CN108197583A - The building change detecting method of optimization and image structure feature is cut based on figure - Google Patents
The building change detecting method of optimization and image structure feature is cut based on figure Download PDFInfo
- Publication number
- CN108197583A CN108197583A CN201810023880.2A CN201810023880A CN108197583A CN 108197583 A CN108197583 A CN 108197583A CN 201810023880 A CN201810023880 A CN 201810023880A CN 108197583 A CN108197583 A CN 108197583A
- Authority
- CN
- China
- Prior art keywords
- building
- dsm
- feature
- image
- variation
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The invention discloses a kind of building change detecting method that optimization and image structure feature are cut based on figure, including：S1, DSM filtering obtain DEM, calculate normalization digital surface model (nDSM)；S2, calculating difference DSM (dDSM)；Feature changes test problems are modeled as binaryzation classification problem by S3, extraction feature changes region, and feature changes region is prospect, other regions are background；S4, it is proposed that a kind of steady image structure feature effectively excludes the non-building areas domains such as the vegetation of variation；S5 changes testing result and altitude data with reference to building, and the building object of variation is further discriminated between to increase, reduces, removes, create four classes.Present invention comprehensive utilization three-dimensional information and image spectral information, the precision and reliability of building variation detection can be significantly improved, is of great significance to fields such as urban planning, city dynamic monitoring, city growth detection, the identification of architecture against regulations object and geographical information updatings.
Description
Technical field
The invention belongs to building variation detection field more particularly to a kind of optimization and image structure feature are cut based on figure
Building change detecting method.
Background technology
One of the important content of building variation detection as geographical national conditions monitoring, moves the identification of architecture against regulations object, city
State monitoring and geographical information updating etc. are of great significance.By taking the architecture against regulations analyte detection of city as an example, with economic society of China
It can continue to develop, urbanization process is constantly accelerated, and City Building constantly increases, and the quantity and scale of illegal building object is not yet
Disconnected to increase, this phenomenon not only destroys urban planning and urban landscape, but also influences city image and resident living, is both common people pass
The hot issue of note, and be the difficulties of city management, even more influence one of socially harmonious negative factor.At present, it is " illegal
At low cost, executive cost is high " it is one of the main reason for illegal building object remains incessant after repeated prohibition, other than relevant law link is short of,
It is also weaker to the context of detection of illegal building object, due to lacking the monitoring means of automation to illegal building object, utilize people
The mode of work inspection has that all kinds of corrupt practices creep in, first, the discovery procedure period is longer, second is that monitoring cost is high on a large scale.In recent years
The cities such as Beijing have been attempted using satellite image data progress illegal building analyte detection, but the automatic analysis technology of image information
Still not mature enough, the ratio that manual identified and verification participate in flow is great.The Executing The Law While Managing Land department in the annual whole nation, city management
Department is billions of for the manpower and materials input of this task.There is an urgent need to a kind of high degree of automations, steady in the market
Reliable method carries out the detection of city architecture against regulations object, so as to push the regulation process of city illegal building object.
Currently, building changes the research of detection automatically can be classified as the following aspects according to data source：It is first, single
The building variation detection of pure spectrum information[1-9], it is limited to ambiguousness of the spectral information in terms of atural object interpretation and current
Machine vision intelligent level, current algorithm are still left to be desired in terms of versatility and reliability；Second is that based on LiDAR point cloud
Building variation detection[10-12], LiDAR technologies can obtain precision height, highly reliable three dimensional point cloud, with reference to synchronization
The spectral information of image can obtain the building variation testing result that precision is high, highly reliable, but LiDAR data cost at present
It is higher, and data popularity rate is low, especially lacks the LiDAR data of an old phase, therefore application range is relatively more limited；Third, based on vertical
The building variation detection of body image data[13-17], have benefited from the progress of image dense Stereo Matching technology in recent years, pass through three-dimensional shadow
As data can extract the high earth's surface three-dimensional information of density, the spectrum of three-dimensional information and corresponding image for passing through comprehensive two phases is believed
Breath can obtain the building change information of high-precision and reliability, but current variation testing result is by erroneous matching result
Image is larger, and the utilization of spectral information is abundant not enough.
Relevant references are as follows：
[1]Walter,V.Object-based classification of remote sensing data for
change detection.ISPRS Journal of photogrammetry and remote sensing 2004,58,
225-238.
[2] season Shunping County；Yuan repaiies a kind of building change detecting method remote sensing journals 2007 based on shadow Detection of filial piety,
323-329.
[3]Pacifici,F.；Del Frate,F.；Solimini,C.；Emery,W.J.An innovative
neural-net method to detect temporal changes in high-resolution optical
satellite imagery.IEEE Transactions on Geoscience and Remote Sensing 2007,45,
2940-2952.
[4]Bouziani,M.；K.；He,D.-C.Automatic change detection of
buildings in urban environment from very high spatial resolution images using
existing geodatabase and prior knowledge.ISPRS Journal of Photogrammetry and
Remote Sensing 2010,65,143-153.
[5]Vu,T.T.；Ban,Y.Context-based mapping of damaged buildings from
high-resolution optical satellite images.International Journal of Remote
Sensing 2010,31,3411-3425.
[6]Li,P.；Xu,H.；Guo,J.Urban building damage detection from very high
resolution imagery using ocsvm and spatial features.International Journal of
Remote Sensing 2010,31,3393-3409.
[7]Tang,Y.；Huang,X.；Zhang,L.Fault-tolerant building change detection
from urban high-resolution remote sensing imagery.IEEE Geoscience and Remote
Sensing Letters 2013,10,1060-1064.
[8]Huang,X.；Zhang,L.；Zhu,T.Building change detection from
multitemporal high-resolution remotely sensed images based on a morphological
building index.IEEE Journal of Selected Topics in Applied Earth Observations
and Remote Sensing2014,7,105-115.
[9]Xiao P.,Yuan M.,Zhang X.,Feng X.,Guo Y.Cosegmentation for object-
based building change detection from high resolution remotely sensed
images.IEEE Transactions on Geoscience and Remote Sensing,2017,55,1587-1603.
[10]Vu T T,Matsuoka M,Yamazaki F.LIDAR-based change detection of
buildings in dense urban areas[C].Proceedings of the Geoscience and Remote
Sensing Symposium(IGARSS'04),2004.
[11]Teo T-A,Shih T-Y.Lidar-based change detection and change-type
determination in urban areas[J].International Journal of Remote Sensing,2013,
34(3):968-981.
[12]Pang,S.；Hu,X.；Wang,Z.；Lu,Y.Object-based analysis of airborne
lidar data for building change detection.Remote Sensing 2014,6,10733-10749.
[13]Qin R.An Object-Based Hierarchical Method for Change Detection
Using Unmanned Aerial Vehicle Images[J].Remote Sensing,2014,6(9):7911-7932.
[14]Tian J,Reinartz P,D’Angelo P,et al.Region-based automatic
building and forest change detection on Cartosat-1stereo imagery[J].ISPRS
Journal of Photogrammetry and Remote Sensing,2013,79:226-239.
[15]Tian J,Cui S,Reinartz P.Building change detection based on
satellite stereo imagery and digital surface models[J].IEEE Transactions on
Geoscience and Remote Sensing,2014,52(1):406-417.
[16]Qin R J.Change detection on LOD 2building models with very high
resolution spaceborne stereo imagery[J].Isprs Journal of Photogrammetry and
Remote Sensing,2014,96:179-192.
[17]Jung F.Detecting building changes from multitemporal aerial
stereopairs[J].ISPRS Journal of Photogrammetry and Remote Sensing,2004,58(3):
187-201.
Invention content
In view of the deficienciess of the prior art, three-dimensional information and image can be comprehensively utilized the object of the present invention is to provide a kind of
The building change detecting method of spectral information cuts algorithm optimization by figure and changes testing result, while make full use of image light
Spectrum information extracts image structure feature, excludes the non-construction zone of variation.
In order to achieve the above object, technical solution provided by the invention is：Optimization and image structure feature are cut based on figure
Building change detecting method, includes the following steps,
Step 1, the DSM data of two phases is filtered respectively using the progressive filtering method of the triangulation network, obtained for two phases
DEM, then rasterizing correspond to the DSM and DEM in period and make difference operation, obtain two phases normalization digital surface model
nDSM；
Step 2, difference operation is carried out to two phase DSM in step 1, obtains difference DSM, be denoted as dDSM；
Step 3, for each issue evidence, feature changes test problems is modeled as figure and cut binaryzation problem, feature changes
Region is prospect, other regions are background, and remove noise fine crushing using region growing algorithm, obtains variation earth object, i.e.,
Candidate change building object；
Step 4, according to the internal and external orientation information of image, the corresponding image block number of candidate change building object is obtained
According to carrying out structure feature extraction to image blocks, image blocks further divided into building and non-according to structure feature statistical value
Building excludes the influence of non-architectural object；
Step 5, change testing result and DSM data with reference to two phase buildings, by the further area of building object of variation
It is divided into and increases, reduce, remove, creates four classes.
Further, feature changes test problems are modeled as figure in step 3 and cut binaryzation problem, feature changes region is
Prospect, other regions are background, and realization method is as follows,
Energy function E is defined, includes data item ENERGY EdataWith smooth item ENERGY Esmooth,
E=Edata+Esmooth (1)
Wherein EdataRepresent the data item cost summation of all the points, value is determined that P represents all the points by nDSM and dDSM
Set, N represent the set of the point pair of consecutive points composition；
D in formulap(lp) represent p points data item cost, ff(nDSMp,dDSMp) it is the classification l for working as p pointspBelong to prospect
(fg) cost function when, fb(nDSMp,dDSMp) it is the classification l for working as p pointspBelong to cost function during background (bg)；TnDSMIt is
NDSM belongs to the empirical value of building, TdDSM1And TdDSM2It is two threshold values of dDSM, T2It is the larger energy cost threshold of value
Value；
Wherein, EsmoothIt is worth and is determined by the DSM differences between consecutive points,
DLen=fabs (DSMp-DSMq) (8)
Fabs expressions take absolute value in formula, and dLen represents the absolute value of the DSM differences between adjacent 2 points p and q；TdLen1With
TdLen2It is two threshold values；T2It is the larger energy cost threshold value of value, with the T in data item2Value is identical；
Using max-flow min-cut principle solving energy function, minimal cut is obtained；Finally, foreground point is obtained according to minimal cut
And background dot, prospect correspond to atural object region of variation.
Further, noise fine crushing is removed using region growing algorithm in step 3, obtains the realization side of variation earth object
Formula is as follows,
Region growth is carried out according to DSM differences to foreground point, if the DSM differences between consecutive points are less than threshold value, is increased
Otherwise a length of same object-point is another object-point, repeat the process until all foreground points judge to terminate；Most
The thin objects according to caused by matching area threshold debug afterwards, obtain candidate change building object.
Further, structure feature extraction is carried out to image blocks in step 4, according to structure feature statistical value further by shadow
As the realization method that block divides into building and non-building is as follows,
1. line feature extraction and simplification：Edge extracting is carried out using Boundary extracting algorithm to image blocks, then in edge spy
The linear feature of image is extracted on the basis of sign extraction, and line feature is simplified using Douglas-Peucker algorithms；
2. structural texture feature, i.e. edge gradient direction histogram：For each straightway, its length and direction is calculated
Dir, wherein direction are in the range of [0,180], then by certain step pitch by direction demarcation interval, calculate structure feature；
The wherein each interval statistics value calculation formula of direction histogram is as follows：
Wherein PjIt is the corresponding probability values of section j, m is all straightways for belonging to section j, n representative images institute in the block
Some straightways, qiIt is the weight of i straightways, corresponding to its length, i.e. straightway number of pixels；
3. calculating structure feature characterising parameter, including up rightness and long linearity, obtained by edge gradient histogram, specifically
It is as follows：
Up rightness：Some threshold value T is set, if PjMore than T, and 90 degree are differed between section i and section j, then recognized
It is portrayed with vertical section number N for there are up rightness, existing up rightness in image block, N ∈ [0,1,2 ..., 9]；
Long linearity：For region existing for building, usually there are long linearity, using P0Value is measured；
In summary two kinds of structure features have：A=(1+N) P0；When A is more than a certain empirical value T1When, it is believed that image block
It is otherwise non-building for building.
Further, the realization method of step 5 is as follows,
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and
DSMt1<DSMt2, then the object type is increases；
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and
DSMt1>DSMt2, then the object type is reduces；
For each variation object, when previous phase classification results are building, latter phase classification results are non-building or ground
Shape, then the object type is removes；
For each variation object, when previous phase classification results are non-building or landform, latter phase classification results is build
Object is built, then the object type is newly-built.
It is of the invention mainly to change detection automatically using dense Stereo Matching point cloud and image spectral information information progress building, it is close
Collect match point cloud and be originated from dense Stereo Matching technology, it is intensive can to obtain for two phases automatically by two phase stereoscopic image datas being registrated in advance
Point cloud data is matched, the dense Stereo Matching technology of current maturation, such as half global optimization dense Stereo Matching may be used in terms of dense Stereo Matching,
Image spectral information refers mainly to the corresponding raw video data of two phase dense Stereo Matching point clouds.By comprehensively utilizing two kinds of information respectively
Advantage come improve building variation detection precision and reliability.
Compared with the existing methods, the method for the present invention takes full advantage of the three-dimensional information of match point cloud and the extraction of image spectrum
Structural information, improve building variation detection precision and reliability, have following features：
1st, in terms of object extraction is changed.The influence of variation landforms is excluded using nDSM information, optimisation strategy is cut using figure
The integrality of extracting object is maintained, feature changes object is obtained using region growing algorithm with reference to classification information and DSM differences,
The influence of discreet region caused by eliminating erroneous matching etc.；
2nd, in terms of non-building object exclusion.A kind of steady image structure feature is devised, this feature does not need to be additional
Training sample, precision is high, adaptable.
3rd, the present invention improves the precision and reliability of building variation detection, maintains the complete of building variation object
Property, to city dynamic monitoring, architecture against regulations identification and geographical information updating etc. are of great significance.
Description of the drawings
Fig. 1 is particular flow sheet of the embodiment of the present invention；
Fig. 2 is dense Stereo Matching point cloud data and point cloud filter result used by the embodiment of the present invention；
Fig. 3 is to extract schematic diagram based on the alternatively object area that figure is cut in the embodiment of the present invention；
Fig. 4 be the embodiment of the present invention in change atural object object extraction schematic diagram, figure (a) be figure cut as a result, scheme (b) for based on
The region of classification and DSM differences increases, and figure (c) is the feature changes object removed after small object；
Fig. 5 is image structure feature differentiation building and non-building schematic diagram in the embodiment of the present invention；
Fig. 6 be the embodiment of the present invention in Experimental Area sketch map, wherein, figure (a) (c) be respectively old phase orthography and
Dense Stereo Matching point cloud, b) (d) is respectively new phase orthography and dense Stereo Matching point cloud to figure；
Fig. 7 is that building changes testing result figure in the embodiment of the present invention, and figure (a) and (b) are respectively an old phase and new one
The figure of phase cuts optimum results, and figure (c) and (d) are that earth object extraction distinguishes building as a result, figure (e) and (f) are structure feature
With non-building as a result, figure (g) is final fusion results.
Specific embodiment
The present invention carries out building using dense Stereo Matching point cloud and image spectral information and changes detection automatically, first using existing
There are the progressive filtering method of the triangulation network (Axelsson P.DEM generation from laser scanner data using
adaptive TIN models[C].International Archives of Photogrammetry&Remote
Sensing, 2000.) acquisition digital elevation model (DEM) is filtered to dense Stereo Matching point cloud data, calculate normalization
Digital surface model (nDSM) excludes the interference of variation geomorphic province, and then cutting optimisation strategy by figure maintains extraction variation
The integrality of earth object analyzes the variation earth object of extraction finally by the image structure feature of design, excludes
The non-architectural objects such as variation vegetation object therein obtain final building variation object.
Technical solution for a better understanding of the present invention is below in conjunction with the accompanying drawings the present invention further specifically
Bright, Fig. 1 is the overview flow chart of the present invention.The present invention is as follows：
Step 1, DSM is filtered using the triangulation network progressive filtering method, obtains digital elevation model (DEM),
Rasterizing DSM and DEM simultaneously makees difference operation acquisition normalization digital surface model (nDSM)；Respectively to the DSM data of two phases into
Row aforesaid operations obtain nDSMt2And nDSMt1, t1 and t2 represent two different times, dense Stereo Matching point cloud and dense Stereo Matching point
Cloud filter result can refer to Fig. 2；
Step 2, by DSMt2And DSMt1Difference operation is carried out, obtains difference DSM (dDSM)；
Step 3, for each issue evidence, feature changes test problems is modeled as figure and cut binaryzation problem, feature changes
Region is prospect, other regions are background, and using a small amount of noise fine crushing of region growing algorithm removal, obtains variation atural object pair
As i.e. candidate change building object.
First, feature changes test problems are modeled as binaryzation classification problem, the ground object area conduct of variation by the present invention
Prospect, other regions are background.The solution of feature changes test problems is cut energy-optimised frame using figure and is realized, energy mainly wraps
Containing data item and smooth item, as shown in formula (1).The feature changes detection process entirely cut based on figure as shown in schematic diagram Fig. 3,
From figure 3, it can be seen that each point has 2 possible classifications (foreground and background), each point belongs to prospect and belongs to background
Probability is data item, value such as formula (4), and the flatness between consecutive points and point is then calculated using formula (7).It defines
After data item and the value of smooth item, the solution of this problem is gone out using max-flow min-cut principle solving, corresponding to the minimum in figure
It cuts.Finally, the foreground and background point in figure is obtained according to minimal cut, prospect corresponds to the feature changes region in the present invention.
E=Edata+Esmooth (1)
Wherein EdataIt represents data item energy, represents the data item cost summation of all the points, EsmoothRepresent smooth item energy
Amount；P represents the set of all the points, corresponding to the set of all the points in Fig. 3；N represents the set of the point pair of consecutive points composition, right
It should be in the side line set in Fig. 3.EdataValue determine that dDSM can extract region of variation by nDSM and dDSM, nDSM rows
Except influence caused by variation landform, specific formula for calculation is as follows：
Wherein Dp(lp) represent p points data item cost, ff(nDSMp,dDSMp) it is the classification l for working as p pointspBelong to prospect
(fg) cost function when, fb(nDSMp,dDSMp) it is the classification l for working as p pointspBelong to cost function during background (bg).TnDSMIt is
NDSM belongs to the empirical value of building, this is defined herein as 2.2 meters.TdDSM1And TdDSM2It is two threshold values of dDSM, value here
For 1 meter and 2 meters.T2It is the larger energy cost threshold value of value, this is defined herein as 20.
EsmoothRepresent smooth item energy, its value is determined by the DSM differences between consecutive points, is used primarily to ensure extraction
The integrality of object, specific formula for calculation are as follows：
DLen=fabs (DSMp-DSMq) (8)
Wherein, fabs expressions take absolute value, and dLen represents the absolute value of the DSM differences between adjacent 2 points p and q, when 2 points
Between difference it is smaller, then flatness cost is smaller, and difference is bigger, and flatness cost is bigger.TdLen1And TdLen2It is two threshold values,
Distinguish value 0.1 and 0.5 meter.T2It is the larger energy cost threshold value of value, it is identical with formula (3) value.
Whole region propagation process is as shown in figure 4, after the feature changes extracted region cut based on figure, feature changes area
Domain is prospect, other regions are background, and as shown in Fig. 4 (a), black color dots are feature changes point, and white point is non-feature changes point.
In order to obtain feature changes object, region growth is carried out according to DSM differences to foreground point, as shown in Fig. 4 (b), if consecutive points
Between DSM differences be less than threshold value (threshold value be DSM tolerance elevation threshold value, general value 0.3-0.5 meters, in the present embodiment
Value is 0.5 meter), then it is same object-point to increase, and is otherwise another object-point, repeats this process before all
Until sight spot judges to terminate.The finally thin objects according to caused by area threshold debug matching etc., obtain candidate change and build
Object object is built, as shown in Fig. 4 (c), area threshold represents the minimum area of detection variation building, general 50-100 square metres of value
Between, 50 square metres of value in the present embodiment.
Step 4, according to the internal and external orientation information of image, the corresponding image block number of candidate change building object is obtained
According to carrying out structure feature extraction to image blocks, image blocks further divided into building and non-according to structure feature statistical value
Building excludes the influence of the non-architectural objects such as vegetation；
In order to improve the accuracy that the non-architectural object such as building and vegetation is distinguished, candidate change building object is being obtained
On the basis of, the present invention devises a kind of image structure feature based on edge gradient direction histogram, and this feature is by image blocks
Extraction can effectively describe image texture information, the reliability that the two is distinguished be improved, as shown in figure 5, this method is extracted first
Then edge feature in image carries out the extraction and simplification of edge line, finally calculate image structure feature description parameter, and root
It is worth according to statistics and building and vegetation object is distinguished.
This method is as follows:
1. line feature extraction and simplification.The corresponding area image of the object is extracted according to the range of feature changes object, it is right
Area image block carries out edge extracting using Boundary extracting algorithm (such as canny Boundary extracting algorithms), is then carried in edge feature
The linear feature of image is extracted on the basis of taking, and line feature is simplified using Douglas-Peucker algorithms.
2. structural texture feature, i.e. edge gradient direction histogram.For each straightway, its length and direction is calculated
Dir, wherein direction are in the range of [0,180], then by certain step pitch (taking 10 degree here) by direction demarcation interval, calculate
Direction histogram, i.e. structure feature.In view of generally having apparent line feature around building, in order to enhance the stabilization of algorithm
Property, weight is increased according to the length of straightway when calculating direction histogram.
The wherein each interval statistics value calculation formula of direction histogram is as follows：
Wherein PjIt is the corresponding probability values of section j, m is all straightways for belonging to section j, n representative images institute in the block
Some straightways, qiIt is the weight of i straightways, corresponding to its length, i.e. straightway number of pixels.It is general in order to incite somebody to action
Rate normalizes, and normalized purpose is to inhibit the phenomenon that those a large amount of short straight line sections, PjValue is bigger, general existing for long straightway
Rate is bigger.
3. calculate structure feature characterising parameter.Here structure feature mainly includes up rightness and long linearity, by edge
Histogram of gradients obtains, specific as follows：
Up rightness：Some threshold value T is set, if PjMore than T, and 90 degree are differed between section i and section j, then recognized
It is portrayed with vertical section number N for there are up rightness, existing up rightness in image block, N ∈ [0,1,2 ..., 9].
Long linearity：For region existing for building, usually there are long linearity, mainly use P here0Value is measured, P0
Value obtains when taking 0 by the j in formula (9).
In summary two kinds of structure features have：A=(1+N) P0；When A is more than a certain empirical value T1When, it is believed that image block
It is otherwise non-building for building.
Step 5, change testing result and DSM data with reference to two phase buildings, by the further area of building object of variation
It is divided into and increases, reduce, remove, creates four classes, it is specific as follows；
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and
DSMt1<DSMt2, then the object type is increases, wherein DSMt1For previous phase DSM data, DSMt2For latter phase DSM data；
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and
DSMt1>DSMt2, then the object type is reduces, wherein DSMt1For previous phase DSM data, DSMt2For latter phase DSM data；
For each variation object, when previous phase classification results are building, latter phase classification results are non-building or ground
Shape, then the object type is removes；
For each variation object, when previous phase classification results are non-building or landform, latter phase classification results is build
Object is built, then the object type is newly-built；
For each variation object, when previous phase classification results are non-building or landform, latter phase classification results are also
Non- building or landform, then the object type is not change, will be removed, without label；
As shown in table 1, wherein not changing here refers to that there are building variations.
1 building change type of table determines
Fig. 6 is Experimental Area overview diagram in the embodiment of the present invention, wherein figure (a) and two phases that (b) is Experimental Area just penetrate
Image, figure (c) and the elevation that (d) is two phase dense Stereo Matching point clouds render figure；Fig. 7 changes testing result for building, wherein scheming
(a) and (b) is respectively that the figure of an old and new phase cuts optimum results, figure (c) and (d) extracted for earth object as a result, scheme (e) with
(f) it is that structure feature distinguishes building with non-building as a result, scheming (g) for final fusion results, it can be clear from figure
Find out the type of building variation.
Specific embodiment described herein is only an example for the spirit of the invention.Technology belonging to the present invention is led
The technical staff in domain can do various modifications or additions to described specific embodiment or replace in a similar way
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (5)
1. the building change detecting method of optimization and image structure feature is cut based on figure, which is characterized in that include the following steps：
Step 1, the DSM data of two phases is filtered respectively using the progressive filtering method of the triangulation network, obtained for two phases
DEM, then rasterizing correspond to the DSM and DEM in period and make difference operation, obtain two phases normalization digital surface model
nDSM；
Step 2, difference operation is carried out to two phase DSM in step 1, obtains difference DSM, be denoted as dDSM；
Step 3, for each issue evidence, feature changes test problems is modeled as figure and cut binaryzation problem, feature changes region
For prospect, other regions are background, and remove noise fine crushing using region growing algorithm, obtain variation earth object, i.e., candidate
Change building object；
Step 4, according to the internal and external orientation information of image, the corresponding image blocks data of candidate change building object are obtained,
Structure feature extraction is carried out to image blocks, image blocks are further divided by building and non-building according to structure feature statistical value
Object excludes the influence of non-architectural object；
Step 5, change testing result and DSM data with reference to two phase buildings, by the building object of variation further discriminate between for
Increase, reduce, remove, create four classes.
2. the building change detecting method of optimization and image structure feature, feature are cut based on figure as described in claim 1
It is：Feature changes test problems are modeled as figure in step 3 and cut binaryzation problem, feature changes region is prospect, other areas
Domain is background, and realization method is as follows,
Energy function E is defined, includes data item ENERGY EdataWith smooth item ENERGY Esmooth,
E=Edata+Esmooth (1)
Wherein EdataRepresenting the data item cost summation of all the points, value is determined that P represents the set of all the points by nDSM and dDSM,
N represents the set of the point pair of consecutive points composition；
D in formulap(lp) represent p points data item cost, ff(nDSMp,dDSMp) it is the classification l for working as p pointspWhen belonging to prospect (fg)
Cost function, fb(nDSMp,dDSMp) it is the classification l for working as p pointspBelong to cost function during background (bg)；TnDSMIt is that nDSM belongs to
In the empirical value of building, TdDSM1And TdDSM2It is two threshold values of dDSM, T2It is the larger energy cost threshold value of value；
Wherein, EsmoothIt is worth and is determined by the DSM differences between consecutive points,
DLen=fabs (DSMp-DSMq) (8)
Fabs expressions take absolute value in formula, and dLen represents the absolute value of the DSM differences between adjacent 2 points p and q；TdLen1And TdLen2It is
Two threshold values,；T2It is the larger energy cost threshold value of value, with the T in data item2Value is identical；
Using max-flow min-cut principle solving energy function, minimal cut is obtained；Finally, foreground point and the back of the body are obtained according to minimal cut
Sight spot, prospect correspond to atural object region of variation.
3. the building change detecting method of optimization and image structure feature, feature are cut based on figure as claimed in claim 2
It is：Noise fine crushing is removed using region growing algorithm in step 3, the realization method for obtaining variation earth object is as follows,
Region growth is carried out according to DSM differences to foreground point, if the DSM differences between consecutive points are less than threshold value, growth is
Otherwise same object-point is another object-point, repeat the process until all foreground points judge to terminate；Last root
Thin objects caused by being matched according to area threshold debug obtain candidate change building object.
4. the building change detecting method that optimization and image structure feature are cut based on figure as described in claims 1 or 2 or 3,
It is characterized in that：Structure feature extraction is carried out to image blocks in step 4, according to structure feature statistical value further by image blocks area
The realization method for being divided into building and non-building is as follows,
1. line feature extraction and simplification：Edge extracting is carried out using Boundary extracting algorithm to image blocks, is then carried in edge feature
The linear feature of image is extracted on the basis of taking, and line feature is simplified using Douglas-Peucker algorithms；
2. structural texture feature, i.e. edge gradient direction histogram：For each straightway, its length and direction Dir is calculated,
Wherein direction is in the range of [0,180], then by certain step pitch by direction demarcation interval, calculates structure feature；
The wherein each interval statistics value calculation formula of direction histogram is as follows：
Wherein PjIt is the corresponding probability values of section j, m is all straightways for belonging to section j, and n representative images are in the block all straight
Line segment, qiIt is the weight of i straightways, corresponding to its length, i.e. straightway number of pixels；
3. calculating structure feature characterising parameter, including up rightness and long linearity, obtained by edge gradient histogram, specifically such as
Under：
Up rightness：Some threshold value T is set, if PjMore than T, and 90 degree are differed between section i and section j, then it is assumed that image
There are up rightness, existing up rightness in block to be portrayed with vertical section number N, N ∈ [0,1,2 ..., 9]；
Long linearity：For region existing for building, usually there are long linearity, using P0Value is measured；
In summary two kinds of structure features have：A=(1+N) P0；When A is more than a certain empirical value T1When, it is believed that image block is builds
It builds, is otherwise non-building.
5. the building change detecting method of optimization and image structure feature, feature are cut based on figure as claimed in claim 4
It is：The realization method of step 5 is as follows,
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and DSMt1
<DSMt2, then the object type is increases；
For each variation object, when previous phase classification results are building, latter phase classification results are also building, and DSMt1
>DSMt2, then the object type is reduces；
For each variation object, when previous phase classification results are building, latter phase classification results are non-building or landform, then
The object type is removes；
For each variation object, when previous phase classification results are non-building or landform, latter phase classification results are building,
Then the object type is newly-built.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023880.2A CN108197583B (en) | 2018-01-10 | 2018-01-10 | Building change detection method based on graph cut optimization and image structure characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023880.2A CN108197583B (en) | 2018-01-10 | 2018-01-10 | Building change detection method based on graph cut optimization and image structure characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108197583A true CN108197583A (en) | 2018-06-22 |
| CN108197583B CN108197583B (en) | 2020-04-24 |
Family
ID=62588609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810023880.2A Active CN108197583B (en) | 2018-01-10 | 2018-01-10 | Building change detection method based on graph cut optimization and image structure characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108197583B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035206A (en) * | 2018-06-28 | 2018-12-18 | 中国地震局地震预测研究所 | A kind of building collapse state detection method |
| CN111222453A (en) * | 2020-01-03 | 2020-06-02 | 武汉大学 | Remote sensing image change detection method based on dense connection and geometric structure constraint |
| CN111323788A (en) * | 2020-01-19 | 2020-06-23 | 北京建筑大学 | Building change monitoring method and device and computer equipment |
| CN111582156A (en) * | 2020-05-07 | 2020-08-25 | 武汉大势智慧科技有限公司 | Oblique photography-based tall and big vegetation extraction method for urban three-dimensional model |
| CN111814715A (en) * | 2020-07-16 | 2020-10-23 | 武汉大势智慧科技有限公司 | Ground object classification method and device |
| CN111986099A (en) * | 2020-06-30 | 2020-11-24 | 武汉大学 | Tillage monitoring method and system based on convolutional neural network with residual error correction fused |
| CN112183434A (en) * | 2020-10-12 | 2021-01-05 | 南方电网能源发展研究院有限责任公司 | Building change detection method and device |
| CN112580137A (en) * | 2020-12-21 | 2021-03-30 | 北京市测绘设计研究院 | Rapid calculation method for scale of house building |
| US11367265B2 (en) * | 2020-10-15 | 2022-06-21 | Cape Analytics, Inc. | Method and system for automated debris detection |
| US11568639B2 (en) | 2015-08-31 | 2023-01-31 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| US11861843B2 (en) | 2022-01-19 | 2024-01-02 | Cape Analytics, Inc. | System and method for object analysis |
| US11875413B2 (en) | 2021-07-06 | 2024-01-16 | Cape Analytics, Inc. | System and method for property condition analysis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093466A (en) * | 2013-01-21 | 2013-05-08 | 武汉大学 | Building three-dimensional change detection method based on LiDAR point cloud and image |
| CN104049245A (en) * | 2014-06-13 | 2014-09-17 | 中原智慧城市设计研究院有限公司 | Urban building change detection method based on LiDAR point cloud spatial difference analysis |
-
2018
- 2018-01-10 CN CN201810023880.2A patent/CN108197583B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093466A (en) * | 2013-01-21 | 2013-05-08 | 武汉大学 | Building three-dimensional change detection method based on LiDAR point cloud and image |
| CN104049245A (en) * | 2014-06-13 | 2014-09-17 | 中原智慧城市设计研究院有限公司 | Urban building change detection method based on LiDAR point cloud spatial difference analysis |
Non-Patent Citations (6)
| Title |
|---|
| J. TIAN 等: ""Region-based automatic building and forest change detection on Cartosat-1"", 《ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING》 * |
| SHIYAN PANG 等: ""Object-Based Analysis of Airborne LiDAR Data for Building Change Detection"", 《REMOTE SENSING》 * |
| XIANGYUN HU 等: ""Semi-Global Filtering of Airborne LiDAR Data for Fast Extraction of Digital"", 《REMOTE SENSING》 * |
| 刘直芳 等: ""基于DSM和影像特征的城市变化检测"", 《遥感技术与应用》 * |
| 刘莹 等: ""图割能量驱动的高分辨率遥感影像震害损毁建筑物检测"", 《测绘学报》 * |
| 杨晓强: ""基于边缘梯度方向直方图的图像检索方法"", 《微计算机信息》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11568639B2 (en) | 2015-08-31 | 2023-01-31 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| CN109035206A (en) * | 2018-06-28 | 2018-12-18 | 中国地震局地震预测研究所 | A kind of building collapse state detection method |
| CN109035206B (en) * | 2018-06-28 | 2020-10-09 | 中国地震局地震预测研究所 | Building damage state detection method |
| CN111222453A (en) * | 2020-01-03 | 2020-06-02 | 武汉大学 | Remote sensing image change detection method based on dense connection and geometric structure constraint |
| CN111222453B (en) * | 2020-01-03 | 2022-06-14 | 武汉大学 | Remote sensing image change detection method based on dense connection and geometric structure constraint |
| CN111323788A (en) * | 2020-01-19 | 2020-06-23 | 北京建筑大学 | Building change monitoring method and device and computer equipment |
| CN111323788B (en) * | 2020-01-19 | 2022-04-15 | 北京建筑大学 | Building change monitoring method and device and computer equipment |
| CN111582156A (en) * | 2020-05-07 | 2020-08-25 | 武汉大势智慧科技有限公司 | Oblique photography-based tall and big vegetation extraction method for urban three-dimensional model |
| CN111582156B (en) * | 2020-05-07 | 2023-12-05 | 武汉大势智慧科技有限公司 | High and large vegetation extraction method based on oblique photography urban three-dimensional model |
| CN111986099A (en) * | 2020-06-30 | 2020-11-24 | 武汉大学 | Tillage monitoring method and system based on convolutional neural network with residual error correction fused |
| CN111814715A (en) * | 2020-07-16 | 2020-10-23 | 武汉大势智慧科技有限公司 | Ground object classification method and device |
| CN111814715B (en) * | 2020-07-16 | 2023-07-21 | 武汉大势智慧科技有限公司 | Ground object classification method and device |
| CN112183434B (en) * | 2020-10-12 | 2022-06-24 | 南方电网能源发展研究院有限责任公司 | Building change detection method and device |
| CN112183434A (en) * | 2020-10-12 | 2021-01-05 | 南方电网能源发展研究院有限责任公司 | Building change detection method and device |
| US11367265B2 (en) * | 2020-10-15 | 2022-06-21 | Cape Analytics, Inc. | Method and system for automated debris detection |
| CN112580137B (en) * | 2020-12-21 | 2023-06-06 | 北京市测绘设计研究院 | Rapid calculation method for house building scale |
| CN112580137A (en) * | 2020-12-21 | 2021-03-30 | 北京市测绘设计研究院 | Rapid calculation method for scale of house building |
| US11875413B2 (en) | 2021-07-06 | 2024-01-16 | Cape Analytics, Inc. | System and method for property condition analysis |
| US11861843B2 (en) | 2022-01-19 | 2024-01-02 | Cape Analytics, Inc. | System and method for object analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108197583B (en) | 2020-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108197583A (en) | The building change detecting method of optimization and image structure feature is cut based on figure | |
| Clode et al. | The automatic extraction of roads from LIDAR data | |
| CN109919944B (en) | Combined superpixel graph-cut optimization method for complex scene building change detection | |
| CN103632363A (en) | Object-level high-resolution remote sensing image change detection method based on multi-scale fusion | |
| CN106340005A (en) | High-resolution remote sensing image unsupervised segmentation method based on scale parameter automatic optimization | |
| CN115690081A (en) | Tree counting method, system, storage medium, computer equipment and terminal | |
| Lambers et al. | Towards detection of archaeological objects in high-resolution remotely sensed images: the Silvretta case study | |
| Ural et al. | Road and roadside feature extraction using imagery and LiDAR data for transportation operation | |
| Yao et al. | Automated detection of 3D individual trees along urban road corridors by mobile laser scanning systems | |
| YU et al. | Object oriented land cover classification using ALS and GeoEye imagery over mining area | |
| CN113240735B (en) | Slope displacement activity monitoring method | |
| CN112166688B (en) | Method for monitoring desert and desertification land based on minisatellite | |
| He et al. | Narrow-linear and small-area forest disturbance detection and mapping from high spatial resolution imagery | |
| Tian et al. | A Process-Oriented Method for Rapid Acquisition of Canopy Height Model From RGB Point Cloud in Semiarid Region | |
| David et al. | Pathway detection and geometrical description from ALS data in forested mountaneous area | |
| Wang et al. | Canopy extraction and height estimation of trees in a shelter forest based on fusion of an airborne multispectral image and photogrammetric point cloud | |
| Goetzke et al. | A method to map land-use change and urban growth in North Rhine-Westphalia (Germany) | |
| Kaur et al. | Automatic road detection of satellite images—A survey | |
| Saliola | Object-based image analysis for the delineation of canopy gaps and individual tree crowns using multi-source data: A case study in Haliburton Forest, Ontario | |
| Wu et al. | Voxel-based marked neighborhood searching method for identifying street trees using vehicle-borne laser scanning data | |
| Yao et al. | Automatic vehicle extraction from airborne LiDAR data of urban areas using morphological reconstruction | |
| Yastikli et al. | Building extraction using multi sensor systems | |
| Kunyuan et al. | Automated object extraction from MLS data: A survey | |
| Mohammadzadeh et al. | A self-organizing fuzzy segmentation (SOFS) method for road detection from high resolution satellite images | |
| Yuan et al. | DEM generation from airborne LIDAR data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |