CN109271928A - A kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image - Google Patents
A kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image Download PDFInfo
- Publication number
- CN109271928A CN109271928A CN201811074936.3A CN201811074936A CN109271928A CN 109271928 A CN109271928 A CN 109271928A CN 201811074936 A CN201811074936 A CN 201811074936A CN 109271928 A CN109271928 A CN 109271928A
- Authority
- CN
- China
- Prior art keywords
- road
- road network
- vector
- edge
- navigation
- 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
- G06V20/182—Network patterns, e.g. roads or rivers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of road network automatic update methods verified based on the fusion of vector road network with high score remote sensing image, and this method is first by history road vectors, new period navigation road network vector and remote sensing image matching；Then successively history road vectors and new period navigation road network vector are merged using thick matching and essence matched method, finds out the road of unchanged road and doubtful variation；Edge feature, spectral signature and vegetation characteristics building multiple features are extracted from high score remote sensing image later and verify model, whether the road for verifying doubtful variation is road, it is thus found that link change；Road finally will not changed according to geometrical characteristic and variation road carries out road network fusion, obtain updated road network.The present invention excavates out position, geometry, topology, semantic information from vector road network, extracts road network in conjunction with the scene characteristic of road in high-resolution remote sensing image, has stronger practicability and higher accuracy rate.
Description
Technical field
The present invention relates to remote sensing image applied technical field, more particularly, to being a kind of based on the fusion of vector road network and high score
The road network automatic update method of remote sensing image verifying.
Background technique
Road network is the tie between connection area.With the development of Urbanization in China, people transport road traffic
Defeated demand constantly increases, to push the fast development of China's road construction.The extracted with high accuracy of road network information with more
Newly there is vital effect to traffic administration, urban planning, self-navigation.The quick obtaining of road factor data and update
Oneself becomes the vital task of China's Fundamental Geographic Information System construction.It is distant from high-resolution with the development of Remote Sensing Image Processing Technology
The research hotspot that road information is increasingly becoming domestic and foreign scholars is automatically extracted in sense image, but there is no a set of robustness stronger at present
Road automatic Extraction Algorithm, the artificial interpretation based on high-resolution remote sensing image is still the main means of road updated core elements,
Have the shortcomings that productivity is low, is unable to satisfy the demand that road network quickly updates.
On high-resolution remote sensing image, road shows as the narrow band target with spectrum homogeneity, road boundary
It is high-visible with pavement markers, these features are based on, many road extraction algorithms have been derived.These method for extracting roads are big
Cause can be divided into 4 classes: method, mathematics morphology and the Active contour of method, knowledge based rule based on classification.It is based on
The method of classification is the construction features space such as geometrical characteristic, radiation feature and textural characteristics using road by pixel or homogeneous
Object is divided into road or non-rice habitats, and support vector machines, markov random file, neural network are most common classification of road sides
It is most important problem in classification method that method, road and non-rice habitats target, which are obscured,.Booming deep learning method in recent years
It is also applicable in road extraction, essence is also to be classified to obtain road target to pixel, but since road has width
Narrow attribute gradually loses in the down-sampled middle information of deep neural network, therefore ineffective to road especially country road.
The method of knowledge based rule is regular based on expertise building road extraction and is combined to extract road, road
Tracking is one of the method for extracting roads of most typical knowledge based rule.Mathematical Morphology Method is calculated using variform
Son carries out Morphological scale-space to remote sensing image, chaff interferent and noise is filtered out while enhancing road long and narrow feature, based on number
The method for extracting roads for learning morphology operations usually extracts road bone with combinations such as other methods such as Image Segmentation, edge detection
Frame.The cardinal principle of active contour model is by construction energy functional, and under the driving of energy function minimum value, contour curve is gradually
It is approached to the edge of examined object, is finally partitioned into target, snake model and Level Set Models are more common in road extraction
The active contour model arrived, has a wide range of applications in semi-automatic road extraction.These road extraction algorithms push significantly
Road network automatically updates the development of technology, but none of these methods perfect can adapt to road field changeable under Road network extraction task
Scape, first is that because influenced by imaging factors: sensor differences, spectrum and spatial resolution difference, light differential these at
As factor leads to the variation of road radiation feature, is automatically extracted for road and bring difficulty；Second is that road oneself factor: rich in details
On rich high-resolution remote sensing image, road shows as the aggregate of a variety of atural objects, such as vehicle, road sign, trade line, trade
Tree etc., so that there is very big heterogeneity, while road object and neighbouring atural object have biggish feature inside road element again
Correlation, this makes automatic method for extracting roads be difficult to accurately recognize road object；Third is that environmental factor: by shade and other
Atural object blocks, and the automation road extraction task based on remote sensing image becomes more difficult, and the auxiliary for seeking other data is come
Extracting road is inexorable trend.
Based on the demand that Fundamental Geographic Information System road network quickly updates, and takes into account and utilize road in high-resolution remote sensing image
Existing difficult point is extracted, present invention introduces the period of history road vectors auxiliary channels in map of navigation electronic and geographical national conditions generaI investigation
Road network extracts, the road network priori with Up-to-date state, reliability for making full use of map of navigation electronic and history road vectors to provide
Information is constructed a set of road network verified based on the fusion of vector road network with high score remote sensing image and automatically updates technical system, from
Existing road extraction, new added road update and verify two levels and study road network automatic update method progressively.Method is divided into
Vector data is merged with remote sensing image matching, navigation road network with history road vectors, the variation road based on multi-feature evidence is tested
Card connects four steps with road network, has stronger practicability and higher accuracy rate.
Summary of the invention
For the problems raised in the background art, the present invention provides a kind of based on the fusion of vector road network and high score remote sensing image
The road network automatic update method of verifying.The technical solution of the present invention is as follows:
A kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image, including following step
It is rapid:
Step 1, vector data and remote sensing image matching, wherein vector data navigation road network and will be gone through according to image capturing range
History road vectors carry out cutting acquisition；
Step 2, the navigation road network after registration is merged with history road vectors, and obtained and history road vectors phase
Matched navigation road network section, is considered as and does not change road；Fail in navigation road network with the matched section of history road vectors, be considered as
It may changed road；
Step 3, may be become using edge feature, spectral signature, vegetation characteristics multi-feature evidence model in step 2
The road of change is verified, further separation variation road and non-rice habitats；
Step 4, the variation road in step 3 is connected with the road that do not change in step 2 according to road segment segment geometrical characteristic,
Constitute road network.
Further, in step 1 using using by the way of human-computer interaction by vector data and remote sensing image matching, specific mistake
Journey are as follows: cut navigation road network and history road vectors to obtain vector data according to remote sensing image range, respectively from vector
Several same places are selected in data and remote sensing image by hand, whole affine transformation is carried out to road network vector data and corrects arrow
Amount.
Further, the specific implementation of the step 2 is as follows,
Step 2.1, using history road vectors as reference, Duan Weiyi section of vector between two nodes is defined, by road
Section generates buffer area as size to have a lot of social connections width；
Step 2.2, in buffer area, navigation road network vector sum history road vectors are slightly matched first, matching according to
According to being to judge whether there is navigation road network vector across this buffer area, thick matching is completed if passing through, otherwise it fails to match；
Step 2.3, for passing through thick matched navigation road network vector, using length similitude SlenWith directional similarity Sang
Two constraint conditions carry out smart matching, SlenAnd SangCalculation formula respectively such as formula (1)-(2):
Sang=cos θ (2)
Wherein, LOSMComprehensive for the length of all vector sections of navigation road network in buffer area, L is in current history road vectors
The length of vector section, θ are the angle of navigation road network vector and history road vectors；
Final index of similarity uses length similitude SlenWith directional similarity SangWeighted sum characterization, such as formula (3)
It is shown,
S=λ Sang+(1-λ)·Slen (3)
Wherein S value is smaller, then navigation road network is more close with history road vectors；
Step 2.4, if TmatchFor smart matching threshold, if S < Tmatch, then navigation road network is by essence matching, for road of navigating
By the matched section of essence in network data, it is considered as and does not change road, it is subsequent no longer to process；For failing and going through in navigation road network
The matched section of history road vectors, being considered as may changed road.
Further, step 3 specific implementation is as follows,
Step 3.1, for not passing through the matched section of essence in navigation road network data, by section building buffer area, and according to
The range of buffer area obtains image slice；
Step 3.2, road is judged whether it is using multi-feature evidence model to each section, in which:
1. edge feature evidence model: carrying out canny edge detection to the high score remote sensing image in buffer area；Calculate each side
The direction of edge, the total length at the corresponding edge of statistics all directions, it is believed that the corresponding direction of extreme length is the section principal direction；
The marginal density in principal direction is calculated as edge feature evidence Redge, calculation formula is shown in (4):
Redge=Lang/L (4)
Wherein ang is section principal direction, LangFor the total length at the corresponding edge of principal direction, L is road section length；If Tedge
For efficient frontier characteristic threshold value, if Redge< Tedge, then probability existing for road is related in linear positive to edge feature value, if
Redge≥Tedge, then it is assumed that road exists；The road existing probability such as formula (5) of edge feature evidence model:
Wherein αedgeFor the corresponding probability right of edge feature, ε is predefined small probability value；
2. spectral signature evidence model: enabling pixel within the scope of buffer area in the gray value standard difference of each wave band is std (DN),
TspecFor effective spectral signature threshold value, then spectral signature Rspec=std (DN)/Tspec；If Rspec< 1, then it is assumed that existing for road
Probability and RspecIn negative linear correlation；If Rspec>=1, then it is assumed that road exists；The road existing probability such as formula of spectral evidence model
(6):
αspecFor the corresponding probability right of spectral signature；
3. vegetation characteristics evidence model: normalized differential vegetation index NDVI is calculated to image in buffer area first, using threshold value
The method of segmentation extracts roughly the vegetation in buffer area；If DplantFor the corresponding deflection of vegetation object in navigation section buffer area
Degree set, DroadFor current road segment principal direction corresponding angle, TplantFor orientation angle difference threshold value, vegetation characteristics RplantCalculating it is public
Formula is shown in formula (7):
Rplant=std (Dplant-Droad)/Tplant (7)
Work as RplantWhen < 1, probability and R existing for roadplantIt is negatively correlated；Work as RplantWhen >=1, it is believed that vegetation direction and road
Road moves towards unrelated, and the probability for non-rice habitats object is larger at this time；The road existing probability such as formula (8) of vegetation evidence model:
4. calculating separately out edge evidence model, spectral evidence model and the corresponding road existing probability of vegetation evidence model
Afterwards, by asking the maximum of three probability density to obtain the road existing probability based on more evidence models, such as formula (7):
P=max { Pedge, Pspec, Pplant} (9)
After the road existing probability based on more evidence models is calculated, if P >=0.5, then it is assumed that be road, as change
Road, otherwise it is assumed that being non-rice habitats.
Further, middle buffer area is dimensioned to have a lot of social connections for twice in step 3.1.
The present invention is using high-resolution remote sensing image, history road vectors, new period navigation road network vector as input number
According to source, navigation road network, the position in history road vectors data, geometry, topology, semantic information and high-resolution shadow are comprehensively utilized
Scene characteristic as in completes road network element and automates extraction task in conjunction with real road structure priori knowledge.Present invention tool
There are stronger practicability and higher accuracy rate, feature is:
(1) extraction for using navigation road network vector sum history road vectors Liang Zhong data supporting road network, makes full use of and leads
Airway net has the advantage of reliability with timeliness and history road vectors, directly extracts not changed road, only
It handles increase/reduction road area may occur, substantially increases road network extraction efficiency；
(2) for changed section, using three kinds of edge feature, spectral signature, vegetation characteristics features as evidence
Verify whether section is road, it is not only effective to the apparent road of feature, also have preferably to the road segment segment for being blocked or being disturbed
Effect.
Detailed description of the invention
Fig. 1 is the road network automatic update method flow chart based on the fusion of vector road network with the verifying of high score remote sensing image.
Fig. 2 is that navigation road network and history road vectors carry out thick matched schematic diagram.
Fig. 3 is road network connection schematic diagram.
Specific embodiment
The present invention provides a kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image,
To understand convenient for those of ordinary skill in the art and implementing the present invention, present invention work is further retouched in detail with reference to the accompanying drawing
It states, it should be understood that implementation example described herein is only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
Technical solution of the present invention flow chart as shown in Figure 1, specifically includes the following steps:
Step 1: vector data and remote sensing image matching.If vector is with image, there are larger position deviations, can mention to road
Take and have an impact, this programme by the way of human-computer interaction by vector data and remote sensing image matching, with reduce vector data with
The deviation of remote sensing image.It is cut navigation road network and history road vectors to obtain vector data according to image capturing range first,
It selects several same places by hand from vector sum image respectively, whole affine transformation is carried out to road network vector.After guaranteeing
Continuous road extraction process is normally carried out, and the offset distance for triggering affine transformation is determined by the buffer area radius of road extraction, and
Buffer area radius is equal to the sum of road half width and acceptable registration error.After navigation road network is superimposed with image, road of the same name
Road position deviation distance is greater than the buffering radius, then needs to carry out manual correction.
Step 2: the navigation road network after registration is merged with history road vectors, and is obtained and history road vectors
The navigation road network section to match, is considered as and does not change road；Fail in navigation road network with the matched section of history road vectors, depending on
For possible changed road.Specific steps are as follows:
1) using history road vectors as reference, Duan Weiyi section of vector between two nodes is defined, by section Yi Lu
Wide width is that size generates buffer area；Specific practice is all node transverse and longitudinals seat for calculating the section history road vectors first
Then boundary of the target maximin as boundary rectangle adds a width to be the buffer area of width, such as Fig. 2 institute to rectangle
Show.
2) in buffer area, navigation road network and history road vectors are slightly matched first, matching foundation is that judgement is
It is no to there is navigation road network to pass through this buffer area, thick matching is completed if passing through, otherwise it fails to match；As shown in Fig. 2, navigation road network
In, the section of black meets thick matching condition, and grey section is not met.
3) for passing through thick matched navigation road network, using length similitude SlenWith directional similarity SangTwo constraint items
Part carries out smart matching, wherein length similitude SlenFor characterize navigation road network and history road vectors section length it is close
Degree, directional similarity SangFor filtering out and the biggish navigation road network of local history road vectors direction difference, calculation formula point
Not such as formula (1)-(2):
Sang=| cos θ | (2)
Wherein, LOSMComprehensive for the length of all vector sections of navigation road network in buffer area, L is in current history road vectors
The length of vector section, θ are the angle of navigation road network and history road vectors.If should be noted that in buffer area comprising multiple
Navigation road network vector section then calculates similarity by vector section.Final index of similarity uses length similitude SlenWith direction phase
Like property SangWeighted sum characterization, as shown in formula (3):
S=λ Sang+(1-λ)·Slen (3)
In formula (3), λ is the balance factor of adjustment length similitude and directional similarity weight, and in general λ takes 0.5, can
It finely tunes according to the actual situation.S value is smaller, then navigation road network is more close with history road vectors.
4) T is setmatchFor smart matching threshold, if S < Tmatch, then navigation road network passes through essence matching.For navigation road network data
In pass through the matched section of essence, it is believed that do not change, it is subsequent no longer to process, for failing in navigation road network and history road
The section of vector matching, is considered as possible changed road, and step 3 is only handled not through the matched section of essence.TmatchValue
For empirical value, need to determine by many experiments.General recommendations takes initial value at [0.5-0.8], wrong matched, then increases
TmatchValue；If there is void matched, reduce TmatchValue.
Step 3: testing for road is changed based on edge feature, spectral signature, vegetation characteristics multi-feature evidence model
Card.Specific steps are as follows:
1) for, not by the matched section of essence, by section building buffer area, (buffer size is set in navigation road network data
Twice is set to have a lot of social connections), and image slice is obtained according to the range of buffer area.It should be noted that if the length in the section is greater than
Certain length (it is recommended that being set as 1000 pixels), then need section splitting into multistage and process, in order to avoid road spectral differences in section
It is different excessive.
2) road is judged whether it is using multi-feature evidence model to each section, in which:
1. edge feature evidence model: in unobstructed situation, path link often with having significant edge with tropism feature, because
And edge feature can be used as the strong evidence of road verifying.Circular are as follows: to the high score remote sensing image in buffer area
Carry out canny edge detection；Calculate the direction (value range be [0- π) at each edge), the corresponding edge of statistics all directions
Total length, it is believed that the corresponding direction of extreme length is the section principal direction；The marginal density in principal direction is calculated as edge spy
Levy evidence Redge, calculation formula is shown in (4):
Redge=Lang/L (4)
Wherein ang is the corresponding angle of section principal direction, LangFor the total length at the corresponding edge of principal direction, L is that section is long
Degree.If TedgeFor efficient frontier characteristic threshold value, if Redge< Tedge, then probability existing for road and edge feature value are in linear positive
Correlation, if Redge≥Tedge, then it is assumed that road exists.The road existing probability such as formula (5) of edge evidence model:
Wherein αedgeFor the corresponding probability right of edge feature, general value is 0.33, can be adjusted according to the actual situation: if
Section edge clear can then increase αedgeValue.ε is predefined small probability value, general value 0.001.
2. spectral signature evidence model: ideal road surface spectral signature variation is smaller, and carries on the back with neighbouring non-rice habitats
There are biggish gray differences for scape.However, road is blocked by other background atural objects or oneself factor leads to SPECTRAL DIVERSITY in section
Larger situation is generally existing, thus the evidence that spectral signature can not disappear as road.But when navigation road network data are corresponding
When spectral signature difference is smaller in section, spectral signature can be used as strong evidence existing for road.
Enabling pixel within the scope of buffer area in the gray value standard difference of each wave band is std (DN), TspecFor effective spectral signature
Threshold value, then spectral signature Rspec=std (DN)/Tspec.If Rspec< 1, then it is assumed that probability and R existing for roadspecIn negative
It is related；If Rspec>=1, then it is assumed that road exists.The road existing probability such as formula (6) of spectral evidence model:
αspecFor the corresponding probability right of spectral signature, general value is 0.33, can be adjusted according to the actual situation: if section
Face spectral signature is uniform, then can increase αspecValue.ε is predefined small probability value, general value 0.001.
3. vegetation characteristics evidence model: road vegetation is atural object common in road scene, in reality by shade tree,
Greenbelt and greenery patches are constituted；From the point of view of geometric shape, road vegetation is usually close to road, and wherein Road greenbelt usually connects along road
Continuous distribution, and have and then surrounded and the geometric shape of formation rule by road with the consistent direction character of trend of road, greenery patches；
On image, road vegetation then reflects the special curve of spectrum.It is found by the above analysis, vegetation object in road scene
Direction character can be used as the evidence for examining road presence or absence.Circular are as follows: image in buffer area is calculated first
Normalized differential vegetation index NDVI extracts roughly the vegetation in buffer area using the method for Threshold segmentation；If DplantFor section of navigating
The corresponding orientation angle set of vegetation object, D in buffer arearoadFor current road segment principal direction corresponding angle, TplantFor deflection
Spend poor threshold value, vegetation characteristics RplantCalculation formula see formula (7):
Rplant=std (Dplant-Droad)/Tplant (7)
Work as RplantWhen < 1, probability and R existing for roadplantIt is negatively correlated；Work as RplantWhen >=1, it is believed that vegetation direction and road
Road moves towards unrelated, and the probability for non-rice habitats object is larger at this time.The road existing probability such as formula (8) of vegetation evidence model:
αplantFor the corresponding probability right of vegetation characteristics, general value is 0.33, can be adjusted according to the actual situation: if section
There is shade tree in both sides, then can increase αplantValue.ε is predefined small probability value, general value 0.001.
4. calculating separately out edge evidence model, spectral evidence model and the corresponding road existing probability of vegetation evidence model
Afterwards, by asking the maximum of three probability density to obtain the road existing probability based on more evidence models, such as formula (7):
P=max { Pedge, Pspec, Pplant} (9)
After the road existing probability based on more evidence models is calculated, if P >=0.5, then it is assumed that be road, as change
Road, otherwise it is assumed that being non-rice habitats.
Step 4: road network connection.Road will be changed according to road segment segment geometrical characteristic and do not change road and connect, composition road
Net.The geometrical characteristic of constraint section connection specifically includes that end-point distances, linkage section direction and existing section direction difference.Based on several
The section connection schematic diagram of what feature is as shown in Figure 3.Tu Zhong branch section E11E12And E21E22To be broken section, need and main road
Section C1C2It is attached.According to distance restraint, with endpoint E12Centered on, R be radius detected to obtain candidate connecting node C1With
C2(R is determined according to road width, is traditionally arranged to be twice and is had a lot of social connections)；By central point E12With candidate connecting node C1And C2Connect respectively
It connects, and calculates linkage section and horizontal angleWithCorresponding node is as connecting node when taking minimum angle
And it is attached.Solid black lines E in Fig. 312C2The section connection result of expression final choice, and black dotted lines E12C1It then indicates not
The candidate linkage section selected.
Specific embodiment described herein is only an example for the spirit of the invention.The neck of technology belonging to the present invention
The technical staff in domain can make various modifications or additions to the described embodiments or replace by a similar method
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (5)
1. a kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image, which is characterized in that
The following steps are included:
Step 1, vector data and remote sensing image matching, wherein vector data is according to image capturing range by navigation road network and history road
Road vector carries out cutting acquisition；
Step 2, the navigation road network after registration is merged with history road vectors, and obtains and matches with history road vectors
Navigation road network section, be considered as and do not change road；Fail in navigation road network with the matched section of history road vectors, be considered as possibility
Changed road；
Step 3, using edge feature, spectral signature, vegetation characteristics multi-feature evidence model to possible changed in step 2
Road is verified, further separation variation road and non-rice habitats；
Step 4, the variation road in step 3 is connected according to road segment segment geometrical characteristic with the road that do not change in step 2, is constituted
Road network.
2. a kind of road network based on the fusion of vector road network and the verifying of high score remote sensing image according to claim 1 is automatically more
New method, it is characterised in that: by vector data and remote sensing image matching by the way of using human-computer interaction in step 1, specifically
Process are as follows: cut navigation road network and history road vectors to obtain vector data according to remote sensing image range, respectively from arrow
Several same places are selected by hand on amount data and remote sensing image, and whole affine transformation is carried out to road network vector data and corrects arrow
Amount.
3. a kind of road network based on the fusion of vector road network and the verifying of high score remote sensing image according to claim 1 is automatically more
New method, it is characterised in that: the specific implementation of the step 2 is as follows,
Step 2.1, using history road vectors as reference, define Duan Weiyi section of vector between two nodes, by section with
The width that has a lot of social connections is that size generates buffer area；
Step 2.2, in buffer area, navigation road network vector sum history road vectors are slightly matched first, matching foundation is
Navigation road network vector is judged whether there is across this buffer area, thick matching is completed if passing through, otherwise it fails to match；
Step 2.3, for passing through thick matched navigation road network vector, using length similitude SlenWith directional similarity SangTwo
Constraint condition carries out smart matching, SlenAnd SangCalculation formula respectively such as formula (1)-(2):
Sang=cos θ (2)
Wherein, LOSMComprehensive for the length of all vector sections of navigation road network in buffer area, L is vector in current history road vectors
The length of section, θ are the angle of navigation road network vector and history road vectors；
Final index of similarity uses length similitude SlenWith directional similarity SangWeighted sum characterization, as shown in formula (3),
S=λ Sang+(1-λ)·Slen (3)
Wherein S value is smaller, then navigation road network is more close with history road vectors；
Step 2.4, if TmatchFor smart matching threshold, if S < Tmatch, then navigation road network is by essence matching, for navigation road network number
By the matched section of essence in, it is considered as and does not change road, it is subsequent no longer to process；For failing in navigation road network and history road
The section of road vector matching, being considered as may changed road.
4. a kind of road network based on the fusion of vector road network and the verifying of high score remote sensing image according to claim 1 is automatically more
New method, it is characterised in that: step 3 specific implementation is as follows,
Step 3.1, for, not by the matched section of essence, constructing buffer area by section, and according to buffering in navigation road network data
The range in area obtains image slice；
Step 3.2, road is judged whether it is using multi-feature evidence model to each section, in which:
1. edge feature evidence model: carrying out canny edge detection to the high score remote sensing image in buffer area；Calculate each edge
Direction, the total length at the corresponding edge of statistics all directions, it is believed that the corresponding direction of extreme length is the section principal direction；It calculates
Marginal density in principal direction is as edge feature evidence Redge, calculation formula is shown in (4):
Redge=Lang/L (4)
Wherein ang is section principal direction, LangFor the total length at the corresponding edge of principal direction, L is road section length；If TedgeIt is effective
Edge feature threshold value, if Redge< Tedge, then probability existing for road is related in linear positive to edge feature value, if Redge≥
Tedge, then it is assumed that road exists；The road existing probability such as formula (5) of edge feature evidence model:
Wherein αedgeFor the corresponding probability right of edge feature, ε is predefined small probability value；
2. spectral signature evidence model: enabling pixel within the scope of buffer area in the gray value standard difference of each wave band is std (DN), Tspec
For effective spectral signature threshold value, then spectral signature Rspec=std (DN)/Tspec；If Rspec< 1, then it is assumed that probability existing for road
With RspecIn negative linear correlation；If Rspec>=1, then it is assumed that road exists；The road existing probability such as formula (6) of spectral evidence model:
αspecFor the corresponding probability right of spectral signature；
3. vegetation characteristics evidence model: normalized differential vegetation index NDVI is calculated to image in buffer area first, using Threshold segmentation
Method extract the vegetation in buffer area roughly；If DplantFor the corresponding orientation angle collection of vegetation object in navigation section buffer area
It closes, DroadFor current road segment principal direction corresponding angle, TplantFor orientation angle difference threshold value, vegetation characteristics RplantCalculation formula see
Formula (7):
Rplant=std (Dplant-Droad)/Tplant (7)
Work as RplantWhen < 1, probability and R existing for roadplantIt is negatively correlated；Work as RrlantWhen >=1, it is believed that walked with road in vegetation direction
To unrelated, the probability at this time for non-rice habitats object is larger；The road existing probability such as formula (8) of vegetation evidence model:
4. after calculating separately out edge evidence model, spectral evidence model and the corresponding road existing probability of vegetation evidence model,
Maximum by seeking three probability density obtains the road existing probability based on more evidence models, such as formula (7):
P=max { Pedge, Pspec, Pplant} (9)
After the road existing probability based on more evidence models is calculated, if P >=0.5, then it is assumed that be road, as change road
Road, otherwise it is assumed that being non-rice habitats.
5. a kind of road network based on the fusion of vector road network and the verifying of high score remote sensing image according to claim 4 is automatically more
New method, it is characterised in that: middle buffer area is dimensioned to have a lot of social connections for twice in step 3.1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811074936.3A CN109271928B (en) | 2018-09-14 | 2018-09-14 | Road network updating method based on vector road network fusion and remote sensing image verification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811074936.3A CN109271928B (en) | 2018-09-14 | 2018-09-14 | Road network updating method based on vector road network fusion and remote sensing image verification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109271928A true CN109271928A (en) | 2019-01-25 |
| CN109271928B CN109271928B (en) | 2021-04-02 |
Family
ID=65189059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811074936.3A Active CN109271928B (en) | 2018-09-14 | 2018-09-14 | Road network updating method based on vector road network fusion and remote sensing image verification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109271928B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110084107A (en) * | 2019-03-19 | 2019-08-02 | 安阳师范学院 | A kind of high-resolution remote sensing image method for extracting roads and device based on improvement MRF |
| CN110543885A (en) * | 2019-08-13 | 2019-12-06 | 武汉大学 | method for interactively extracting high-resolution remote sensing image road and generating road network |
| CN111145157A (en) * | 2019-12-27 | 2020-05-12 | 国交空间信息技术（北京）有限公司 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
| CN112115817A (en) * | 2020-09-01 | 2020-12-22 | 国交空间信息技术（北京）有限公司 | Remote sensing image road track correctness checking method and device based on deep learning |
| CN112509453A (en) * | 2020-12-14 | 2021-03-16 | 广西壮族自治区自然资源遥感院 | Scenic spot live-action guide picture electronic guide method and system based on mobile equipment |
| CN113065594A (en) * | 2021-04-01 | 2021-07-02 | 中科星图空间技术有限公司 | Road network extraction method and device based on Beidou data and remote sensing image fusion |
| CN113076387A (en) * | 2021-04-08 | 2021-07-06 | 北京星天地信息科技有限公司 | Road network matching method and device based on multi-element map matching |
| CN113177456A (en) * | 2021-04-23 | 2021-07-27 | 西安电子科技大学 | Remote sensing target detection method based on single-stage full convolution network and multi-feature fusion |
| CN113408457A (en) * | 2021-06-29 | 2021-09-17 | 西南交通大学 | Road information intelligent extraction method combining high-resolution image and video image |
| CN114064835A (en) * | 2021-11-18 | 2022-02-18 | 中国公路工程咨询集团有限公司 | Multisource vector road network updating method based on change point detection and electronic equipment |
| CN114078335A (en) * | 2020-07-30 | 2022-02-22 | 通用汽车环球科技运作有限责任公司 | System and method for detecting road edges by fusing aerial images and telemetry data |
| CN116860906A (en) * | 2023-09-05 | 2023-10-10 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
| CN112115817B (en) * | 2020-09-01 | 2024-06-07 | 国交空间信息技术（北京）有限公司 | Remote sensing image road track correctness checking method and device based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101364259A (en) * | 2008-04-09 | 2009-02-11 | 武汉大学 | Method for extracting road various information of multi-level knowledge driven panchromatic remote sensing image |
| KR101476172B1 (en) * | 2013-07-15 | 2014-12-24 | (주)시터스 | Method for Automatic Change Detection Based on Areal Feature Matching in Different Network Datasets |
| CN106778605A (en) * | 2016-12-14 | 2017-05-31 | 武汉大学 | Remote sensing image road net extraction method under navigation data auxiliary |
-
2018
- 2018-09-14 CN CN201811074936.3A patent/CN109271928B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101364259A (en) * | 2008-04-09 | 2009-02-11 | 武汉大学 | Method for extracting road various information of multi-level knowledge driven panchromatic remote sensing image |
| KR101476172B1 (en) * | 2013-07-15 | 2014-12-24 | (주)시터스 | Method for Automatic Change Detection Based on Areal Feature Matching in Different Network Datasets |
| CN106778605A (en) * | 2016-12-14 | 2017-05-31 | 武汉大学 | Remote sensing image road net extraction method under navigation data auxiliary |
Non-Patent Citations (2)
| Title |
|---|
| DEREN L 等: "Automatic change detection of geo-spatial data from imagery", 《GEO-SPATIAL INFORMATION SCIENCE》 * |
| 徐阳 等: "基于道路网矢量数据的遥感影像道路损毁检测", 《测绘通报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110084107A (en) * | 2019-03-19 | 2019-08-02 | 安阳师范学院 | A kind of high-resolution remote sensing image method for extracting roads and device based on improvement MRF |
| CN110543885A (en) * | 2019-08-13 | 2019-12-06 | 武汉大学 | method for interactively extracting high-resolution remote sensing image road and generating road network |
| CN110543885B (en) * | 2019-08-13 | 2022-03-04 | 武汉大学 | Method for interactively extracting high-resolution remote sensing image road and generating road network |
| CN111145157A (en) * | 2019-12-27 | 2020-05-12 | 国交空间信息技术（北京）有限公司 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
| CN111145157B (en) * | 2019-12-27 | 2023-08-04 | 国交空间信息技术（北京）有限公司 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
| CN114078335A (en) * | 2020-07-30 | 2022-02-22 | 通用汽车环球科技运作有限责任公司 | System and method for detecting road edges by fusing aerial images and telemetry data |
| CN114078335B (en) * | 2020-07-30 | 2024-04-12 | 通用汽车环球科技运作有限责任公司 | System and method for detecting road edges by fusing aerial images and telemetry data |
| CN112115817B (en) * | 2020-09-01 | 2024-06-07 | 国交空间信息技术（北京）有限公司 | Remote sensing image road track correctness checking method and device based on deep learning |
| CN112115817A (en) * | 2020-09-01 | 2020-12-22 | 国交空间信息技术（北京）有限公司 | Remote sensing image road track correctness checking method and device based on deep learning |
| CN112509453A (en) * | 2020-12-14 | 2021-03-16 | 广西壮族自治区自然资源遥感院 | Scenic spot live-action guide picture electronic guide method and system based on mobile equipment |
| CN113065594A (en) * | 2021-04-01 | 2021-07-02 | 中科星图空间技术有限公司 | Road network extraction method and device based on Beidou data and remote sensing image fusion |
| CN113076387B (en) * | 2021-04-08 | 2022-09-20 | 北京星天地信息科技有限公司 | Road network matching method and device based on multi-element map matching |
| CN113076387A (en) * | 2021-04-08 | 2021-07-06 | 北京星天地信息科技有限公司 | Road network matching method and device based on multi-element map matching |
| CN113177456B (en) * | 2021-04-23 | 2023-04-07 | 西安电子科技大学 | Remote sensing target detection method based on single-stage full convolution network and multi-feature fusion |
| CN113177456A (en) * | 2021-04-23 | 2021-07-27 | 西安电子科技大学 | Remote sensing target detection method based on single-stage full convolution network and multi-feature fusion |
| CN113408457A (en) * | 2021-06-29 | 2021-09-17 | 西南交通大学 | Road information intelligent extraction method combining high-resolution image and video image |
| CN114064835B (en) * | 2021-11-18 | 2023-05-26 | 中国公路工程咨询集团有限公司 | Multi-source vector road network updating method based on change point detection and electronic equipment |
| CN114064835A (en) * | 2021-11-18 | 2022-02-18 | 中国公路工程咨询集团有限公司 | Multisource vector road network updating method based on change point detection and electronic equipment |
| CN116860906A (en) * | 2023-09-05 | 2023-10-10 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
| CN116860906B (en) * | 2023-09-05 | 2023-11-28 | 高德软件有限公司 | Track generation method, track generation device, track generation equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109271928B (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109271928A (en) | A kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image | |
| CN106778605B (en) | Automatic remote sensing image road network extraction method under assistance of navigation data | |
| Zhang | Towards an operational system for automated updating of road databases by integration of imagery and geodata | |
| Ünsalan et al. | A system to detect houses and residential street networks in multispectral satellite images | |
| CN107690840B (en) | Unmanned plane vision auxiliary navigation method and system | |
| Shao et al. | Application of a fast linear feature detector to road extraction from remotely sensed imagery | |
| CN108445480A (en) | Mobile platform based on laser radar adaptively extends Target Tracking System and method | |
| CN109740483A (en) | A kind of rice growing season detection method based on deep-neural-network | |
| CN108376247A (en) | Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection | |
| CN112163622B (en) | Global and local fusion constrained aviation wide-baseline stereopair line segment matching method | |
| CN109389062A (en) | Utilize the method for High Resolution Spaceborne SAR image zooming-out lake land and water cut-off rule | |
| Li et al. | A novel method for urban road damage detection using very high resolution satellite imagery and road map | |
| Yang et al. | A pattern‐based approach for matching nodes in heterogeneous urban road networks | |
| Gerke et al. | Graph-supported verification of road databases | |
| CN112017243A (en) | Medium visibility identification method | |
| CN110084107A (en) | A kind of high-resolution remote sensing image method for extracting roads and device based on improvement MRF | |
| CN109543498A (en) | A kind of method for detecting lane lines based on multitask network | |
| Sun et al. | Building outline extraction from aerial imagery and digital surface model with a frame field learning framework | |
| CN108615452A (en) | A kind of unknown method for extracting roads based on people's wheel paths point multi-resolution hierarchy | |
| CN104463896A (en) | Image corner point detection method and system based on kernel similar region distribution characteristics | |
| Sohn et al. | Sequential modelling of building rooftops by integrating airborne LiDAR data and optical imagery: preliminary results | |
| JP2005241886A (en) | Extraction method of changed area between geographical images, program for extracting changed area between geographical images, closed area extraction method and program for extracting closed area | |
| Haverkamp et al. | Complementary methods for extracting road centerlines from IKONOS imagery | |
| Baumgartner et al. | Multi-scale road extraction using local and global grouping criteria | |
| Crommelinck | Automating image-based cadastral boundary mapping |
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 |