CN108491826A - A kind of extraction method of remote sensing image building - Google Patents
A kind of extraction method of remote sensing image building Download PDFInfo
- Publication number
- CN108491826A CN108491826A CN201810307327.1A CN201810307327A CN108491826A CN 108491826 A CN108491826 A CN 108491826A CN 201810307327 A CN201810307327 A CN 201810307327A CN 108491826 A CN108491826 A CN 108491826A
- Authority
- CN
- China
- Prior art keywords
- layer
- num
- building
- value
- pixel
- 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
- G06V10/20—Image preprocessing
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/247—Aligning, centring, orientation detection or correction of the image by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of extraction methods of remote sensing image building.Include the following steps:Step 1, multi-spectrum remote sensing image is inputted;Step 2, traversal all pixels point extracts two layerings to the distance between two points randomly selected;Step 3, the cymomotive force of two layerings is calculated;Step 4, differentiate candidate architecture object figure layer and non-building figure layer;Step 5, initialization area quantity;Step 6, candidate architecture object figure layer is handled;Step 7, the iteration stopping condition of determination step 6;Step 8, mark is handled;Step 9, it post-processes;Step 10, result is exported.The building in multi-spectrum remote sensing image can be accurately extracted, especially intensive building can be applied to the update of building in the geographical basis information database of city.
Description
Technical field
The present invention relates to a kind of remote sensing image process field, specifically a kind of side of automatically extracting of remote sensing image building
Method.
Background technology
Building is one of main geographic element in city, is the important content of various city thematic maps, studies building
Extraction is of great significance to integrated survey urban geographic information environment.It is quick with high-resolution remote sensing image acquiring technology
Development, the processing of remote sensing image, analysis and application have a better data source, digital product then have more extensively, deeper into
Application.Computer image processing technology, pattern-recognition, artificial intelligence etc. all obtain different degrees of progress, for height
Effective information in effect ground extraction huge image data provides possibility.But the extraction of building information than other information such as road,
The acquisition of water body is much more difficult, and main cause is as follows:
(1) data source is mainly two-dimensional remote sensing image, in most cases lacks direct three-dimensional data;
(2) factors such as different remote sensing image Chang Yinwei spectral regions, resolution ratio, the several picture of sensor and image-forming condition
Difference and have larger difference;
(3) its appearance for being showed of different types of building and grain details etc. are ever-changing, show remote sensing images
Upper widely different, unified building model library is difficult to set up, this makes automatically extracting for information become extremely difficult;
(4) complexity of scene residing for building, when as relatively low such as contrast, house mutually block, the shade of building itself with
And it is in the shade etc. of other atural objects, so thinking that the building for automatically extracting sharpness of border from background is more difficult.
Invention content
The present invention provides a kind of extraction methods of remote sensing image building, can overcome in current remote sensing image and build
The difficult problem of object extraction, makes full use of R in remote sensing image, the feature of tri- components of G, B, between feature based vector away from
From, the building target in remote sensing image can be detected, be not necessarily to manual intervention, high degree of automation.
Technical solution is used by target to realize the present invention:Method includes the following steps:
Step 1:Input multi-spectrum remote sensing image comprising R, tri- color components of G, B is pre-processed, image I is obtainedin;
Step 2:In image IinRGB color SPA in randomly select at 2 points, be denoted as P respectively1And P2, calculate image IinIn picture
Vegetarian refreshments PxP is arrived respectively1And P2The distance between, it is denoted as d respectively1xAnd d2x, work as d1x≤d2xWhen, by pixel PxWith P1It merges into same
One set, and it is denoted as S1, otherwise by pixel PxWith P2Identity set is merged into, and is denoted as S2, work as S1In there is new pixel to be added after,
By P1Location updating be S1The mean place (the equal coordinate value rounding of contraposition horizontalization) of middle all pixels, works as S2In there is new pixel to add
After entering, by P2Location updating be S2The mean place (the equal coordinate value rounding of contraposition horizontalization) of middle all pixels;Iterative step 2
The above process, until traversing image IinIn all pixels point, obtain two layering Layer1And Layer2;
Step 3:With following formula to two layering Layer in step 21And Layer2It is handled:
In formula (1), Fltlayer_numTo be layered Layerlayer_numCymomotive force, meanlayer_numTo be layered Layerlayer_num
Mean value in RGB color SPA, layer_num are two layering numbers in step 2, and value is 1 and 2, IiIt is i-th
Value of a pixel in RGB color SPA, Nlayer_numTo be layered Layerlayer_numIn number of pixels;
Step 4:Work as Fltlayer_numMeet condition T0When, Layer will be layeredlayer_numIt is determined as candidate architecture object figure layer
Layercb, otherwise, Layer will be layeredlayer_numIt is determined as non-building figure layer Layernb;
Step 5:By the candidate architecture object figure layer Layer in step 4cbRegion quantity be initialized as C0;
Step 6:With following formula to candidate architecture object figure layer LayercbIt is handled:
In formula (2), ObjF is extracted region object function, HiFor feature of the ith pixel point in RGB color SPA to
Amount, RiFor value of the ith pixel point in R component, GiFor value of the ith pixel point in G components, BiFor ith pixel
Value of the point in B component, HQFor feature vectors of the barycenter pixel Qc in RGB color SPA of region Q, RQcFor picture
Values of the vegetarian refreshments Qc in R component, GQcFor values of the pixel Qc in G components, BQcFor pixel Qc taking in B component
Value, pQ(i) belong to the probability of region Q for ith pixel point, | | Hi-HQ| | feature vector H is sought in expressioniAnd HQThe distance between;
Step 7:When | ObjF(k+1)-ObjF(k)When |≤Thr, regional ensemble RS is obtained, enters step 8, otherwise, by C0Value
It is updated to C0+ 1, return to step 5, ObjF(k)For the value of ObjF after kth time iteration;
Step 8:Processing is labeled to the regional ensemble RS in step 7;
Step 9:It deletes area and is less than S0Non- construction zone, use rectangular degree and length-width ratio as constraints extract build
Object;Step 10:Output extraction building result.
Pretreatment in the step 1 includes geometric correction, radiant correction and contrast enhancing.
Distance in the step 6 | | Hi-HQ| | acquiring method use Chebyshev's distance.
The beneficial effects of the invention are as follows:The building in multi-spectrum remote sensing image can be accurately extracted, it is especially intensive to build
Object is built, the update of building in the geographical basis information database of city is can be applied to.
Description of the drawings
Fig. 1 is the overall process flow figure of the present invention.
Specific implementation mode
Detailed description of the present invention specific implementation mode below in conjunction with the accompanying drawings.
In step 101, input includes R, the multi-spectrum remote sensing image of tri- color components of G, B.
In step 102, the input multi-spectrum remote sensing image of step 101 is pre-processed, including geometric correction, radiation school
Just enhance with contrast, obtains image Iin。
In step 103, in image IinRGB color SPA in randomly select at 2 points, be denoted as P respectively1And P2。
In step 104, image I is calculatedinIn pixel PxP is arrived respectively1And P2The distance between, it is denoted as d respectively1xWith
d2x, work as d1x≤d2xWhen, by pixel PxWith P1Identity set is merged into, and is denoted as S1, otherwise by pixel PxWith P2It merges into same
One set, and it is denoted as S2, work as S1In there is new pixel to be added after, by P1Location updating be S1The mean place of middle all pixels is (right
Position average coordinates value rounding), work as S2In there is new pixel to be added after, by P2Location updating be S2The average bit of middle all pixels
It sets (the equal coordinate value rounding of contraposition horizontalization).
In step 105, judge whether to traverse image IinIn all pixels point, if it is, obtain two layering
Layer1And Layer2And 106 are entered step, otherwise enter step 104.
In step 106, with following formula to two layering Layer in step 1051And Layer2It is handled:
In formula (3), Fltlayer_numTo be layered Layerlayer_numCymomotive force, meanlayer_numTo be layered Layerlayer_num
Mean value in RGB color SPA, layer_num are two layering numbers in step 2, and value is 1 and 2, IiIt is i-th
Value of a pixel in RGB color SPA, Nlayer_numTo be layered Layerlayer_numIn number of pixels.
In step 107, Flt is judgedlayer_numWhether condition T is met0, if it is, Layer will be layeredlayer_numDifferentiate
For candidate architecture object figure layer LayercbAnd 109 are entered step, it otherwise will be layered Layerlayer_numIt is determined as non-building figure layer
LayernbAnd 108 are entered step, and according to experimental data, when for extracting intensive building, condition T0It need to be set as
Fltlayer_num>=465, when for extracting sparse building, condition T0It need to be set as 250≤Fltlayer_num< 465.
In step 108, non-building figure layer Layer is obtainednb, indicate that the input multi-spectrum remote sensing image of step 101 is free of
Building.
In step 109, by the candidate architecture object figure layer Layer in step 107cbRegion quantity be initialized as C0, it is simultaneous
The speed of service and building extraction effect are cared for, by C0It is set as 10.
In step 110, structure realm extracts object function ObjF:
In formula (4), ObjF is extracted region object function, HiFor feature of the ith pixel point in RGB color SPA to
Amount, RiFor value of the ith pixel point in R component, GiFor value of the ith pixel point in G components, BiFor ith pixel
Value of the point in B component, HQFor feature vectors of the barycenter pixel Qc in RGB color SPA of region Q, RQcFor picture
Values of the vegetarian refreshments Qc in R component, GQcFor values of the pixel Qc in G components, BQcFor pixel Qc taking in B component
Value, pQ(i) belong to the probability of region Q for ith pixel point, | | Hi-HQ| | feature vector H is sought in expressioniAnd HQBetween cut ratio
Avenge husband's distance.
In step 111, with the extracted region object function ObjF in step 110 to candidate architecture object figure layer LayercbIt carries out
Processing.
In step 112, ObjF is used(k)The value of ObjF after expression kth time iteration, judges whether ObjF meets stop condition |
ObjF(k+1)-ObjF(k)|≤Thr enters step 113 if it is, obtaining regional ensemble RS, otherwise by C0Value be updated to
C0It+1 and enters step 111, in order to ensure to extract the accuracy of result, threshold value Thr is set as 10-4。
In step 113, processing is labeled to the regional ensemble RS in step 112.
It in step 114, is post-processed, including deletes area and be less than S0Non- construction zone, with rectangular degree and length and width
Than extracting building, the interference of small area atural object in order to prevent, by S as constraints0It is set as 50.
In step 115, result is exported.
Claims (3)
1. a kind of extraction method of remote sensing image building, it is characterised in that include the following steps:
Step 1:Input multi-spectrum remote sensing image comprising R, tri- color components of G, B is pre-processed, image I is obtainedin;
Step 2:In image IinRGB color SPA in randomly select at 2 points, be denoted as P respectively1And P2, calculate image IinIn
Pixel PxP is arrived respectively1And P2The distance between, it is denoted as d respectively1xAnd d2x, work as d1x≤d2xWhen, by pixel PxWith P1Merge
For identity set, and it is denoted as S1, otherwise by pixel PxWith P2Identity set is merged into, and is denoted as S2, work as S1In there is new pixel to add
After entering, by P1Location updating be S1The mean place (the equal coordinate value rounding of contraposition horizontalization) of middle all pixels, works as S2In have new picture
After element is added, by P2Location updating be S2The mean place (the equal coordinate value rounding of contraposition horizontalization) of middle all pixels;Iterative step
2 above process, until traversing image IinIn all pixels point, obtain two layering Layer1And Layer2;
Step 3:With following formula to two layering Layer in step 21And Layer2It is handled:
In formula (1), Fltlayer_numTo be layered Layerlayer_numCymomotive force, meanlayer_numTo be layered Layerlayer_num
Mean value in RGB color SPA, layer_num are two layering numbers in step 2, and value is 1 and 2, IiIt is i-th
Value of a pixel in RGB color SPA, Nlayer_numTo be layered Layerlayer_numIn number of pixels;
Step 4:Work as Fltlayer_numMeet condition T0When, Layer will be layeredlayer_numIt is determined as candidate architecture object figure layer
Layercb, otherwise, Layer will be layeredlayer_numIt is determined as non-building figure layer Layernb;
Step 5:By the candidate architecture object figure layer Layer in step 4cbRegion quantity be initialized as C0;
Step 6:With following formula to candidate architecture object figure layer LayercbIt is handled:
In formula (2), ObjF is extracted region object function, HiFor feature of the ith pixel point in RGB color SPA to
Amount, RiFor value of the ith pixel point in R component, GiFor value of the ith pixel point in G components, BiFor ith pixel
Value of the point in B component, HQFor feature vectors of the barycenter pixel Qc in RGB color SPA of region Q, RQcFor picture
Values of the vegetarian refreshments Qc in R component, GQcFor values of the pixel Qc in G components, BQcFor pixel Qc taking in B component
Value, pQ(i) belong to the probability of region Q for ith pixel point, | | Hi-HQ| | feature vector H is sought in expressioniAnd HQThe distance between;
Step 7:When | ObjF(k+1)-ObjF(k)When |≤Thr, regional ensemble RS is obtained, enters step 8, otherwise, by C0Value
It is updated to C0+ 1, return to step 5, ObjF(k)For the value of ObjF after kth time iteration;
Step 8:Processing is labeled to the regional ensemble RS in step 7;
Step 9:It deletes area and is less than S0Non- construction zone, use rectangular degree and length-width ratio as constraints extract building;
Step 10:Output extraction building result.
2. a kind of extraction method of remote sensing image building according to claim 1, it is characterised in that in step 1
Pretreatment includes geometric correction, radiant correction and contrast enhancing.
3. a kind of extraction method of remote sensing image building according to claim 1, it is characterised in that in step 6
Distance | | Hi-HQ| | acquiring method use Chebyshev's distance.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810307327.1A CN108491826B (en) | 2018-04-08 | 2018-04-08 | Automatic extraction method of remote sensing image building |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810307327.1A CN108491826B (en) | 2018-04-08 | 2018-04-08 | Automatic extraction method of remote sensing image building |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108491826A true CN108491826A (en) | 2018-09-04 |
CN108491826B CN108491826B (en) | 2021-04-30 |
Family
ID=63315035
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810307327.1A Expired - Fee Related CN108491826B (en) | 2018-04-08 | 2018-04-08 | Automatic extraction method of remote sensing image building |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108491826B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635715A (en) * | 2018-12-07 | 2019-04-16 | 福建师范大学 | A kind of remote sensing images building extracting method |
CN110298348A (en) * | 2019-06-12 | 2019-10-01 | 苏州中科天启遥感科技有限公司 | Remote sensing image building sample areas extracting method and system, storage medium, equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102015200260A1 (en) * | 2014-01-10 | 2015-07-16 | Mitsubishi Electric Corporation | Method of creating a descriptor for a scene image |
CN104794478A (en) * | 2015-05-04 | 2015-07-22 | 福建师范大学 | Method for extracting buildings with uniform spectral characteristics from remote sensing images |
CN105761266A (en) * | 2016-02-26 | 2016-07-13 | 民政部国家减灾中心 | Method of extracting rectangular building from remote sensing image |
-
2018
- 2018-04-08 CN CN201810307327.1A patent/CN108491826B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102015200260A1 (en) * | 2014-01-10 | 2015-07-16 | Mitsubishi Electric Corporation | Method of creating a descriptor for a scene image |
CN104794478A (en) * | 2015-05-04 | 2015-07-22 | 福建师范大学 | Method for extracting buildings with uniform spectral characteristics from remote sensing images |
CN105761266A (en) * | 2016-02-26 | 2016-07-13 | 民政部国家减灾中心 | Method of extracting rectangular building from remote sensing image |
Non-Patent Citations (2)
Title |
---|
吴宁: "基于目标识别和参数化技术的城市建筑群三维重建研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
施文灶 等: "多光谱遥感影像建筑物提取", 《计算机***应用》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635715A (en) * | 2018-12-07 | 2019-04-16 | 福建师范大学 | A kind of remote sensing images building extracting method |
CN110298348A (en) * | 2019-06-12 | 2019-10-01 | 苏州中科天启遥感科技有限公司 | Remote sensing image building sample areas extracting method and system, storage medium, equipment |
CN110298348B (en) * | 2019-06-12 | 2020-04-28 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image building sample region, storage medium and equipment |
Also Published As
Publication number | Publication date |
---|---|
CN108491826B (en) | 2021-04-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106127791B (en) | A kind of contour of building line drawing method of aviation remote sensing image | |
CN104331699B (en) | A kind of method that three-dimensional point cloud planarization fast search compares | |
Medioni et al. | Matching images using linear features | |
CN107292234B (en) | Indoor scene layout estimation method based on information edge and multi-modal features | |
CN109949340A (en) | Target scale adaptive tracking method based on OpenCV | |
Delmerico et al. | Building facade detection, segmentation, and parameter estimation for mobile robot localization and guidance | |
CN113706482A (en) | High-resolution remote sensing image change detection method | |
CN111340939B (en) | Indoor three-dimensional semantic map construction method | |
Wang et al. | Window detection from mobile LiDAR data | |
CN103646391A (en) | Real-time camera tracking method for dynamically-changed scene | |
CN114612488A (en) | Building-integrated information extraction method, computer device, and storage medium | |
CN113192200B (en) | Method for constructing urban real scene three-dimensional model based on space-three parallel computing algorithm | |
Zhang et al. | Vehicle global 6-DoF pose estimation under traffic surveillance camera | |
Zhang et al. | Lidar-guided stereo matching with a spatial consistency constraint | |
CN109754440A (en) | A kind of shadow region detection method based on full convolutional network and average drifting | |
Yang et al. | Visual SLAM based on semantic segmentation and geometric constraints for dynamic indoor environments | |
Matei et al. | Image to lidar matching for geotagging in urban environments | |
CN113487631A (en) | Adjustable large-angle detection sensing and control method based on LEGO-LOAM | |
Wang et al. | A method for detecting windows from mobile LiDAR data | |
CN108491826A (en) | A kind of extraction method of remote sensing image building | |
Mu et al. | McGET: A rapid image-based method to determine the morphological characteristics of gravels on the Gobi desert surface | |
CN105631849B (en) | The change detecting method and device of target polygon | |
Zhang et al. | A robust visual odometry based on RGB-D camera in dynamic indoor environments | |
CN109635715A (en) | A kind of remote sensing images building extracting method | |
Wang et al. | Multidimensional particle swarm optimization-based unsupervised planar segmentation algorithm of unorganized point clouds |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210430 |