profile extracting and process, the profile after here Line vectorization being processed is called vector outline.Current contour vector method is edge pixel is extracted and form list, then this list is carried out to straight-line segment matching, a distance threshold is wherein set, be defined as the ultimate range of off-straight, after surpassing this distance threshold, straight-line segment will be divided into two in proportion.The impact that the result of this method is chosen by distance threshold is larger, and in order to obtain good effect, each buildings in image needs corresponding different distance thresholds, and the method rule of thumb manual setting often of distance threshold was set in the past, and operating efficiency is low.
Summary of the invention
In order to overcome the deficiencies in the prior art, the invention provides a kind of from single width groups of building satellite image the automatic mode of rapid extraction buildings horizontal pixel profile in batches; A kind of new Line vectorization automatic processing method of choosing without distance threshold and other threshold values is further proposed, the buildings horizontal pixel profile extracting is carried out to Line vectorization processing, be met follow-up buildings three-dimensional and rebuild required horizontal vector profile.It is the batch rapid extraction of the horizontal vector profile of the polygonal common building thing of linear section structure that the method is specially adapted to vertical view in satellite image, applicable equally to these class groups of building of taking photo by plane in groups of building image.
The technical solution adopted for the present invention to solve the technical problems is: first utilize K-mean algorithm to carry out Images Classification, obtain the trunk portion of buildings, get the On The Choice that each construction zone barycenter solves initial seed point.Grow behind all Seed Points region, by image edge information, construction zone and peripheral region are separated, and get rid of non-construction zone according to characteristic parameters such as rectangular degree, banded indexes, thereby realize the automatic extraction of buildings horizontal pixel profile.Then utilize the technology such as Hough conversion, piece processing to carry out Line vectorization processing to horizontal pixel profile, the final horizontal vector profile that obtains in batches owned building.
In the present invention, the initial point of image coordinate system is positioned at the image upper left corner, and straight down, y axle positive dirction level to the right for x axle positive dirction.Concrete steps are as follows:
1. the automatic generation of different buildings seed points, comprises the following steps:
1) at Lab color space, utilize K-mean algorithm, on a of representative color information, b color subspace to source images I
origcarry out construction zone and cut apart, the class categories number that K-mean algorithm needs in carrying out equals image I
origpeak value number in the two-dimensional histogram of a, b color subspace, is stored as I by the image that is partitioned into buildings
seg;
2) by image I
segbe converted to gray level image, then be binary image I by greyscale image transitions
bw, wherein, construction zone is white, background is black, to binary image I
bwcarry out successively following processing:
(1) adopting window is that the median filter of 5*5 pixel removes pixel isolated in image;
(2) adopt the square structure operator of 5*5 pixel to carry out opening and closing operation, remove the region that area is less than square structure operator coverage;
(3) remove area and be less than s
areathe connected region of individual pixel, s
areaspan be the integer in [700,2000];
(4) remove area and be greater than I
areathe connected region of individual pixel, I
areaspan be the integer in [8000,10000];
3) construction zone is numbered respectively, and record its barycenter as the initial seed point of each buildings; If Seed Points position coordinates is (x
k, y
k), k is buildings numbering, x
kand y
kbe respectively the line number of k buildings seed point in image and row number,
the line number of i presentation video, the row number of j presentation video, f (i, j) represents that (i, j) locates the gray-scale value of pixel, D represents UNICOM region, symbol
represent downward rounding operation;
2. extract construction zone, comprise the following steps:
1) utilize region-growing method, generate the region of each candidate architecture thing, concrete steps are as follows:
(1) set up one and source images I
origequally size, pixel grey scale are all 0 image I
out, establishing growing threshold is thresh, its span is 8-10 pixel;
(2) orient an initial seed point (x
k, y
k), k is buildings numbering, and I is set
out(x
k, y
k)=1;
(3) with (x
k, y
k) centered by, consider (x
k, y
k) four neighborhood territory pixel (x
k-1, y
k), (x
k+ 1, y
k), (x
k, y
k-1), (x
k, y
k+ 1),, in Lab color space, according to Euclidean distance, calculate respectively the colour-difference of pixel and Seed Points in Seed Points neighborhood
Wherein, L (x, y), a (x, y), b (x, y) represent respectively L, a, the b component value of current pixel (x, y) in neighborhood, and L
seed, a
seed, b
seedl, a, the b component value that represent current Seed Points; If the I of current pixel (x, y) in neighborhood
out(x, y)=0, and d (x, y) <thresh, arrange I
out(x, y)=1 is pressed into storehouse by (x, y) simultaneously;
(4) from storehouse, eject a pixel and be used as new Seed Points, assignment is given (x
k, y
k), get back to step (3); When storehouse is sky, growth finishes;
(5) get back to step (2), orient next Seed Points, repeating step (3) and (4), until all Seed Points have scanned, whole growth course finishes;
2) to source images I
origafter gray processing, with Canny operator, detect strong edge, the pixel on strong edge is white, and other pixels are black, and the image that this is contained to strong marginal information is stored as I
edge;
3) traversal image I
out, work as I
outin pixel when white construction zone, check image I
edgewhether the pixel of middle correspondence position is in white edge, if so, by I
outmiddle respective pixel becomes black;
4) in image I
outin, calculate rectangular degree
length breadth ratio
with banded index
a wherein, l is respectively the area and perimeter in each region, A
e, h, w is area, length and the width of the minimum boundary rectangle in corresponding each region respectively; When d>0.7 and s<3 and r<5, determine that this region is buildings, being confirmed to be is that the region of buildings will remain, otherwise this region is become to black;
(1) at I
out_kin, utilize Hough change detection to go out the longest corresponding in buildings horizontal pixel profile straight-line segment, as the main shaft p of buildings; Take main shaft one end as fixed point, horizontal pixel profile is rotated to main shaft parallel with a coordinate axis of image coordinate system, specific practice is as follows:
If theta angle is the angle of x axle positive dirction in buildings main shaft p and image coordinate system, counterclockwise for just, unit is for spending; When | cos (theta) | during > cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the x axle of image coordinate system; When | cos (theta) | during≤cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the y axle of image coordinate system;
(2) by postrotational horizontal pixel profile by its extraneous rectangle at x, piecemeal uniformly-spaced in y direction, every block size is 5*3 pixel, is defined as cell block, remembers and is unit; After rotation, in the direction at main shaft place, cell block is defined as to 3 pixels, other direction is defined as 5 pixels; Calculating by a cell block of last column and a cell block of last row less than;
(3) in image I
out_kin, from top to bottom, each cell block of from left to right lining by line scan, calculates the number percent of construction zone area and cell block area in each cell block
wherein, A
buildthe area of construction zone in cell block, A
unitfor the area of cell block, square measure is number of pixels; When Percentage is more than or equal to 0.45, think that this cell block is construction zone, retain this cell block and be all filled to white, otherwise abandoning this cell block;
(4) the linearize profile obtaining is gone back to original position around the fixed point rotary-theta of main shaft angular setting, obtain the linearize profile of single buildings;
(5) respectively all construction zone are carried out the operation of above-mentioned steps (1)-(4), and each buildings linearize profile after processing is merged in same image, and be stored as polar plot, finally complete the batch extracting to owned building horizontal vector profile in groups of building on raw video.
The invention has the beneficial effects as follows: solved that existing groups of building horizontal pixel profile based on satellite image extracts and pixel profile Line vectorization process in the defect of semi-automation and manual methods, automatic Intelligent treatment technology has been proposed.First, the method of the different buildings initial seed point of a kind of automatic generation has been proposed, and utilize the over-segmentation of region growing and image edge information separated buildings and object around, obtain the horizontal pixel profile of buildings, overcome in existing region-growing method, initial seed point and growing threshold need the drawback of manually rule of thumb choosing one by one.Secondly, the needs of rebuilding for meeting follow-up three-dimensional, the horizontal pixel profile extracting is carried out to Line vectorization processing automatically, utilize Hough change detection to go out the main shaft in buildings pixel profile, coordinating block treatment technology carries out Line vectorization, thereby obtain the horizontal vector profile of buildings, having avoided zones of different in classic method needs the rule of thumb manual deficiency that different distance threshold value is set.The present invention can realize rapid extraction in batches to the horizontal vector profile that in satellite image, a large amount of vertical views that exist are the polygonal buildings of linear structure, for follow-up groups of building three-dimensional reconstruction provides modeling data efficiently.Same applicable to this class building of taking photo by plane in groups of building image.Can be widely used in many three-dimensional modelings and the virtual emulation fields such as physical construction planning, military scene simulation.
Accompanying drawing explanation
Fig. 1 is the angle theta schematic diagram of buildings horizontal pixel profile main shaft p and x axle, and p is obtained by Hough change detection.
Fig. 2 is horizontal pixel profile Line vectorization schematic diagram.
Fig. 3 is the process flow diagram of the inventive method.
In figure, 1 – buildings horizontal pixel profile, the main shaft p(fixed point that left upper end is p herein of 2 – buildings horizontal pixel profiles), the angle theta of 3 – main shaft p and x axle, 4 – imagination image coordinate systems move to initial point and overlap with main shaft fixed point, to show that theta angle forms, the extraneous rectangle of 5 – horizontal pixel profiles, 6 – cell block unit, a 7 – 5*3 pixel.
Embodiment
Below in conjunction with drawings and Examples, the present invention is further described, the present invention includes but be not limited only to following embodiment.
Concrete steps of the present invention are as follows:
First adopt Gaussian filter to carry out smoothing processing to groups of building satellite image, smoothly filter out some speckle noises that exist in image.Here adopt variances sigma=0.8, the Gaussian filter that window size is 7*7 carries out filtering to former image.Then in Hsv color space, keep the tone h value of each pixel constant, according to s=s+ Δ s, adjust saturation degree, according to v=v+ Δ v, adjust brightness, satellite image is carried out to image enhancement processing, the color contrast of buildings color and periphery scenery is strengthened, be convenient to distinguish buildings and non-construction zone, wherein the scope of Δ s is between 0.06-0.08, and the scope of Δ v is between 0.05-0.07.Image for subsequent treatment after pre-service is called to source images, is stored as I
orig.
1. the automatic generation of different buildings seed points
1) at Lab color space, utilize K-mean algorithm, on a of representative color information, b color subspace to image I
origcarry out construction zone and cut apart, the class categories number that K-mean algorithm needs in carrying out equals image I
origpeak value number in the two-dimensional histogram of a, b color subspace.The image that is partitioned into buildings is stored as to I
seg.Why selecting Lab color space, is that color is abundanter because it is more suitable in human eye perceives.
2) by K-mean algorithm, complete after cutting apart roughly of region, on satellite image, some ground close with buildings color, roof, road etc. are also divided in the classification of buildings, need to further process.First by image I
segby coloured image, be converted to gray level image, then be binary image by greyscale image transitions, be stored as I
bw.At binary image I
bwin, construction zone is white, background is black.To image I
bwcarry out successively following processing:
(1) medium filtering: employing window is 5*5(unit: median filter pixel), removes pixel isolated in image.
(2) opening and closing operation: adopt 5*5(unit: square structure operator pixel) carries out opening and closing operation, removes the region that area is less than square structure operator coverage.
(3) remove little connected region: remove area and be less than s
areathe connected region of individual pixel, main target is to remove the connected region on similar roof, ground and so on.Here s
areaspan be the integer in [700,2000].
(4) remove large connected region: remove area and be greater than I
areathe connected region of individual pixel; Here for the region that object is mainly ground, road homalographic is larger, I
areaspan be the integer in [8000,10000].
After above-mentioned a series of processing, trunk portion that can preserved building thing is got rid of most of non-partial building simultaneously, and wherein residual non-partial building will be in subsequent step 2 4) time removes.
3) finally construction zone is numbered respectively, and record its barycenter as the initial seed point of each buildings.If Seed Points position coordinates is (x
k, y
k), k is buildings numbering, x
kand y
kbe respectively the line number of k buildings seed point in image and row number, here using the position coordinates of initial seed point as (x
k, y
k) initial value, computing method are as follows:
In above formula, the line number of i presentation video, the row number of j presentation video, f (i, j) represents that (i, j) locates the gray-scale value of pixel, D represents UNICOM region, symbol
represent downward rounding operation.
2. extract construction zone
Mainly comprise two parts: use the buildings initial seed point initiated area growth obtaining in above-mentioned steps 1; Utilize source images I
origthe separated construction zone of marginal information and non-construction zone, avoid the deficiency that region-growing method in the past needs manually choose different buildingss different growing thresholds.Specifically be divided into the following steps:
1) utilize region-growing method, generate the region of each candidate architecture thing, by the selection assurance candidate architecture object area of growing threshold thresh, present the situation of over-segmentation, so that candidate architecture object area is able to complete reservation.Concrete steps are as follows:
(1) one of model and source images I
origequally size, pixel grey scale are all 0 image I
out, establishing growing threshold is thresh, its span is 8-10 pixel.
(2) by formula (1) and formula (2), orient an initial seed point (x
k, y
k), k is buildings numbering, and I is set
out(x
k, y
k)=1.
(3) with (x
k, y
k) centered by, consider (x
k, y
k) four neighborhood territory pixel (x
k-1, y
k), (x
k+ 1, y
k), (x
k, y
k-1), (x
k, y
k+ 1), in Lab color space, calculate respectively the colour-difference d (x, y) of pixel and Seed Points in Seed Points neighborhood according to Euclidean distance, computing method are:
In above formula, L (x, y), a (x, y), b (x, y) represent respectively L, a, the b component value of current pixel (x, y) in neighborhood, and L
seed, a
seed, b
seedl, a, the b component value that represent current Seed Points.
If the I of current pixel (x, y) in neighborhood
out(x, y)=0, and d (x, y) <thresh, arrange I
out(x, y)=1 is pressed into storehouse by (x, y) simultaneously.
(4) from storehouse, eject a pixel, it is used as to new Seed Points, assignment is given (x
k, y
k), get back to previous step (3).When storehouse is sky, growth finishes.Result makes I
outnear initial seed point, formed gradually construction zone (white portion).
(5) get back to step (2), orient next Seed Points, repeat above step (3) and (4), until all Seed Points have scanned, whole growth course finishes.The mode of choosing of growing threshold and Seed Points neighborhood remains unchanged during this time.
With source images I
origcontrast mutually the image I obtaining in said process
outshow as candidate architecture object area and occur over-segmentation, some non-construction zone of edges of regions also generate simultaneously, and this mode has guaranteed the integrality of candidate architecture object area, can guarantee the complete extraction of subsequent operation to candidate architecture object area.
2) to source images I
origafter gray processing, with Canny operator, detect strong edge, the pixel on strong edge is white, and other pixels are black, and the image that this is contained to strong marginal information is stored as I
edge.
3) traversal image I
out, work as I
outin pixel when white construction zone, check image I
edgewhether the pixel of middle correspondence position is in white edge, if so, by I
outmiddle respective pixel becomes black, i.e. the color of background.After above-mentioned processing, in image I
outmiddle candidate architecture thing will be separated with peripheral region, shows as large stretch of candidate architecture object area and be accompanied by trifling, scattered non-construction zone in image.
4) in image I
outin, utilize the characteristic parameters such as rectangular degree d, length breadth ratio r, banded index s to get rid of the non-buildings in non-construction zone and candidate architecture object area.The computing method of each characteristic parameter are respectively:
A wherein, l is respectively the area and perimeter in each region, A
e, h, w is area, length and the width of the minimum boundary rectangle in corresponding each region respectively.
In groups of building satellite image, the vertical view of common building thing is all generally the closed polygon of linear section structure, the feature with regular rectangular shape or similar rectangle, therefore can have compared with high rectangle degree, lower banded index and certain length breadth ratio, buildings is met conventionally: rectangular degree d is greater than 0.7, banded index s is less than 3.0, and length breadth ratio r is less than 5.0.According to above-mentioned three characteristic parameters, can get rid of the non-buildings in non-construction zone and candidate architecture object area.Judgment criterion is:
When d>0.7 and s<3 and r<5, determine that this region is buildings.
Being confirmed to be is that the region of buildings will remain, otherwise this region is become to black, i.e. the color of background.
The Line vectorization of step 3, horizontal pixel profile is processed
Can be uneven through the construction zone marginal portion pixel obtaining after step above, former should be the edge of straight line, has become the broken line consisting of many trifling little line segments, such situation is unfavorable for the carrying out of follow-up three-dimensional modeling work.In three-dimensional modeling, the medium and small line segment of the horizontal profile of buildings increases the surge that can cause the unilateral number of three-dimensional building model intermediate cam shape, directly affects the real-time of dummy emulation system operation, especially, in simulation on wide scene, affects more serious.Therefore this part utilizes the technology such as Hough conversion, piece processing to carry out Line vectorization processing to the horizontal pixel profile of buildings.
First, on the basis of step 2, utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering.To the single buildings image I splitting
out_kdo following processing:
(1) at I
out_kin, utilize Hough change detection to go out the longest corresponding in buildings horizontal pixel profile straight-line segment, as the main shaft p of buildings.Take main shaft one end as fixed point, horizontal pixel profile is rotated to main shaft parallel with a coordinate axis of image coordinate system, specific practice is as follows:
If theta angle is the angle of x axle positive dirction in buildings main shaft p and image coordinate system, counterclockwise for just, unit is for spending.
When | cos (theta) | during > cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the x axle of image coordinate system;
When | cos (theta) | during≤cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the y axle of image coordinate system.
(2) by postrotational horizontal pixel profile by its extraneous rectangle at x, piecemeal uniformly-spaced in y direction, every block size is 5*3 pixel, is defined as cell block, remembers and is unit.Owing to having the detailed information of these classes of protrusion such as balcony on common major axes orientation, therefore after rotation, in the direction at main shaft place, cell block is defined as to 3 pixels, in order to avoid lose detailed information, other direction is defined as 5 pixels.Calculating by a cell block of last column and a cell block of last row less than.
(3) in image I
out_kin, from top to bottom, each cell block of from left to right lining by line scan, calculates the number percent of construction zone area and cell block area in each cell block:
Wherein, A
buildthe area of construction zone in cell block, A
unitfor the area of cell block, square measure is number of pixels.
When Percentage is more than or equal to 0.45, think that this cell block is construction zone, retain this cell block and be all filled to white, otherwise abandoning this cell block.After the end of scan, all cell blocks that finally retain have formed new construction zone, and the edge of tracing area just can obtain the profile of linearize.
(4) the linearize profile obtaining is gone back to original position around the fixed point rotary-theta of main shaft angular setting, obtain the linearize profile of single buildings.
(5) respectively all construction zone are carried out the operation of above-mentioned (1)-(4), and each buildings linearize profile after processing is merged in same image, and be stored as polar plot, finally complete the batch extracting to owned building horizontal vector profile in groups of building on raw video.
In this example, the extraction of satellite image groups of building comprises following three steps:
First adopt the Gaussian filter that variances sigma=0.8, window size are 7*7, smoothly filter out some speckle noises that exist in groups of building satellite image.Then in Hsv color space, satellite image is done to following adjustment: each pixel tone h is constant, saturation degree s=s+ Δ s, brightness v=v+ Δ v, wherein the scope of Δ s is between 0.06-0.08, and the scope of Δ v, between 0.05-0.07, completes the image enhancement processing to image.Image for subsequent treatment after pre-service is called to source images, is stored as I
orig.
Operation steps is as follows in detail:
1. the automatic generation of different buildings seed points
1) at Lab color space, utilize K-mean algorithm, on a of representative color information, b color subspace to image I
origcarry out construction zone and cut apart, the image obtaining is stored as to I
seg.The class categories number that wherein K-mean algorithm needs equals image I
origpeak value number in the two-dimensional histogram of a, b color subspace.
2) by image I
segby coloured image, be converted to gray level image, then be binary image by greyscale image transitions, be stored as I
bw, wherein construction zone is white, background is black.To image I
bwcarry out successively following processing, get rid of territory, most non-building area:
(1) medium filtering: employing window is 5*5(unit: median filter pixel), removes pixel isolated in image.
(2) opening and closing operation: adopt 5*5(unit: square structure operator pixel) carries out opening and closing operation, removes the region that area is less than square structure operator coverage.
(3) remove little connected region: remove area and be less than s
areathe connected region of individual pixel, main target is to remove the connected region on similar roof, ground and so on.Here s
areaspan be the integer in [700,2000].
(4) remove large connected region: remove area and be greater than I
areathe connected region of individual pixel; Here for the region that object is mainly ground, road homalographic is larger, I
areaspan be the integer in [8000,10000].
After above-mentioned a series of processing, the residual non-partial building of minority will be in subsequent step 2 4) time removes.
3) finally construction zone is numbered respectively, and record its barycenter as the initial seed point of each buildings.If Seed Points position coordinates is (x
k, y
k), k is buildings numbering, x
kand y
kbe respectively the line number of k buildings seed point in image and row number, here using the position coordinates of initial seed point as (x
k, y
k) initial value, computing method are as follows:
In above formula, the line number of i presentation video, the row number of j presentation video, f (i, j) represents that (i, j) locates the gray-scale value of pixel, D represents UNICOM region, symbol
represent downward rounding operation.
2. extract construction zone
1) the buildings initial seed point based on obtaining in above-mentioned steps 1, utilize region-growing method, generate the region of each candidate architecture thing, by the selection assurance candidate architecture object area of growing threshold thresh, present the situation of over-segmentation, so that candidate architecture object area is able to complete reservation.Concrete steps are as follows:
(1) one of model and source images I
origequally size, pixel grey scale are all 0 image I
out, establishing growing threshold is thresh, its span is 8-10 pixel.
(2) by formula (1) and formula (2), orient an initial seed point (x
k, y
k), k is buildings numbering, and I is set
out(x
k, y
k)=1.
(3) with (x
k, y
k) centered by, consider (x
k, y
k) four neighborhood territory pixel (x
k-1, y
k), (x
k+ 1, y
k), (x
k, y
k-1), (x
k, y
k+ 1), in Lab color space, calculate respectively the colour-difference d (x, y) of pixel and Seed Points in Seed Points neighborhood according to Euclidean distance, computing method are:
In above formula, L (x, y), a (x, y), b (x, y) represent respectively L, a, the b component value of current pixel (x, y) in neighborhood, and L
seed, a
seed, b
seedl, a, the b component value that represent current Seed Points.
If the I of current pixel (x, y) in neighborhood
out(x, y)=0, and d (x, y) <thresh, arrange I
out(x, y)=1 is pressed into storehouse by (x, y) simultaneously.
(4) from storehouse, eject a pixel, it is used as to new Seed Points, assignment is given (x
k, y
k), get back to previous step (3).When storehouse is sky, growth finishes.Result makes I
outnear initial seed point, formed gradually construction zone (white portion).
(5) get back to step (2), orient next Seed Points, repeat above step (3) and (4), until all Seed Points have scanned, whole growth course finishes.The mode of choosing of growing threshold and Seed Points neighborhood remains unchanged during this time.
With source images I
origcontrast mutually the image I obtaining in said process
outshow as candidate architecture object area and occur over-segmentation, to guarantee the complete extraction of subsequent operation to candidate architecture object area.
2) to source images I
origafter gray processing, by Canny rim detection, go out strong edge, the pixel on strong edge is white, and other pixels are black.The image that this is contained to strong marginal information is stored as I
edge.
3) traversal image I
out, work as I
outin while being white construction zone, check image I
edgewhether middle correspondence position is white edge, when being white edge, by I
outmiddle respective pixel becomes black, i.e. the color of background.Like this, in image I
outmiddle candidate architecture thing will be separated with peripheral region, shows as large stretch of candidate architecture object area and be accompanied by trifling, scattered non-construction zone in image.
4) in image I
outin, utilize the characteristic parameters such as rectangular degree d, length breadth ratio r, banded index s to get rid of the non-buildings in non-construction zone and candidate architecture object area.The computing method of each characteristic parameter are respectively:
A wherein, l is respectively the area and perimeter in each region, A
e, h, w is area, length and the width of the minimum boundary rectangle in corresponding each region respectively.
According to criterion, " when the d>0.7 of connected domain and s<3 and the r<5, determine that this region is buildings." judgement I
outin each region whether be buildings, if buildings retains this region; Otherwise this region is become to black, i.e. the color of background.
The Line vectorization of step 3, horizontal pixel profile is processed
This part utilizes the technology such as Hough conversion, piece processing to carry out Line vectorization processing to the horizontal pixel profile of buildings.First, in image I
outin, utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering.To image I
out_kdo respectively following processing:
(1) at I
out_kin, utilize Hough change detection to go out the longest corresponding in buildings horizontal pixel profile straight-line segment, as the main shaft p of buildings.Take main shaft one end as fixed point, horizontal pixel profile is rotated to main shaft parallel with a coordinate axis of image coordinate system.If theta angle is the angle of x axle positive dirction in buildings main shaft p and image coordinate system, counterclockwise for just, unit is for spending.Theta equals 90 ° and represents this buildings horizontal positioned in image, and theta equals 0 ° and represents vertical placement.As theta, neither 0 °, in the time of neither 90 °, buildings shows as slant setting in image, therefore need to be rotated operation to it, makes its level or vertically places:
When | cos (theta) | during >=cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the x axle of image coordinate system;
When | cos (theta) | during≤cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the y axle of image coordinate system.
(2) by postrotational horizontal pixel profile by its extraneous rectangle at x, piecemeal uniformly-spaced in y direction, every block size is 5*3 pixel, is defined as cell block, remembers and is unit.Owing to having the detailed information of these classes of protrusion such as balcony on common major axes orientation, therefore after rotation, in the direction at main shaft place, cell block is defined as to 3 pixels, in order to avoid lose detailed information, other direction is defined as 5 pixels.Calculating by a cell block of last column and a cell block of last row less than.
(3) in image I
out_kin, from top to bottom, each cell block of from left to right lining by line scan, calculates the number percent of construction zone area and cell block area in each cell block:
Wherein, A
buildthe area of construction zone in cell block, A
unitfor the area of cell block, square measure is number of pixels.
When Percentage is more than or equal to 0.45, think that this cell block is construction zone, retain this cell block and be all filled to white, otherwise abandoning this cell block.The all cell blocks that retain after the end of scan have formed new construction zone, and the edge of tracing area just can obtain the profile of linearize.
(4) the linearize profile obtaining is gone back to original position around the fixed point rotary-theta of main shaft angular setting, obtain the linearize profile of single buildings.
(5) respectively all construction zone are carried out the operation of above-mentioned (1)-(4), and each buildings linearize profile after processing is merged in same image, and be stored as polar plot, finally complete the batch extracting to owned building horizontal vector profile in groups of building on raw video.
Embodiment 1: somewhere, Beijing satellite image that is a 702*902 pixel to original image carries out the extraction of contour of building.
Adopt variances sigma=0.8, the Gaussian filter that window size is 7*7 carries out smothing filtering to former image; In Hsv space, keep the h value of each pixel constant, v=v+0.08, s=s+0.07 carries out image enhancement processing, and the color contrast of buildings and background is strengthened.Image after processing is stored as to I
orig, as the source images of subsequent treatment.
Image I
origin a of Lab color space, b color subspace, peak value number is 3, and as it being carried out to the classification number of construction zone while cutting apart by K-mean algorithm, the Image Saving after cutting apart is I
seg, in this embodiment, construction zone is present in the 3rd class.Then to image I
segcarry out gray processing and binary conversion treatment, obtain its binary image I
bw, I now
bwin connected region number be 100, comprising buildings and non-construction zone.Then travel through each connected region, get rid of area and be less than s
area=2000 and area be greater than I
area=10000 region, utilizing size is 5*5(unit: square structure operator pixel) carries out opening operation to image, removes the region that area is less than square structure operator coverage, now bianry image I
bwthe number of middle connected region has become 9, and the trunk portion of buildings has 9.Calculate respectively the center-of-mass coordinate of each connected region as the initial seed point coordinate of each buildings, as shown in table 1:
The initial seed point coordinate of each buildings of table 1
Zone number | Seed Points x coordinate | Seed Points y coordinate |
1 | 100 | 352 |
2 | 245 | 479 |
3 | 319 | 107 |
4 | 345 | 700 |
5 | 352 | 217 |
6 | 463 | 230 |
7 | 540 | 739 |
8 | 563 | 314 |
9 | 655 | 366 |
Utilize 9 initial seed point of trying to achieve above, choose growing threshold thresh=10, utilize region-growing method to obtain 9 candidate architecture object areas, image saves as I
out.In conjunction with Canny operator from source images I
origin detected strong edge, by image I
outin candidate architecture object area separated with territory, non-building area.In image I
outin, calculate respectively the rectangular degree d in each region, banded index s, tri-characteristic parameters of length breadth ratio r, wherein the characteristic parameter of candidate architecture object area is as shown in table 2 (because non-construction zone is trifling, scattered, irregular and be easy to differentiation, no longer enumerates here.):
The characteristic parameter of nine candidate architecture object areas of table 2
Zone number | Rectangular degree d | Banded index s | Length |
1 | 0.75 | 0.0029 | 1.9 |
2 | 0.77 | 0.0027 | 2.26 |
3 | 0.83 | 0.0035 | 1.98 |
4 | 0.71 | 0.003 | 2.83 |
5 | 0.76 | 0.004 | 2.49 |
6 | 0.82 | 0.0031 | 1.91 |
7 | 0.76 | 0.0028 | 2.53 |
8 | 0.85 | 0.003 | 2.06 |
9 | 0.55 | 0.0093 | 1.12 |
According to criterion, " when d>0.7 and s<3 and the r<5, determine that this region is buildings." judge whether each region is buildings.Wherein the discontented toe mark in region 9, therefore casts out this region, has so far tentatively obtained the horizontal pixel profile of all 8 buildingss in image.
Utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering.Utilize Hough change detection to go out each I
out_kthe main shaft p of middle buildings horizontal pixel profile, the angle theta of difference calculated level pixel profile main shaft and x axle, according to condition | cos (theta) |>=cos (45 °) and | cos (theta) |≤cos (45 °) rotates horizontal pixel profile to level or upright position, facilitates next step operation.
The angle theta of each buildings horizontal pixel profile main shaft of table 3 and x axle
Zone number | The angle theta of main shaft and x axle |
1 | -42.02° |
2 | -41.28° |
3 | -41.11° |
4 | -30.37° |
5 | -22.38° |
6 | -40.99° |
7 | -30.04° |
8 | -41.66° |
In this example, in table 3, buildings, all in obliquity, therefore all needs it to be rotated.Utilize the cell block unit that size is 5*3 pixel respectively single postrotational construction zone to be divided, judge that construction zone in each cell block (white portion) accounts for the number percent Percentage of whole cell block region area, when Percentage>0.45, think that this cell block is construction zone and is all filled to white, handle after whole cell blocks, buildings horizontal pixel profile is rotated back to initial position.Successively as above operation is done in owned building region, the edge of tracing area just can obtain the contour of building after linearize, finally each buildings linearize profile is merged in same image, and is stored as polar plot.
Embodiment 2: the satellite image that is the dormitory area, Purdue University of a 775*401 pixel to original image carries out the extraction of contour of building.
Adopt variances sigma=0.8, the Gaussian filter that window size is 7*7 carries out smothing filtering to former image; In Hsv space, keep the h of each pixel constant, v=v+0.06, s=s+0.05 carries out image enhancement processing, and the color contrast of buildings and background is strengthened.Image after processing is stored as to I
orig, as the source images of subsequent treatment.
Image I
origin a of Lab color space, b color subspace, peak value number is 3, and as it being carried out to the classification number of construction zone while cutting apart by K-mean algorithm, the Image Saving after cutting apart is I
seg, in this embodiment, construction zone is present in the 2nd class.Then to image I
segcarry out gray processing and binary conversion treatment, obtain its binary image I
bw, I now
bwin connected region number be 16, comprising buildings and non-construction zone.Then travel through each connected region, get rid of area and be less than s
area=700 and area be greater than I
area=8000 region, utilizing size is 5*5(unit: square structure operator pixel) carries out opening operation to image, removes the region that area is less than square structure operator coverage, this bianry image I
bwthe number of middle connected region has become 9, and the trunk portion of buildings has 9.Now calculate respectively the center-of-mass coordinate of each connected region as the initial seed point coordinate of each buildings, as shown in table 4:
The initial seed point coordinate of each buildings of table 4
Zone number | Seed Points x coordinate | Seed Points y coordinate |
1 | 108 | 45 |
2 | 96 | 194 |
3 | 182 | 366 |
4 | 207 | 209 |
5 | 359 | 179 |
6 | 437 | 58 |
7 | 542 | 333 |
8 | 513 | 179 |
9 | 665 | 118 |
Utilize 9 initial seed point of trying to achieve above, choose growing threshold thresh=8, utilize region-growing method to obtain 9 candidate architecture object areas, image saves as I
out.In conjunction with Canny operator from source images I
origin detected strong edge, by image I
outin candidate architecture object area separated with territory, non-building area.In image I
outin, calculate respectively the rectangular degree d in each region, banded index s, tri-characteristic parameters of length breadth ratio r, wherein the characteristic parameter of candidate architecture object area is as shown in table 5 (because non-construction zone is trifling, scattered, irregular and be easy to differentiation, no longer enumerates here.):
The characteristic parameter of nine candidate architecture object areas of table 5
Zone number | Rectangular degree d | Banded index s | Length |
1 | 0.85 | 0.003 | 4.75 |
2 | 0.83 | 0.0036 | 4.35 |
3 | 0.92 | 0.0035 | 5.32 |
4 | 0.87 | 0.0032 | 4.46 |
5 | 0.82 | 0.0032 | 4.11 |
6 | 0.97 | 0.006 | 1.93 |
7 | 0.84 | 0.0033 | 4.7 |
8 | 0.86 | 0.0032 | 4.37 |
9 | 0.94 | 0.0032 | 4.83 |
According to criterion, " when d>0.7 and s<3 and the r<5, determine that this region is buildings." judge whether each region is buildings.Above-mentioned 9 connected regions all meet the parameter index of construction area, have so far tentatively obtained the horizontal pixel profile of all 9 buildingss in image.
Utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering.Utilize Hough change detection to go out each I
out_kthe main shaft p of middle buildings horizontal pixel profile, the angle theta of difference calculated level pixel profile main shaft and x axle, according to condition | cos (theta) |>=cos (45 °) and | cos (theta) |≤cos (45 °) rotates horizontal pixel profile to level or upright position, facilitates next step operation.
The angle theta of each buildings horizontal pixel profile main shaft of table 6 and x axle
Zone number | The angle theta of main shaft and |
1 | 90° |
2 | 0° |
3 | 90° |
4 | 0° |
5 | 90° |
6 | 90° |
7 | 89.5° |
8 | 0° |
9 | 90° |
In this example, in table 6, buildings, all in level or upright position, therefore does not need it to be rotated.Utilize the cell block unit that size is 5*3 pixel respectively single postrotational construction zone to be divided, judge that construction zone in each cell block (white portion) accounts for the number percent Percentage of whole cell block region area, when Percentage>0.45, think that this cell block is construction zone and is all filled to white, handle after whole cell blocks, buildings horizontal pixel profile is rotated back to initial position.Successively as above operation is done in owned building region, the edge of tracing area just can obtain the contour of building after linearize, finally each buildings linearize profile is merged in same image, and is stored as polar plot.
Embodiment 3: the satellite image that is the somewhere, Xi'an of a 874*383 pixel to original image carries out the extraction of contour of building.
Adopt variances sigma=0.8, the Gaussian filter that window size is 7*7 carries out smothing filtering to former image; In Hsv space, keep the h of each pixel constant, v=v+0.07, s=s+0.06 carries out image enhancement processing, and the color contrast of buildings and background is strengthened.Image after processing is stored as to I
orig, as the source images of subsequent treatment.
Image I
origin a of Lab color space, b color subspace, peak value number is 5, and as it being carried out to the classification number of construction zone while cutting apart by K-mean algorithm, the Image Saving after cutting apart is I
seg, in this embodiment, construction zone is present in the 1st class.Then to image I
segcarry out gray processing and binary conversion treatment, obtain its binary image I
bw, I now
bwin connected region number be 84, comprising buildings and non-construction zone.Then travel through each connected region, get rid of area and be less than s
area=1000 and area be greater than I
area=9000 region, utilizing size is 5*5(unit: square structure operator pixel) carries out opening operation to image, removes the region that area is less than square structure operator coverage, this bianry image I
bwthe number of middle connected region has become 12, and the trunk portion of buildings has 12.Calculate respectively the center-of-mass coordinate of each connected region as the initial seed point coordinate of each buildings, as shown in table 7:
The initial seed point coordinate of each buildings of table 7
Zone number | Seed Points x coordinate | Seed Points y coordinate |
1 | 15 | 39 |
2 | 37 | 146 |
3 | 93 | 72 |
4 | 152 | 168 |
5 | 167 | 346 |
6 | 239 | 195 |
7 | 404 | 192 |
8 | 406 | 310 |
9 | 356 | 10 |
10 | 566 | 253 |
11 | 778 | 308 |
12 | 764 | 191 |
Utilize 12 initial seed point of trying to achieve above, choose growing threshold thresh=9, utilize region-growing method to obtain 12 candidate architecture object areas, image saves as I
out.In conjunction with Canny operator from source images I
origin detected strong edge, by image I
outin candidate architecture object area separated with territory, non-building area.In image I
outin, calculate respectively the rectangular degree d in each region, banded index s, tri-characteristic parameters of length breadth ratio r, wherein the characteristic parameter of candidate architecture object area is as shown in table 8 (because non-construction zone is trifling, scattered, irregular and be easy to differentiation, no longer enumerates here.):
The characteristic parameter of 12 candidate architecture object areas of table 8
Zone number | Rectangular degree d | Banded index s | Length |
1 | 0.71 | 0.0083 | 9.81 |
2 | 0.59 | 0.011 | 5.99 |
3 | 0.65 | 0.005 | 2.35 |
4 | 0.85 | 0.0036 | 1.93 |
5 | 0.7 | 0.0093 | 1.18 |
6 | 0.45 | 0.0046 | 2.5 |
7 | 0.81 | 0.004 | 4.65 |
8 | 0.83 | 0.0039 | 4.62 |
9 | 0.55 | 0.0093 | 1.12 |
10 | 0.89 | 0.0037 | 4.74 |
11 | 0.89 | 0.0036 | 4.5 |
12 | 0.79 | 0.0046 | 3.36 |
According to criterion, " when d>0.7 and s<3 and the r<5, determine that this region is buildings." judge whether each region is buildings.Wherein the discontented toe mark in region 1,2,3,5,6,9, therefore casts out this 6 regions, and remaining 6 connected regions all meet the parameter index of construction area, have so far tentatively obtained the horizontal pixel profile of all 6 buildingss in image.
Utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering.Utilize Hough change detection to go out each I
out_kthe main shaft p of middle buildings horizontal pixel profile, the angle theta of difference calculated level pixel profile main shaft and x axle, according to condition | cos (theta) |>=cos (45 °) and | cos (theta) |≤cos (45 °) rotates horizontal pixel profile to level or upright position, facilitates next step operation.
The angle theta of each buildings horizontal pixel profile main shaft of table 9 and x axle
Zone number | The angle theta of main shaft and |
1 | 45° |
2 | -80.5° |
3 | 90° |
4 | 0° |
5 | -48.8° |
6 | 90° |
In this example, in table 9, buildings 1,2,5, in obliquity, need to be rotated it, and remaining construction thing does not need to be rotated processing.Utilize the cell block unit that size is 5*3 pixel respectively the construction zone of (if needs) after single rotation to be divided, judge that construction zone in each cell block (white portion) accounts for the number percent Percentage of whole cell block region area, when Percentage>0.45, think that this cell block is construction zone and is all filled to white, handle after whole cell blocks, the buildings horizontal pixel profile rotating through is rotated back to initial position again.Successively as above operation is done in owned building region, the edge of tracing area just can obtain the contour of building after linearize, finally each buildings linearize profile is merged in same image, and is stored as polar plot.
Claims (2)
1. a buildings horizontal vector profile automatic batch extracting method in satellite image, is characterized in that comprising following step
Rapid:
Step 1, employing Gaussian filter carry out smoothing processing to groups of building satellite image, variances sigma=0.8 of described Gaussian filter, and window size is 7*7 pixel; Then in Hsv color space, keep the tone of each pixel constant, according to step value Δ s, adjust saturation degree, according to step value Δ v, adjust brightness, satellite image is carried out to image enhancement processing, obtain source images I
orig, wherein Δ s is 0.06-0.08, Δ v is 0.05-0.07;
Step 2, profile extract automatically, comprise the following steps:
1. the automatic generation of different buildings seed points, comprises the following steps:
1) at Lab color space, utilize K-mean algorithm, on a of representative color information, b color subspace to source images I
origcarry out construction zone and cut apart, the class categories number that K-mean algorithm needs in carrying out equals image I
origpeak value number in the two-dimensional histogram of a, b color subspace, is stored as I by the image that is partitioned into buildings
seg;
2) by image I
segbe converted to gray level image, then be binary image I by greyscale image transitions
bw, wherein, construction zone is white, background is black, to binary image I
bwcarry out successively following processing:
(1) adopting window is that the median filter of 5*5 pixel removes pixel isolated in image;
(2) adopt the square structure operator of 5*5 pixel to carry out opening and closing operation, remove the region that area is less than square structure operator coverage;
(3) remove area and be less than s
areathe connected region of individual pixel, s
areaspan be the integer in [700,2000];
(4) remove area and be greater than I
areathe connected region of individual pixel, I
areaspan be the integer in [8000,10000];
3) construction zone is numbered respectively, and record its barycenter as the initial seed point of each buildings; If Seed Points position coordinates is (x
k, y
k), k is buildings numbering, x
kand y
kbe respectively the line number of k buildings seed point in image and row number,
the line number of i presentation video, the row number of j presentation video, f (i, j) represents that (i, j) locates the gray-scale value of pixel, D represents UNICOM region, symbol
represent downward rounding operation.
2. extract construction zone, comprise the following steps:
1) utilize region-growing method, generate the region of each candidate architecture thing, concrete steps are as follows:
(1) set up one and source images I
origequally size, pixel grey scale are all 0 image I
out, establishing growing threshold is thresh, its span is 8-10 pixel;
(2) orient an initial seed point (x
k, y
k), k is buildings numbering, and I is set
out(x
k, y
k)=1;
(3) with (x
k, y
k) centered by, consider (x
k, y
k) four neighborhood territory pixel (x
k-1, y
k), (x
k+ 1, y
k), (x
k, y
k-1), (x
k, y
k+ 1),, in Lab color space, according to Euclidean distance, calculate respectively the colour-difference of pixel and Seed Points in Seed Points neighborhood
wherein, L (x, y), a (x, y), b (x, y) represent respectively L, a, the b component value of current pixel (x, y) in neighborhood, and L
seed, a
seed, b
seedl, a, the b component value that represent current Seed Points; If the I of current pixel (x, y) in neighborhood
out(x, y)=0, and d (x, y) <thresh, arrange I
out(x, y)=1 is pressed into storehouse by (x, y) simultaneously;
(4) from storehouse, eject a pixel and be used as new Seed Points, assignment is given (x
k, y
k), get back to step (3); When storehouse is sky, growth finishes;
(5) get back to step (2), orient next Seed Points, repeating step (3) and (4), until all Seed Points have scanned, whole growth course finishes;
2) to source images I
origafter gray processing, with Canny operator, detect strong edge, the pixel on strong edge is white, and other pixels are black, and the image that this is contained to strong marginal information is stored as I
edge;
3) traversal image I
out, work as I
outin pixel when white construction zone, check image I
edgewhether the pixel of middle correspondence position is in white edge, if so, by I
outmiddle respective pixel becomes black;
4) in image I
outin, calculate rectangular degree
length breadth ratio
with banded index
a wherein, l is respectively the area and perimeter in each region, A
e, h, w is area, length and the width of the minimum boundary rectangle in corresponding each region respectively; When d>0.7 and s<3 and r<5, determine that this region is buildings, being confirmed to be is that the region of buildings will remain, otherwise this region is become to black;
Step 3, utilize the method for mark connected region by image I
outin single buildings split, be stored as respectively I
out_k, k is buildings numbering; To the single buildings image I splitting
out_kdo following processing:
(1) at I
out_kin, utilize Hough change detection to go out the longest corresponding in buildings horizontal pixel profile straight-line segment, as the main shaft p of buildings; Take main shaft one end as fixed point, horizontal pixel profile is rotated to main shaft parallel with a coordinate axis of image coordinate system, specific practice is as follows:
If theta angle is the angle of x axle positive dirction in buildings main shaft p and image coordinate system, counterclockwise for just, unit is for spending; When | cos (theta) | during > cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the x axle of image coordinate system; When | cos (theta) | during≤cos (45 °), buildings horizontal pixel profile is rotated to main shaft parallel with the y axle of image coordinate system;
(2) by postrotational horizontal pixel profile by its extraneous rectangle at x, piecemeal uniformly-spaced in y direction, every block size is 5*3 pixel, is defined as cell block, remembers and is unit; After rotation, in the direction at main shaft place, cell block is defined as to 3 pixels, other direction is defined as 5 pixels; Calculating by a cell block of last column and a cell block of last row less than;
(3) in image I
out_kin, from top to bottom, each cell block of from left to right lining by line scan, calculates the number percent of construction zone area and cell block area in each cell block
wherein, A
buildthe area of construction zone in cell block, A
unitfor the area of cell block, square measure is number of pixels; When Percentage is more than or equal to 0.45, think that this cell block is construction zone, retain this cell block and be all filled to white, otherwise abandoning this cell block;
(4) the linearize profile obtaining is gone back to original position around the fixed point rotary-theta of main shaft angular setting, obtain the linearize profile of single buildings;
(5) respectively all construction zone are carried out the operation of above-mentioned steps (1)-(4), and each buildings linearize profile after processing is merged in same image, and be stored as polar plot, finally complete the batch extracting to owned building horizontal vector profile in groups of building on raw video.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410003206.XA CN103699900B (en) | 2014-01-03 | 2014-01-03 | Building horizontal vector profile automatic batch extracting method in satellite image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410003206.XA CN103699900B (en) | 2014-01-03 | 2014-01-03 | Building horizontal vector profile automatic batch extracting method in satellite image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103699900A true CN103699900A (en) | 2014-04-02 |
CN103699900B CN103699900B (en) | 2016-10-05 |
Family
ID=50361422
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410003206.XA Active CN103699900B (en) | 2014-01-03 | 2014-01-03 | Building horizontal vector profile automatic batch extracting method in satellite image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103699900B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927541A (en) * | 2014-04-21 | 2014-07-16 | 西北工业大学 | Target edge line detection method applicable to space tethered robot capture process |
CN107330979A (en) * | 2017-06-30 | 2017-11-07 | 电子科技大学中山学院 | Vector diagram generation method and device for building house type and terminal |
CN108323389A (en) * | 2018-01-18 | 2018-07-27 | 华南农业大学 | The detection method and device of the rice transplanting rice shoot spacing in the rows and cave rice shoot number of rice transplanter |
CN108665489A (en) * | 2017-03-27 | 2018-10-16 | 波音公司 | The method and data processing system of variation for detecting geographical space image |
CN109034073A (en) * | 2018-07-30 | 2018-12-18 | 深圳大学 | Predict method, system, equipment and the medium of building demolition waste yield |
CN109345454A (en) * | 2018-09-18 | 2019-02-15 | 徐庆 | Method, storage medium and the system of bitmap images vector quantization |
CN110619282A (en) * | 2019-08-26 | 2019-12-27 | 海南撰云空间信息技术有限公司 | Automatic extraction method for unmanned aerial vehicle orthoscopic image building |
CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
CN111523391A (en) * | 2020-03-26 | 2020-08-11 | 上海刻羽信息科技有限公司 | Building identification method, system, electronic device and readable storage medium |
CN111652892A (en) * | 2020-05-02 | 2020-09-11 | 王磊 | Remote sensing image building vector extraction and optimization method based on deep learning |
CN114581786A (en) * | 2021-12-28 | 2022-06-03 | 深圳市城市产业发展集团有限公司 | Method and device for estimating building area according to ground image |
CN114998664A (en) * | 2022-07-18 | 2022-09-02 | 中国科学院烟台海岸带研究所 | Rapid detection method and device for micro-plastic in seawater by multiple optical platforms |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101978395A (en) * | 2008-04-23 | 2011-02-16 | 株式会社博思科 | Building roof outline recognizing device, building roof outline recognizing method, and building roof outline recognizing program |
CN103020342A (en) * | 2012-12-04 | 2013-04-03 | 南京大学 | Method for extracting contour and corner of building from ground LiDAR data |
CN103198494A (en) * | 2013-04-03 | 2013-07-10 | 武汉海达数云技术有限公司 | Building elevation extraction method based on building outer contour lines and laser-point cloud |
CN103218618A (en) * | 2013-01-09 | 2013-07-24 | 重庆交通大学 | Highway route automatic extraction method based on remote-sensing digital image |
US20130222375A1 (en) * | 2008-11-05 | 2013-08-29 | Hover, Inc. | 3d building modeling |
-
2014
- 2014-01-03 CN CN201410003206.XA patent/CN103699900B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101978395A (en) * | 2008-04-23 | 2011-02-16 | 株式会社博思科 | Building roof outline recognizing device, building roof outline recognizing method, and building roof outline recognizing program |
US20130222375A1 (en) * | 2008-11-05 | 2013-08-29 | Hover, Inc. | 3d building modeling |
CN103020342A (en) * | 2012-12-04 | 2013-04-03 | 南京大学 | Method for extracting contour and corner of building from ground LiDAR data |
CN103218618A (en) * | 2013-01-09 | 2013-07-24 | 重庆交通大学 | Highway route automatic extraction method based on remote-sensing digital image |
CN103198494A (en) * | 2013-04-03 | 2013-07-10 | 武汉海达数云技术有限公司 | Building elevation extraction method based on building outer contour lines and laser-point cloud |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927541A (en) * | 2014-04-21 | 2014-07-16 | 西北工业大学 | Target edge line detection method applicable to space tethered robot capture process |
CN103927541B (en) * | 2014-04-21 | 2017-02-22 | 西北工业大学 | Target edge line detection method applicable to space tethered robot capture process |
CN108665489A (en) * | 2017-03-27 | 2018-10-16 | 波音公司 | The method and data processing system of variation for detecting geographical space image |
CN108665489B (en) * | 2017-03-27 | 2023-03-21 | 波音公司 | Method for detecting changes in geospatial images and data processing system |
CN107330979A (en) * | 2017-06-30 | 2017-11-07 | 电子科技大学中山学院 | Vector diagram generation method and device for building house type and terminal |
CN107330979B (en) * | 2017-06-30 | 2020-09-18 | 电子科技大学中山学院 | Vector diagram generation method and device for building house type and terminal |
CN108323389A (en) * | 2018-01-18 | 2018-07-27 | 华南农业大学 | The detection method and device of the rice transplanting rice shoot spacing in the rows and cave rice shoot number of rice transplanter |
CN109034073B (en) * | 2018-07-30 | 2021-05-25 | 深圳大学 | Method, system, equipment and medium for predicting generation amount of building demolition waste |
CN109034073A (en) * | 2018-07-30 | 2018-12-18 | 深圳大学 | Predict method, system, equipment and the medium of building demolition waste yield |
CN109345454A (en) * | 2018-09-18 | 2019-02-15 | 徐庆 | Method, storage medium and the system of bitmap images vector quantization |
CN109345454B (en) * | 2018-09-18 | 2023-01-06 | 徐庆 | Bitmap image vectorization method, storage medium and system |
CN110619282A (en) * | 2019-08-26 | 2019-12-27 | 海南撰云空间信息技术有限公司 | Automatic extraction method for unmanned aerial vehicle orthoscopic image building |
CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
CN110674735B (en) * | 2019-09-23 | 2022-05-24 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
CN111523391A (en) * | 2020-03-26 | 2020-08-11 | 上海刻羽信息科技有限公司 | Building identification method, system, electronic device and readable storage medium |
CN111523391B (en) * | 2020-03-26 | 2021-11-05 | 上海刻羽信息科技有限公司 | Building identification method, system, electronic device and readable storage medium |
CN111652892A (en) * | 2020-05-02 | 2020-09-11 | 王磊 | Remote sensing image building vector extraction and optimization method based on deep learning |
CN114581786B (en) * | 2021-12-28 | 2022-11-25 | 深圳市城市产业发展集团有限公司 | Method and device for estimating building area according to ground image |
CN114581786A (en) * | 2021-12-28 | 2022-06-03 | 深圳市城市产业发展集团有限公司 | Method and device for estimating building area according to ground image |
CN114998664A (en) * | 2022-07-18 | 2022-09-02 | 中国科学院烟台海岸带研究所 | Rapid detection method and device for micro-plastic in seawater by multiple optical platforms |
Also Published As
Publication number | Publication date |
---|---|
CN103699900B (en) | 2016-10-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103699900A (en) | Automatic batch extraction method for horizontal vector contour of building in satellite image | |
CN102968782B (en) | In a kind of coloured image, remarkable object takes method automatically | |
CN107862667B (en) | Urban shadow detection and removal method based on high-resolution remote sensing image | |
CN110232389B (en) | Stereoscopic vision navigation method based on invariance of green crop feature extraction | |
Chen et al. | An improved license plate location method based on edge detection | |
CN103186904B (en) | Picture contour extraction method and device | |
CN103559500B (en) | A kind of multi-spectral remote sensing image terrain classification method based on spectrum Yu textural characteristics | |
Chen et al. | A novel color edge detection algorithm in RGB color space | |
WO2015010451A1 (en) | Method for road detection from one image | |
CN106651872A (en) | Prewitt operator-based pavement crack recognition method and system | |
CN102609917B (en) | Image edge fitting B spline generating method based on clustering algorithm | |
CN105631455A (en) | Image main body extraction method and system | |
CN102800094A (en) | Fast color image segmentation method | |
CN107154026B (en) | Method for eliminating road surface shadow based on self-adaptive brightness elevation model | |
CN103984946A (en) | High resolution remote sensing map road extraction method based on K-means | |
CN103208123A (en) | Image segmentation method and system | |
CN102147867B (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
CN108038416A (en) | Method for detecting lane lines and system | |
CN104299263A (en) | Method for modeling cloud scene based on single image | |
CN103824309A (en) | Automatic extracting method of urban built-up area border | |
CN104966285A (en) | Method for detecting saliency regions | |
CN104809709A (en) | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition | |
CN102663760A (en) | Location and segmentation method for windshield area of vehicle in images | |
CN103198479A (en) | SAR image segmentation method based on semantic information classification | |
CN102930538B (en) | The self-adaptive projection method method that Gauss potential and spatial histogram merge |
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 |