CN103279951A - Object-oriented remote sensing image building and shade extraction method of remote sensing image building - Google Patents
Object-oriented remote sensing image building and shade extraction method of remote sensing image building Download PDFInfo
- Publication number
- CN103279951A CN103279951A CN2013101764874A CN201310176487A CN103279951A CN 103279951 A CN103279951 A CN 103279951A CN 2013101764874 A CN2013101764874 A CN 2013101764874A CN 201310176487 A CN201310176487 A CN 201310176487A CN 103279951 A CN103279951 A CN 103279951A
- Authority
- CN
- China
- Prior art keywords
- buildings
- shade
- image
- potential district
- remote sensing
- 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
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to an object-oriented remote sensing image building and a shade extraction method of the remote sensing image building. The method comprises the following steps that the images are preprocessed; the multi-scale division is carried out, and an image scale layer corresponding to the optimum division size of a ground object is selected; non-buildings and non-shades (roads, water bodies, vegetation and bare land) are extracted, and images comprising buildings and the shade candidate object regions are obtained; shade primary extraction is carried out by using spectrum characteristics, the nearest classifier is adopted, a characteristic space is built by combining with the spectrum characteristics, the building primary extraction is carried out, and building potential region and a shade potential region are obtained; and buildings and shades of the buildings are obtained by adopting optimization processing methods of space characteristics, context relationship, smoothening and the like. The object-oriented remote sensing image building and the shade extraction method can effectively solve the problem of confusion among the buildings, the roads and the bare land as well as between the shades of the buildings and the water bodies, the extraction precision of the buildings and the shades of the buildings is high, the process flow is simple, and a certain practical value is realized in the work of basic data updating, urban planning, urban disaster prevention and reduction in the urban construction.
Description
Technical field
The present invention relates to remote sensing technology ground object target information extraction field, particularly relate to the method that a kind of OO remote sensing image building and shade thereof extract.
Background technology
Buildings is the main part on the high-resolution remote sensing image of city, and obtaining for the work such as basic data renewal in city planning, Urban Disaster Prevention and Mitigation and the urban construction of buildings and shadow information thereof has great importance.Remote sensing technology relies on characteristics such as its large tracts of land simultaneous observation, high-timeliness, economy, for the extraction of buildings and shadow information thereof provides a new technological means.Widespread use along with high-resolution remote sensing image, information such as the spectrum of ground object target, shape, texture are more obvious on the image, this identification for ground object target provides rich data source, in recent years, how to obtain buildings and shadow information thereof quickly and accurately is focus and difficult point in the ground object target Study of recognition always.
Aspect the buildings extraction, Chinese scholars has proposed a large amount of model methods and strategy, mainly can be divided three classes:
(1) utilizes the method for rim detection, by some pattern-recognition means, from image, extract earlier the edge of linear feature, identify the edge of buildings again by some means, these class methods have certain effect to the extraction of buildings, but because ambiguity and the similar linear feature of other atural object of linear edge as road, square etc., make detected buildings have the situation that mistake is divided and leakage divides.Contour of building is carried out rim detection and testing result is strengthened as employing Sobel operators such as Li Weiyue, realize the extraction [Li Weiyue of buildings, Hu Zhibin, Zhu Jiaojun, Li Na, Ma Cong, 2009. utilize rim detection to the extraction research of buildings in the high resolution image, remote sensing technology and application, 502-506.].
(2) utilize the elevation information of buildings such as DSM, LIDAR, stereogram image equal altitudes data, come the extraction of ancillary building, but these data often are not easy to obtain, therefore can be subjected to the restriction of data source.As Cheng Liang etc. the auxiliary contour of building method [Cheng Liang, Gong Jianya, the auxiliary ultrahigh resolution image that utilizes down of 2008.LiDAR extracts the contour of building method, surveys and draws journal, 391-393+399.] of extracting down at LiDAR has been proposed.
(3) adopt OO method to extract buildings.These class methods at first are divided into object with image, utilize the spectrum, how much, texture of object and contextual information to come building target is extracted on this basis, and the precision of identification is higher, and can satisfy the needs that multiple dimensioned sensor information is used.Scholars have also proposed some strategies in succession, but the method for current development generally is at the specific region, and the buildings roof Material is comparatively single, and leaching process is comparatively loaded down with trivial details, and universality is not strong, and context relation use few, effect is not fine.As the buildings extracting method in conjunction with spectral signature and shape facility of propositions such as Wu Wei, detect quality and have only 75.5%.[Wu Wei, et al., the buildings extracting method of the high-resolution remote sensing image that spectrum and shape facility combine. Wuhan University's journal (information science version), 2012 (07): p.800-805.].
For the identification of buildings shade, mainly contain two class methods: a class is based on the method for model, and another kind of is to extract target according to the characteristic of shade.First kind method generally needs priori, obtain after enough parameters to calculate by model to obtain, but these parameters obtain the comparison difficulty, and shade easily and water body obscure, thereby nicety of grading is not high; The corresponding usefulness of second class methods is more extensive, and it is to utilize shade to extract in reasonable characteristics of aspect consistance such as texture, gray scale, edges, and generally can not be subjected to the influence of view field, and effect is relatively good.List of references: [Lorenzi, L., Melgani, F., Mercier, G., 2012.A Complete Processing Chain for Shadow Detection and Reconstruction in VHR Images.Geoscience and Remote Sensing, IEEE Transactions on50,3440-3452.] and [Zhao Hongyan, Wang Junyu, 2008. the research of city high resolution image buildings shade, pp.560-561-562-563.], in object-oriented method, general second class methods that adopt utilize the spectral signature of shade to carry out the shade extraction.
Object-oriented method in the past is in the process that buildings is extracted, consider its spectrum and space characteristics more, and to the less utilization of its contextual feature, and the method for extracting buildings and shade thereof is simultaneously arranged seldom, though Zhou Yanan and Zhou Xiaocheng etc. have utilized the space characteristics of buildings shade and buildings to extract buildings and shade [Zhou Yanan thereof, Shen Zhanfeng, Luo Jiancheng, Chen Qiuxiao, Hu Xiaodong, Shen Jinxiang, 2010. the object-oriented City Building under shade is auxiliary extracts. geographical and geography information science, 37-40.] and [Zhou Xiaocheng, 2008.12. the high-resolution satellite image buildings recognition methods in conjunction with the object relationship feature. land resources remote sensing .], but still there is not fully to excavate the potential of context relation, mutual spatial relationship between each object particularly, thus solve buildings and shade inside " cavity " phenomenon thereof effectively and obtain comparatively level and smooth problems such as edge.
Summary of the invention
Technical matters to be solved by this invention is: a kind of OO image analysing computer method is provided, only utilizes the panchromatic and multi light spectrum hands of high-resolution remote sensing image, realize the extraction of buildings and shade thereof simultaneously.
The present invention proposes a kind of object-oriented policies, can extract buildings and shade thereof simultaneously: at first in high-resolution remote sensing image, reject most of road, water body, the influence of vegetation and bare area, tentatively extract the potential district of shade and the potential district of buildings then, and then take full advantage of spatial relationship, context relation is (as the relation of buildings and shade, relation between buildings and the bare area easily obscured, the inside that buildings or shadow object surround also should be that buildings or shade solve " cavity " problem) and progressively optimize the buildings that extracts and shade precision height thereof by the means such as edge that the standardization processing algorithm smoothly extracts the result.
The method that OO remote sensing image building provided by the invention and shade thereof extract, its step comprises:
(1) image pre-service.According to the situation of study area, the panchromatic wave-band image of the same area and multi light spectrum hands image are carried out cutting handle, obtain the image I of study area
1
(2) multi-scale division.Through pretreated remote sensing image I
1, according to spectrum standard and the shape criteria parameter set, select 30 to cut apart yardstick: 5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,110,120,130,140,150,160,170,180,190,200 pairs of images carry out multi-scale division, form the imaged object layer of different scale, utilize time maximum area to select the optimum segmentation yardstick in conjunction with visualization, consider the treatment effeciency of image, yardstick 200,55 and 30 are selected as the comparatively suitable yardstick of final atural object extraction, have formed 3 yardstick imaged object layers, i.e. image I thus
2, I
3, I
4
(3) non-buildings and unshaded extraction.At 200 yardstick image I
2On, for each object, if length characteristic and length breadth ratio characteristic threshold value satisfy logical and, then belong to road class R, otherwise remain unfiled, obtained extracting the image I of road
5At 55 yardstick image I
3On, if NDWI, NDVI and red band ratios satisfy certain threshold condition, then this object belongs to water body class W, vegetation class V and bare area class B respectively, has obtained to extract the image I of water body, vegetation, bare area
6Then with image I
5And I
6Classification results inherit 30 yardstick image I
4On, thereby obtain image I
7, the image that can reject road, water body, greenery patches and bare area like this improves the accuracy rate that buildings and its shade extract;
(4) extraction in buildings and the potential district of shade thereof.At image I
7On, never utilize spectral signature to extract the potential district of shade SH in the object of classification; By closing on most sorter, extract buildings potential district BH in conjunction with spectral signature, this spectral signature comprises wave band average, brightness average, the potential district of shade that misses when the atural object that obtains simultaneously also has the potential district of part shade TS(also to be front extraction SH), shoofly class TR and interim bare area class TB, TS is referred to the potential district of shade SH, thereby obtains comprising the image I of shade potential district SH and the potential district of buildings BH
8
(5) optimization process of buildings and shade thereof.At image I
8On, adopt space characteristics, context relation and smoothing algorithm shade to be optimized processing, the shadow category S that is finally identified; Simultaneously, utilize context relation, smoothing algorithm, be optimized processing, the buildings classification B that is finally identified to having a lot of wrong branches among the potential district of the buildings BH and leaking the buildings that divides;
Through above-mentioned steps, obtain the extraction of described OO remote sensing image building and shade thereof.
The spectrum standard of setting in the above-mentioned steps (2) and shape criteria parameter obtain in conjunction with visualization by expertise, and wherein, the ratio of spectrum and shape criteria is being 1:1, and the ratio of smoothness parameter and degree of compacting parameter also is 1:1.
In the above-mentioned steps (3), described NDWI, NDVI and red band ratios satisfy certain threshold condition and refer to: the NDWI threshold value is that 0.3～0.4, NDVI threshold value is 0.25～0.35, and red band ratios threshold value is 0.1～0.2.
The shadow category S that the present invention can adopt following optimization process method finally to be determined, its step comprises:
(1) at first merges the potential district of adjacent shade object, in order to carry out follow-up space characteristics, context relation and smoothing algorithm shade is optimized processing;
When (2) object area of the potential district of the shade after merging SH satisfies certain threshold condition, it is classified as unfiled, thereby reject the wrong buildings shade that divides of part; Described certain threshold condition refers to: area threshold Area is 250～300;
(3) object of the potential district of shade SH encirclement also should belong to shadow category, utilizes this context relation, is referred to the potential district of shade SH with leaking the shadow object that divides;
(4) adopt smoothing algorithm to optimize the shade edge, make it become smooth, obtain described shadow category S like this.
The buildings classification B that the present invention can adopt following optimization process method finally to be determined, its step comprises:
(1) object with the potential district of adjacent buildings BH merges;
(2) buildings is all adjacent with shade, utilizes these neighbouring relations can reject a lot of wrong buildingss that divide;
(3) object of the potential district of buildings BH encirclement also should belong to the buildings classification, utilizes this context relation, can be referred to the potential district of buildings BH with leaking the buildings object that divides;
(4) still have some buildingss to be missed, be categorized into other atural object classification, at this moment judge whether it to be referred to the buildings classification by its adjacent degree with the buildings of classifying;
(5) adopt smoothing algorithm standardization buildings edge, make it become smooth, obtain final buildings classification B like this.
Above-mentioned smoothing algorithm can increase and the Shrinking contraction algorithm for: Growing, and the part that falls in is filled up by the window of 5*5, if more than or equal to 0.5, then continues to increase, and so circulates; The classification of projection is shunk by the window of 5*5, if smaller or equal to 0.5, then continues to shrink so circulation.By they can smooth object the edge.
The present invention compared with prior art has following main advantage:
(1) the optimum segmentation yardstick that atural object extracts is chosen in combination time maximum area method and visualization, compares traditional great majority and only adopts repeatedly trial acquisition optimum segmentation yardstick, and is more objective, reduced the workload of man-machine interaction.
(2) before extracting buildings and shade thereof, the influence of at first having rejected other classifications is done like this and can be improved the precision that target atural object extracts.Spectral signature, space characteristics by object, choose to extract road, water body, greenery patches and bare area by combination, as use object length and length breadth ratio to extract road, the NDWI of the spectral value calculating object by wave band four and wave band three, extract water body thereby utilize, utilize the NDVI value of calculating to extract the greenery patches, adopt red band ratio to extract bare area.
(3) pass through knowledge rule flexibly, its method is closing on classification most in conjunction with the spectral signature employing to have replenished the shade that partly leaks branch, and utilize optimization means such as going up space feature, contextual feature that the potential district of shade is handled, utilize the relation of shade and buildings and the knowledge rule that other contextual feature is set up, potential district progressively optimizes to buildings, final buildings and shade thereof extract accuracy rate height as a result, verification and measurement ratio reaches 100%, compare the superfine house that detects more than 90% of making pottery, the accuracy of detection height.[Tao Chao, OO high-resolution remote sensing image city buildings fractional extraction method. mapping journal, 2010.2.]
(4) integrated buildings and shade thereof extract flow process, can extract buildings and shade thereof simultaneously.And the precision of extracting is higher, and the resultnat accuracy that namely obtains the classification of buildings and shade thereof is 80.67%.
Description of drawings
Fig. 1 is schematic flow sheet of the present invention.
Fig. 2 is non-buildings and unshaded extraction process flow diagram.
Fig. 3 is the extraction figure as a result in buildings and the potential district of shade thereof.
Fig. 4 is buildings and the shadow vectors figure thereof that derives.
Embodiment
The invention will be further described below in conjunction with embodiment and accompanying drawing, but do not limit the present invention.
With reference to Fig. 1, being the present invention carries out the process flow diagram that buildings and shade thereof extract to the high-resolution remote sensing image of a certain survey region in optical valley area, Wuhan City, may further comprise the steps:
One, image pre-service
The present invention has selected a panel height resolution remote sense image in optical valley area, Wuhan City, these data are remote sensing images that World View-II satellite was taken on July 29th, 2011, wherein multi light spectrum hands (comprising indigo plant, green, red, near red four wave bands) image has the spatial resolution of 2m, and the panchromatic wave-band image spatial resolution is 0.5m.This image has carried out overshoot and geometry correction, and the Geographic Reference system is arranged, and according to the concrete condition of study area, image has been carried out the cutting processing, the study area image I that obtains after the cutting
1Size is 1664*1672pixels, as Fig. 2.Observe image, can find that features such as interior three the subdistrict architecture spectrum of this study area, shape have diversity, the shape of buildings shade is also not too regular, and complicated scene has increased the difficulty of buildings and shade extraction thereof;
Two, multi-scale division
Through pretreated remote sensing image I
1, in conjunction with expertise and visualization, the spectrum that we select and the ratio of shape criteria are 1:1, the ratio of smoothness parameter and degree of compacting parameter also is 1:1, selects 30 to cut apart the yardstick threshold value: 5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,110,120,130,140,150,160,170,180,190,200 carry out multi-scale division, form the imaged object layer of different scale.What multi-scale division adopted is heterogeneous minimum region merging algorithm, it is to carry out to cut apart task under the partitioning parameters of setting: by given yardstick threshold value, homogeneity criterion according to color and shape, based on the heterogeneous minimum principle of object, the similar adjacent pixel of spectral information is merged into significant object.The selection principle of multi-scale division parameter is: under the prerequisite that satisfies necessary shape criteria, adopt color standard as far as possible, this is that the weight of shape criteria is too high, can reduce the quality of cutting apart on the contrary because the most important thing is spectral information in the image.If the border smoother of ground object target is just set higher smoothness parameter, for the type of ground objects that shape is compacted, bigger degree of compacting parameter should be set.Multi-scale division is tied the object that generates according to certain hierarchical organization, form the imaged object hierarchical structure of different scale, this with face of land entity class like hierarchical organization make the remote sensing image of intrinsic resolution can satisfy different application demands.After multi-scale division is finished, utilize time maximum area to select the optimum segmentation yardstick in conjunction with visualization for each atural object, because we only pay close attention to the extraction of buildings and shade thereof, so under the prerequisite that satisfies suitable yardstick, should choose less yardstick layer as far as possible.Therefore, 200,55 and 30 these three yardsticks are selected the imaged object layer that extracts as atural object, i.e. image I
2, I
3, I
4Wherein, 200 yardsticks can extract water body, vegetation and bare area as the yardstick layer of road extraction on 55 yardsticks, and it is the most suitable that buildings and shade thereof extract at 30 yardstick imaged object layers;
Three, non-buildings and unshaded extraction
Extract most of more accurate non-buildings and non-shade (road, water body, vegetation, bare area) at suitable yardstick layer earlier, and this classification results is inherited on the yardstick layer of buildings and shade extraction thereof, therefrom extract the classification (buildings and shade) that we pay close attention to again, do like this and can get rid of them to the some effects of buildings and shade extraction thereof, specific implementation process is: at first, and at 200 yardstick imaged object layer I
2On, for each object, if length characteristic and length breadth ratio feature satisfy logical and, the threshold condition lower limit of preseting length and length breadth ratio is respectively 308,4.7, and the object that satisfies this condition is classified as road class R, otherwise remain unfiled, obtain extracting the image I behind the road
5Then, at 55 yardstick imaged object layer I
3On, the calculated characteristics Parameter N DWI of elder generation, and to set its lower limit be 0.32, extract water body class W, on this basis, if the feature NDVI value of unfiled object is more than or equal to 0.3, then be classified as vegetation class V, and then set under the red band ratios that remains unfiled object and be limited to 0.144, extract the bare area class B that satisfies this threshold condition, obtain extracting the image I of water body class W, vegetation class V and bare area class B thus
6At last, with image I
5With image I
6Classification results inherit 30 yardstick imaged object layer I
4On, obtain image I
7As shown in Figure 3, be the flow process frame diagram of this step.This process can solve the problem that buildings and road, bare area are obscured effectively, also can reject the influence of the part vegetation similar with the shade spectral signature simultaneously, thereby improves the accuracy rate of buildings and the extraction of its shade;
Four, the extraction in buildings and the potential district of shade thereof
On remote sensing image, shade has lower gray-scale value, and different local buildings shades has stronger consistance, thereby simple spectral signature just can realize that shade just extracts, the potential district of acquisition shade.But there is diversity in the buildings roof Material, be difficult to unified feature it be identified, in addition, buildings is obscured easily and with the similar road of its spectral signature etc., these have all increased the difficulty of buildings identification greatly, and therefore, we adopt and close on sorter most, the feature space that constitutes in conjunction with spectral signature carries out buildings and just extracts, and obtains the potential district of buildings.Concrete implementation process is: through after the step 3, and image I
7The ground class that obtains has road class R, water body class W, vegetation class V, bare area class B, buildings and shade thereof are included in the remaining unfiled object, utilize brightness on be limited to 254.6, extract the potential district of the shade SH of the overwhelming majority, then by closing on most sorter, in conjunction with spectral signature (comprising wave band average, brightness average), in the unfiled object of residue, extract buildings potential district BH, the atural object that obtains simultaneously also has the potential district of part shade TS, shoofly class TR and interim bare area class TB.Wherein, TS is the fraction shade of missing when the potential district of shade SH is extracted in the front in order to replenish, and therefore TS is referred to the potential district of shade SH.Finally obtain comprising the image I of shade potential district SH and the potential district of buildings BH
8As Fig. 3, be the extraction result in buildings and the potential district of shade thereof, wherein, the potential district of shade is thick line mark (being numbered 1), the potential district of buildings marks (being numbered 2) with fine rule, examine, can find, some water body objects have been divided into shadow category by mistake, and the mistake of buildings is divided and it is just more serious to leak branch, and some bare areas and road are divided into buildings by mistake, bulk zone as the lower left corner, in addition, the object that some buildingss surround fails to be classified as the buildings object, and namely there is " cavity " phenomenon in part buildings classification the inside.
Five, the optimization process of buildings and shade thereof
Through after the previous step, there are some wrong branches and the leakage branch in buildings and shade thereof, need carry out follow-up optimization process.
The optimization process flow of shade is:
(1) in order to use following feature and algorithm, adopt this algorithm of merge region to merge the adjacent potential district of shade object earlier;
(2) object area of the potential district of the shade after merging SH is smaller or equal to 288 the time, then is classified as unfiledly, can reject some little objects that are not shade by this threshold condition, and this is because the buildings shaded area has certain size;
(3) there is " cavity " phenomenon in some shade inside, and the shade internal object should also belong to shadow category, and the internal object of utilizing " find enclosed by class " this algorithm just the potential district of shade SH can be surrounded correctly classifies as shade;
(4) by smoothing algorithm " pixel-based object resizing " the shade edge is carried out smoothing processing, can select growing and shrinking pattern according to actual conditions, the shade of observing us extracts the result, the part that falls in is filled up by the window of 5*5, if more than or equal to 0.5, then continue to increase so circulation; The classification of projection is shunk by the window of 5*5, if smaller or equal to 0.5, then continues to shrink so circulation.By their smooth shade edge well, the shadow category S that is finally determined.
The optimization process flow of buildings is:
(1) adopt merge region algorithm that the object of the potential district of adjacent buildings BH is merged;
(2) observe image, can see that a large amount of greenery patchess is divided into buildings by mistake, therefore set when the retive boundary of the object of the object of shade S and the potential BH of district of adjacent buildings is 0, be re-classified as vegetation, because buildings has certain shade, and the vegetation shade of study area almost can be ignored.
(3) object of the potential district of buildings BH encirclement also should belong to the buildings classification, utilizes this context relation, interior of building can be leaked the object that divides by " find enclosed by class " algorithm and reclassify the buildings classification;
(4) further observations, can see that some buildings edges should be that the object of buildings classification has been divided into the bare area classification by mistake, at first, adopt merge region algorithm that the object of adjacent bare area B is merged, then, if adjacent relative edge's dividing value of the object of the potential district of the object of bare area B and adjacent buildings BH classified as buildings more than or equal to 0.4 o'clock with it;
(5) growing and the shrinking algorithm specification buildings edge in the employing smoothing algorithm " pixel-based object resizing ", make it become smooth, same, the part that falls in is by the window of 5*5, threshold condition is made as more than or equal to 0.5, utilizes Growing to fill up; The classification of projection is carried out the Shrinking contraction by the window of 5*5, and threshold condition is made as smaller or equal to 0.5, obtains final buildings classification B like this;
Through after the above optimization process step, the extraction result of buildings and shade thereof obtains very big improvement, derives final buildings and the shade thereof that extracts, the polar plot that obtains such as Fig. 4, wherein, shade is thick line mark (being numbered 1), and buildings marks (being numbered 2) with fine rule.
In order to estimate the result that buildings and shade thereof extract more exactly, the present invention will extract result's stack of result and manual sort, adopt remote sensing evaluation of classification confusion matrix commonly used to estimating, utilize confusion matrix can calculate precision indexs such as resultnat accuracy, producer's precision and user's precision, the resultnat accuracy that obtains the classification of buildings and shade thereof is 80.67%, wherein buildings and shade class user precision reach 100%, shade class producer precision reaches 82.76%, and buildings class producer precision reaches 78.69%.
The present invention can solve buildings and road, bare area effectively, obscure problem between buildings shade and the water body, buildings and shade thereof extract the precision height, treatment scheme is simple, has certain practical value in the basic data renewal in urban construction, city planning, the Urban Disaster Prevention and Mitigation work.
Claims (7)
1. the method extracted of an OO remote sensing image building and shade thereof is characterized in that this method may further comprise the steps:
(1) image pre-service: according to the situation of study area, the panchromatic wave-band image of the same area and multi light spectrum hands image are carried out cutting handle, obtain the image I of study area
1
(2) multi-scale division: through pretreated remote sensing image I
1, according to spectrum standard and the shape criteria parameter set, select 30 to cut apart yardstick: 5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,110,120,130,140,150,160,170,180,190,200 pairs of images carry out multi-scale division, form the imaged object layer of different scale, utilize time maximum area to select the optimum segmentation yardstick in conjunction with visualization, consider the treatment effeciency of image, yardstick 200,55 and 30 are selected as the comparatively suitable yardstick of final atural object extraction, have formed 3 yardstick imaged object layers, i.e. image I thus
2, I
3, I
4
(3) non-buildings and unshaded extraction: at 200 yardstick image I
2On, for each object, if length characteristic and length breadth ratio characteristic threshold value satisfy logical and, then belong to road class R, otherwise remain unfiled, obtained extracting the image I of road
5At 55 yardstick image I
3On, if NDWI, NDVI and red band ratios satisfy certain threshold condition, then this object belongs to water body class W, vegetation class V and bare area class B respectively, has obtained to extract the image I of water body, vegetation, bare area
6Then with image I
5And I
6Classification results inherit 30 yardstick image I
4On, thereby obtain image I
7, the image that can reject road, water body, greenery patches and bare area like this improves the accuracy rate that buildings and its shade extract;
(4) extraction in buildings and the potential district of shade thereof: at image I
7On, never utilize spectral signature to extract the potential district of shade SH in the object of classification; By closing on most sorter, extract buildings potential district BH in conjunction with spectral signature, this spectral signature comprises wave band average, brightness average, the potential district of shade that misses when the atural object that obtains simultaneously also has the potential district of part shade TS(also to be front extraction SH), shoofly class TR and interim bare area class TB, TS is referred to the potential district of shade SH, thereby obtains comprising the image I of shade potential district SH and the potential district of buildings BH
8
(5) optimization process of buildings and shade thereof: at image I
8On, adopt space characteristics, context relation and smoothing algorithm shade to be optimized processing, the shadow category S that is finally identified; Simultaneously, utilize context relation, smoothing algorithm, be optimized processing, the buildings classification B that is finally identified to having a lot of wrong branches among the potential district of the buildings BH and leaking the buildings that divides;
Through above-mentioned steps, obtain the extraction of described OO remote sensing image building and shade thereof.
2. the method for claim 1, it is characterized in that spectrum standard and the shape criteria parameter set in the step (2) obtain in conjunction with visualization by expertise, wherein, the ratio of spectrum and shape criteria is being 1:1, and the ratio of smoothness parameter and degree of compacting parameter also is 1:1.
3. the method for claim 1, it is characterized in that in the step (3), described NDWI, NDVI and red band ratios satisfy certain threshold condition and refer to: the NDWI threshold value is that 0.3～0.4, NDVI threshold value is 0.25～0.35, and red band ratios threshold value is 0.1～0.2.
4. the method for claim 1 is characterized in that in the step (5), the shadow category S that adopts following optimization process method finally to be determined, and its step comprises:
(1) at first merges the potential district of adjacent shade object, in order to carry out follow-up space characteristics, context relation and smoothing algorithm shade is optimized processing;
When (2) object area of the potential district of the shade after merging SH satisfies certain threshold condition, it is classified as unfiled, thereby reject the wrong buildings shade that divides of part;
(3) object of the potential district of shade SH encirclement also should belong to shadow category, utilizes this context relation, is referred to the potential district of shade SH with leaking the shadow object that divides;
(4) adopt smoothing algorithm to optimize the shade edge, make it become smooth, obtain described shadow category S like this.
5. method as claimed in claim 4 is characterized in that described smoothing algorithm is: Growing increases and the Shrinking contraction algorithm, and the part that falls in is filled up by the window of 5*5, if more than or equal to 0.5, then continues to increase, and so circulates; The classification of projection is shunk by the window of 5*5, if smaller or equal to 0.5, then continues to shrink so circulation.By they can smooth object the edge.
6. method as claimed in claim 4, it is characterized in that in the step (2), described certain threshold condition refers to: area threshold Area is 250～300.
7. the method for claim 1 is characterized in that in the step (5), the buildings classification B that adopts following optimization process method finally to be determined, and its step comprises:
(1) object with the potential district of adjacent buildings BH merges;
(2) buildings is all adjacent with shade, utilizes these neighbouring relations can reject a lot of wrong buildingss that divide;
(3) object of the potential district of buildings BH encirclement also should belong to the buildings classification, utilizes this context relation, can be referred to the potential district of buildings BH with leaking the buildings object that divides;
(4) still have some buildingss to be missed, be categorized into other atural object classification, at this moment judge whether it to be referred to the buildings classification by its adjacent degree with the buildings of classifying;
(5) adopt smoothing algorithm standardization buildings edge, make it become smooth, obtain final buildings classification B like this.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310176487.4A CN103279951B (en) | 2013-05-13 | 2013-05-13 | A kind of method of OO remote sensing image building and shadow extraction thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310176487.4A CN103279951B (en) | 2013-05-13 | 2013-05-13 | A kind of method of OO remote sensing image building and shadow extraction thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103279951A true CN103279951A (en) | 2013-09-04 |
| CN103279951B CN103279951B (en) | 2016-03-09 |
Family
ID=49062458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310176487.4A Active CN103279951B (en) | 2013-05-13 | 2013-05-13 | A kind of method of OO remote sensing image building and shadow extraction thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103279951B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103839267A (en) * | 2014-02-27 | 2014-06-04 | 西安科技大学 | Building extracting method based on morphological building indexes |
| CN104794456A (en) * | 2015-05-04 | 2015-07-22 | 福建师范大学 | Remote-sensing image building detecting method based on colors and shadows |
| CN104915925A (en) * | 2015-06-12 | 2015-09-16 | 西南科技大学 | Dry valley extraction method by preparing high-precision mask file |
| CN105528596A (en) * | 2016-02-03 | 2016-04-27 | 长江大学 | High-resolution remote sensing image building automatic extraction method and system by using shadow |
| CN106683112A (en) * | 2016-10-10 | 2017-05-17 | 中国交通通信信息中心 | High-resolution image-based road region building change extraction method |
| CN107194857A (en) * | 2017-05-23 | 2017-09-22 | 国家地理空间信息中心 | Based on high score satellite remote sensing date using land resource analysis system |
| CN107977968A (en) * | 2017-12-22 | 2018-05-01 | 长江勘测规划设计研究有限责任公司 | The building layer detection method excavated based on buildings shadow information |
| CN109086666A (en) * | 2018-06-29 | 2018-12-25 | 中国水利水电科学研究院 | One kind realizing the fast automatic extracting method of building and system based on Object-Oriented Method |
| CN109815894A (en) * | 2019-01-23 | 2019-05-28 | 中国石油大学(华东) | A kind of building effects extraction process method for sentry's 2A image |
| CN110390267A (en) * | 2019-06-25 | 2019-10-29 | 东南大学 | A kind of mountain landscape Building extraction method and apparatus based on high score remote sensing image |
| CN110796042A (en) * | 2019-10-16 | 2020-02-14 | 长江大学 | High-resolution remote sensing image building extraction method based on form transformation of offset shadow sample |
| CN110852207A (en) * | 2019-10-29 | 2020-02-28 | 北京科技大学 | Blue roof building extraction method based on object-oriented image classification technology |
| CN111126482A (en) * | 2019-12-23 | 2020-05-08 | 自然资源部国土卫星遥感应用中心 | Remote sensing image automatic classification method based on multi-classifier cascade model |
| CN111415357A (en) * | 2020-03-19 | 2020-07-14 | 长光卫星技术有限公司 | Portable shadow extraction method based on color image |
| CN111598048A (en) * | 2020-05-31 | 2020-08-28 | 中国科学院地理科学与资源研究所 | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image |
| CN113139994A (en) * | 2021-04-13 | 2021-07-20 | 宁波四象径宇科技有限公司 | High-resolution optical remote sensing satellite image building height monitoring method based on angular points |
| CN113705523A (en) * | 2021-09-06 | 2021-11-26 | 青岛星科瑞升信息科技有限公司 | Layered city impervious surface extraction method based on optical and dual-polarization SAR fusion |
| CN114092815A (en) * | 2021-11-29 | 2022-02-25 | 自然资源部国土卫星遥感应用中心 | Remote sensing intelligent extraction method for large-range photovoltaic power generation facility |
| CN116258958A (en) * | 2022-12-22 | 2023-06-13 | 二十一世纪空间技术应用股份有限公司 | Building extraction method and device for homologous high-resolution images and DSM data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609726A (en) * | 2012-02-24 | 2012-07-25 | 中国科学院东北地理与农业生态研究所 | Method for classifying remote sensing images blended with high-space high-temporal-resolution data by object oriented technology |
| CN103034863A (en) * | 2012-12-24 | 2013-04-10 | 重庆市勘测院 | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction |
-
2013
- 2013-05-13 CN CN201310176487.4A patent/CN103279951B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609726A (en) * | 2012-02-24 | 2012-07-25 | 中国科学院东北地理与农业生态研究所 | Method for classifying remote sensing images blended with high-space high-temporal-resolution data by object oriented technology |
| CN103034863A (en) * | 2012-12-24 | 2013-04-10 | 重庆市勘测院 | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction |
Non-Patent Citations (1)
| Title |
|---|
| 吴浩: "基于景观指数的武汉市轨道交通沿线土地利用变化研究", 《安徽农业科学》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103839267A (en) * | 2014-02-27 | 2014-06-04 | 西安科技大学 | Building extracting method based on morphological building indexes |
| CN103839267B (en) * | 2014-02-27 | 2017-05-10 | 西安科技大学 | Building extracting method based on morphological building indexes |
| CN104794456A (en) * | 2015-05-04 | 2015-07-22 | 福建师范大学 | Remote-sensing image building detecting method based on colors and shadows |
| CN104794456B (en) * | 2015-05-04 | 2018-01-16 | 福建师范大学 | A kind of remote sensing image building detection method based on color and shade |
| CN104915925A (en) * | 2015-06-12 | 2015-09-16 | 西南科技大学 | Dry valley extraction method by preparing high-precision mask file |
| CN105528596A (en) * | 2016-02-03 | 2016-04-27 | 长江大学 | High-resolution remote sensing image building automatic extraction method and system by using shadow |
| CN105528596B (en) * | 2016-02-03 | 2018-10-09 | 长江大学 | Utilize the high-resolution remote sensing image automatic building extraction method and system of shade |
| CN106683112A (en) * | 2016-10-10 | 2017-05-17 | 中国交通通信信息中心 | High-resolution image-based road region building change extraction method |
| CN107194857A (en) * | 2017-05-23 | 2017-09-22 | 国家地理空间信息中心 | Based on high score satellite remote sensing date using land resource analysis system |
| CN107194857B (en) * | 2017-05-23 | 2020-05-12 | 国家地理空间信息中心 | Available land resource analysis system based on high-resolution satellite remote sensing data |
| CN107977968A (en) * | 2017-12-22 | 2018-05-01 | 长江勘测规划设计研究有限责任公司 | The building layer detection method excavated based on buildings shadow information |
| CN109086666A (en) * | 2018-06-29 | 2018-12-25 | 中国水利水电科学研究院 | One kind realizing the fast automatic extracting method of building and system based on Object-Oriented Method |
| CN109815894A (en) * | 2019-01-23 | 2019-05-28 | 中国石油大学(华东) | A kind of building effects extraction process method for sentry's 2A image |
| CN110390267A (en) * | 2019-06-25 | 2019-10-29 | 东南大学 | A kind of mountain landscape Building extraction method and apparatus based on high score remote sensing image |
| CN110390267B (en) * | 2019-06-25 | 2021-06-01 | 东南大学 | Mountain landscape building extraction method and device based on high-resolution remote sensing image |
| CN110796042A (en) * | 2019-10-16 | 2020-02-14 | 长江大学 | High-resolution remote sensing image building extraction method based on form transformation of offset shadow sample |
| CN110796042B (en) * | 2019-10-16 | 2023-04-21 | 长江大学 | High-resolution remote sensing image building extraction method based on offset shadow sample form transformation |
| CN110852207A (en) * | 2019-10-29 | 2020-02-28 | 北京科技大学 | Blue roof building extraction method based on object-oriented image classification technology |
| CN111126482A (en) * | 2019-12-23 | 2020-05-08 | 自然资源部国土卫星遥感应用中心 | Remote sensing image automatic classification method based on multi-classifier cascade model |
| CN111415357A (en) * | 2020-03-19 | 2020-07-14 | 长光卫星技术有限公司 | Portable shadow extraction method based on color image |
| CN111415357B (en) * | 2020-03-19 | 2023-04-07 | 长光卫星技术股份有限公司 | Portable shadow extraction method based on color image |
| CN111598048A (en) * | 2020-05-31 | 2020-08-28 | 中国科学院地理科学与资源研究所 | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image |
| CN113139994A (en) * | 2021-04-13 | 2021-07-20 | 宁波四象径宇科技有限公司 | High-resolution optical remote sensing satellite image building height monitoring method based on angular points |
| CN113139994B (en) * | 2021-04-13 | 2022-07-08 | 宁波四象径宇科技有限公司 | High-resolution optical remote sensing satellite image building height monitoring method based on angular points |
| CN113705523A (en) * | 2021-09-06 | 2021-11-26 | 青岛星科瑞升信息科技有限公司 | Layered city impervious surface extraction method based on optical and dual-polarization SAR fusion |
| CN114092815A (en) * | 2021-11-29 | 2022-02-25 | 自然资源部国土卫星遥感应用中心 | Remote sensing intelligent extraction method for large-range photovoltaic power generation facility |
| CN114092815B (en) * | 2021-11-29 | 2022-04-15 | 自然资源部国土卫星遥感应用中心 | Remote sensing intelligent extraction method for large-range photovoltaic power generation facility |
| CN116258958A (en) * | 2022-12-22 | 2023-06-13 | 二十一世纪空间技术应用股份有限公司 | Building extraction method and device for homologous high-resolution images and DSM data |
| CN116258958B (en) * | 2022-12-22 | 2023-12-05 | 二十一世纪空间技术应用股份有限公司 | Building extraction method and device for homologous high-resolution images and DSM data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103279951B (en) | 2016-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103279951B (en) | A kind of method of OO remote sensing image building and shadow extraction thereof | |
| CN105956557B (en) | A kind of sequential remote sensing image cloud covered areas domain automatic testing method of object-oriented | |
| CN103839267B (en) | Building extracting method based on morphological building indexes | |
| CN101840581B (en) | Method for extracting profile of building from satellite remote sensing image | |
| CN103632363B (en) | Object level high-resolution remote sensing image change detecting method based on Multiscale Fusion | |
| Ghaffarian | Automatic building detection based on supervised classification using high resolution Google Earth images | |
| CN107067405B (en) | Remote sensing image segmentation method based on scale optimization | |
| CN103578110B (en) | Multiband high-resolution remote sensing image dividing method based on gray level co-occurrence matrixes | |
| CN109544497A (en) | Image interfusion method and electronic equipment for transmission line faultlocating | |
| CN102855487A (en) | Method for automatically extracting newly added construction land change image spot of high-resolution remote sensing image | |
| CN106683112A (en) | High-resolution image-based road region building change extraction method | |
| Zhang et al. | The analysis of object-based change detection in mining area: A case study with pingshuo coal mine | |
| Uzar | Automatic building extraction with multi-sensor data using rule-based classification | |
| Attarzadeh et al. | Object-based building extraction from high resolution satellite imagery | |
| CN109829426A (en) | Railway construction temporary building monitoring method and system based on high score remote sensing image | |
| CN101403796A (en) | City ground impermeability degree analyzing and drawing method | |
| Zou et al. | Object based image analysis combining high spatial resolution imagery and laser point clouds for urban land cover | |
| Ji et al. | Discriminating saltcedar (Tamarix ramosissima) from sparsely distributed cottonwood (Populus euphratica) using a summer season satellite image | |
| CN110427961A (en) | A kind of rule-based and samples fusion architecture information extracting method and system | |
| Grigillo et al. | Classification based building detection from GeoEye-1 images | |
| Ujjinakoppa et al. | Multimodal and Multitemporal Spatial Data Analysis in Google Earth Engine Cloud Computing Platform to Detect Human Settlements Without Electricity: A Case Study of Bangalore City | |
| Yue et al. | The extraction of water information based on SPOT5 image using object-oriented method | |
| Forghani et al. | Object-based classification of multi-sensor optical imagery to generate terrain surface roughness information for input to wind risk simulation | |
| Li et al. | Automatic extraction of urban built-up area based on object-oriented method and remote sensing data | |
| Hashim et al. | Multi-level image segmentation for urban land-cover classifications |
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 |