CN111145157B - Road network data automatic quality inspection method based on high-resolution remote sensing image - Google Patents
Road network data automatic quality inspection method based on high-resolution remote sensing image Download PDFInfo
- Publication number
- CN111145157B CN111145157B CN201911375647.1A CN201911375647A CN111145157B CN 111145157 B CN111145157 B CN 111145157B CN 201911375647 A CN201911375647 A CN 201911375647A CN 111145157 B CN111145157 B CN 111145157B
- Authority
- CN
- China
- Prior art keywords
- road
- remote sensing
- data
- matching
- model
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- 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
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Abstract
The invention discloses a road network data automatic quality inspection method based on high-resolution remote sensing images, which automatically realizes quantitative quality inspection of road tracks and assists in updating and optimizing road data; the method comprises the following steps: 1) Breaking at the road feature points according to the road morphological features to generate a simple road section structure; 2) Constructing a general road cross section model according to road section structural features; 3) And obtaining image data block data corresponding to the road section structure. 4) And establishing an image block coordinate system, and performing template matching between the road cross section model and the image to obtain road matching points. 5) And obtaining an actual road extraction result in the image after calculating the RANSAC least square curve fitting error. 6) The actual road extraction result is matched with the road section structural similarity; 7) And (3) comprehensively evaluating by combining the similarity matching result of the road segment structure in each road, and screening out a 'problem road' for road data quality evaluation and further data modification and updating.
Description
Technical Field
The invention relates to the technical field of high-resolution remote sensing images, in particular to a road network data automatic quality inspection method based on high-resolution remote sensing images.
Background
The road is an artery and a booster for economic development, the road network is used as an important component of traffic infrastructure, the accuracy and the comprehensiveness of the data are directly related to civilian life, and the road network has important significance and value in the aspects of traffic management, urban planning, mass travel and the like. Because the road network data acquisition mode based on traditional measurement is long in period in early stage, the latest real information of the road can not be reflected in time, so that the problems of inaccurate data base of part of historical road network, incapability of estimating data quality, inconsistent new and old data and the like are caused, and analysis, decision making and data updating of a data user are affected to a certain extent.
The remote sensing image has the characteristics of wide instantaneous imaging range and real ground feature live reflection, and the roads in the image show obvious geometric texture characteristics, so that the remote sensing image is an effective technical means for assisting in acquiring real road network track information and realizing historical road network inspection and updating. The related departments and enterprises have developed a series of road network vector data quality inspection and updating works based on remote sensing images, and are applied to management of road network basic geographic data, and compared with the traditional mapping and mobile terminal navigation data acquisition modes, the efficiency and accuracy are greatly improved.
At present, quality inspection and updating work of road data are carried out based on high-resolution remote sensing images, and the road data are mainly dependent on a large amount of manual visual interpretation work, so that subjectivity is realized, time and labor are wasted, and automation is required to improve the production efficiency. In the prior art, although the remote sensing image road network extraction technology has been developed for many years, most of the remote sensing image road network extraction technology is still in a laboratory stage, and the accuracy of the remote sensing image road network extraction technology is difficult to reach the data production standard of business and commercialization. Under the condition of multiple sources (such as navigation positioning data and mobile Internet) of the road data acquisition channel, a data user is more concerned about the quality conditions such as the authenticity of the data and how to continuously update the information based on the historical data so as to ensure the consistency and timeliness of the data. In the prior art, on one hand, the traditional remote sensing image road extraction method is difficult to apply to quality inspection and update based on the existing data; on the other hand, although related researches on remote sensing ground object change detection exist in the fields of China and the like, the method is widely applied to planar ground object pattern spots, and has certain limitation on road linear track detection.
Disclosure of Invention
The invention aims to provide a road network data automatic quality inspection method based on high-resolution remote sensing images, which can automatically realize quantitative quality inspection of road track data based on the remote sensing images, complete marking and integrity evaluation of road error track parts, and form an automatic tool for road data quality inspection and auxiliary updating based on the remote sensing images so as to solve the problems in the background art.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a road network data automatic quality inspection method based on high-resolution remote sensing images comprises the following steps:
step 1): breaking at characteristic points of road data according to road morphological characteristics to form a road section structure with simple characteristics;
step 2): constructing a general road cross section model according to road section structural features;
step 3): combining the geometric positioning offset error of the remote sensing image and the road data to extract image data blocks within a certain range of the road section structure;
step 4): an X-Y coordinate system is established for the remote sensing image block in a rotating mode, and a road cross section model and the image block are subjected to template matching along the Y direction at certain intervals in the X direction, so that a series of road matching points are obtained;
step 5): calculating to obtain an actual road extraction result in the image block by using a RANSAC least squares curve fitting method;
step 6): combining the spatial relationship between the actual road extraction result and the road section structure in the image, and carrying out quick similarity matching, wherein the road section structure with lower similarity is used as a part with problems in data quality inspection;
step 7): and comprehensively evaluating the structural similarity matching results of all the road sections in each road, screening out a problem road, and analyzing the road data quality evaluation and further modifying and updating the data.
Still further, the morphological features in step 1) include 4 classes: node bending angle, track bending direction, constituent unit length, track bending degree, for a curve composed of P 0 P 1 …P i P i+1 …P n Road track data composed of points, if point P i If either of the following conditions is not satisfied, then at P i Breaking at the point to form a new constituent unit:
condition 1) the turning and bending angle of the road is too large and not smooth enough, namely +.P i-1 P i P i+1 ≤Th angle Wherein Th is angle Is a bend angle threshold;
condition 2) the point bending tendency is inconsistent with the track bending direction, namely the judgment vectorAt vectorIs the left side or the right side of the road section structure and is consistent with the bending trend of all points in the same road section structure;
condition 3):therein LTh curve A threshold value for the length of the constituent units, the threshold value representing the minimum road segment error granularity allowed by the quality check;
condition 4) track bending degree utilizationThe ratio of the actual length of the curve to the length of the straight line between the ends of the curve is measured, the greater the ratio, the higher the track complexity, i.e
Further, the specific method for constructing the universal road cross section model in the step 2) is as follows: according to different characteristics of roads of different grades in the image, the establishment of the road cross section model is divided into the following two modes:
1) Constructing a model of 0-1-0 for a lower-grade or narrower road, specifically setting the model to be m pixels wide and 3w pixels high, taking 1 as a middle value and taking 0 as two sides;
2) For a higher-grade or wider road, an m-w model is constructed, and all values take 1, w as the number of pixels calculated according to the road width value and the spatial resolution of the remote sensing image.
Furthermore, in the method for constructing the road cross section model in step 2), if the road width value information cannot be obtained, a corresponding road width initial value can be set according to the road class to be checked and the road construction engineering requirement.
Further, the specific method for extracting the corresponding image block of the road section structure in the step 3) is as follows: for the L 0 L 1 …L n The road section structure formed by the points is formed by a line segment L 0 L n Taking geometric positioning deviation Dis of the remote sensing image and road data as a buffer area radius as a center, creating a flat-head buffer area, and extracting images in the buffer area to obtain a corresponding rectangular remote sensing image block.
Furthermore, the road cross section model in the step 4) is subjected to template matching with the image block, and the specific method comprises the following steps:
step 401): establishing an X-Y coordinate system: in vectorsThe direction is X axis, anticlockwise vertical vector +.>The direction is the Y axis, the origin of coordinates is the point L 0 Translating point 0 a distance Dis in the opposite direction of the Y-axis;
step 402): in the X-Y coordinate system, a certain distance d is set at intervals in the X direction, and each interval d i Template matching is carried out on the road cross section model and the remote sensing image block along the Y direction, and certain interval X in the X axis direction is carried out i Obtaining a matching point with the largest correlation coefficient in the Y-axis direction, if the correlation coefficient value is larger than a set threshold value TH corr If the matching point is considered valid, the coordinates (x i ,y i ) Sequentially processing each interval in the X-axis direction to obtain a series of initial matching point sets with the largest correlation coefficients: (x) 0 ,y 0 ),(x 1 ,y 1 ),(x 2 ,y 2 )…(x n ,y n )。
Further, the curve equation used for the curve fitting in step 5) is:
y=a 0 +a 1 x+a 2 x 2
the matching error points are removed by adopting a RANSAC and least squares curve fitting method, so that an actual road in the image is obtained, and the specific flow is as follows:
step 501) model threshold setting: let N be the iteration number of RANSAC, the number proportion of the initial matching points meeting the curve model is RTh num The coordinate difference threshold is Th y ;
Step 502) randomly selecting n points (n is more than or equal to 3) from the initial matching point set, and performing least square curve equation fitting to obtain initial parameter value a of parabola 0 ,a 1 ,a 2 ;
Step 503) sequentially determining the coordinates (x) of each store in the initial matching point set i ,y i ) And points (x) i ,a 0 +a 1 x i +a 2 x i 2 ) Comparing if the Y-axis coordinate difference is |y i -a 0 +a 1 x i +a 2 x i 2 I, less than threshold Th y Then the point is counted as the point meeting the curve modelCentralizing;
step 504) versus step 503), if the number of points satisfying the curve model is greater than the threshold RTh num The RANSAC iteration loop is ended, and all points meeting the model are applied to least square fitting to obtain a final curve parameter A 0 ,A 1 ,A 2 。
Step 505) versus step 503), if the number of points satisfying the model is less than the proportional threshold RTh num The RANSAC iterative loop continues and steps 502-504 are repeated until the end.
Further, the specific process of road data similarity matching in step 6) is as follows:
under the X-Y coordinates, calculating and obtaining each coordinate point (X) of the actual road in the image block according to the curve model parameters curve0 ,y curve0 )...(x curvei ,y curvei ) Coordinates of points in the road section structure (x road0 ,y road0 )…(x roadi ,y roadi ) Sequentially comparing their Y-axis coordinate differences |y curvei -y roadi I and find out the maximum coordinate difference Dis max And a minimum coordinate difference Dis min If Dis max -Dis min ≤Th dis The matching degree of the road vector to be checked and the road track in the image is high in the local range, and the road vector to be checked and the road track in the image are considered to be correct tracks; otherwise, the two tracks are considered to have larger difference, and the error track is output as an error track part which is used as a part with problems in data quality inspection.
Further, the comprehensive evaluation in each road in the step 7) is specifically as follows: taking the Ratio of the correct track part in the whole road mileage in each road obtained in the step 6) as a quality index for quantitatively evaluating certain road data, wherein the calculation formula is ratio=length correct /Length total Wherein Length is correct Length for mileage of all correct track parts in a certain road total The road overall mileage length; if Ratio is less than or equal to RTh false And judging the road as a problem road and marking if the proportion of the correct track part in the road is smaller.
Compared with the prior art, the invention has the beneficial effects that:
compared with the traditional methods such as the detection of the change of the universal remote sensing image, the road network data automatic quality inspection method based on the high-resolution remote sensing image is closer to business operation and use modes in the aspect of application scenes, meanwhile, the method adopts the universal road cross section model characteristics, so that the algorithm has stronger adaptability and robustness, the adjustment of excessive model parameters caused by image chromatic aberration and the like can be avoided to a certain extent, in addition, the road network data inspection tool provided by the invention can provide quantitative reference indexes for evaluating the quality of the data, marks the suspected error places in the data, avoids the inconsistency of artificial subjectivity evaluation, and provides standardized reference basis for quality inspection and further optimization of the data.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a schematic illustration of a road track split into a series of simple road segment structures;
FIG. 3 is a schematic illustration of a road cross-sectional model;
FIG. 4 is a schematic diagram of image block extraction and its local coordinate establishment;
FIG. 5 is a schematic diagram of a road cross-sectional model matching an image block template;
fig. 6 is a schematic illustration of a quality check result "error trace" flag.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-6, in an embodiment of the present invention, a road network data automatic quality inspection method based on high resolution remote sensing images is provided, which is described by domestic high resolution two-number remote sensing images (resolution is 0.8 m) and road network track data in shape format, wherein shape road vectors have attribute information describing road "technical grade", "administrative grade" and the like, and part of roads have road width attribute, part of road missing width attribute, comprising the following steps:
step one: breaking at characteristic points of road data according to road morphological characteristics to form a road section structure with simple characteristics;
step two: constructing a general road cross section model according to road section structural features;
step three: combining the geometric positioning offset error of the remote sensing image and the road data to extract image data blocks within a certain range of the road section structure;
step four: an X-Y coordinate system is established for the remote sensing image block in a rotating mode, and a road cross section model and the image block are subjected to template matching along the Y direction at certain intervals in the X direction, so that a series of road matching points are obtained;
step five: calculating to obtain an actual road extraction result in the image block by using a RANSAC least squares curve fitting method;
step six: combining the spatial relationship between the actual road extraction result and the road section structure in the image, and carrying out quick similarity matching, wherein the road section structure with lower similarity is used as a part with problems in data quality inspection;
step seven: and comprehensively evaluating the structural similarity matching results of all the road sections in each road, screening out a problem road, and analyzing the road data quality evaluation and further modifying and updating the data.
In the above embodiment, the morphological features in the first step include 4 types: node bending angle, track bending direction, constituent unit length, track bending degree, for a curve composed of P 0 P 1 …P i P i+1 …P n Road track data composed of points, if point P i If either of the following conditions is not satisfied, then at P i Breaking at the point to form a new constituent unit:
condition 1) road turningThe bending angle is too large and is not smooth enough, namely +.P i-1 P i P i+1 ≤Th angle Wherein Th is angle Is a bend angle threshold;
condition 2) the point bending tendency is inconsistent with the track bending direction, namely the judgment vectorAt vectorIs the left side or the right side of the road section structure and is consistent with the bending trend of all points in the same road section structure;
condition 3):therein LTh curve A threshold value for the length of the constituent units, the threshold value representing the minimum road segment error granularity allowed by the quality check;
condition 4): the track bending degree is measured by the ratio of the actual length of the curve to the length of the straight line between the head and the tail of the curve, and the greater the ratio is, the higher the track complexity is, namely
In the above embodiment, in step one, at P i The threshold values in the 4 conditions of the break are respectively set as Th in combination with the related regulations of the highway construction requirements angle =120°,LTh curve =50m,RTh ratio =1.5. For a certain point P in condition 2) j Is defined by the bending tendency ofAnd->And (5) performing cross multiplication calculation of the space. Fig. 2 shows a schematic diagram of a road track split into a series of simple road segments.
In the above embodiment, the specific method for constructing the universal road cross section model in the second step is that if a certain road has road width attribute information, the road cross section model is established according to different characteristics of roads of different grades in the image in the following two ways:
1) Constructing a model of 0-1-0 for a lower-grade or narrower road (the width is smaller than 10 m), specifically setting the width of the model to be m pixels, the height to be 3w pixels, taking 1 as a middle value and taking 0 as two sides;
2) For higher-grade or wider roads (the width is more than 10 m), an m-w model is constructed, all values take 1, and w is the number of pixels calculated according to the road width value and the spatial resolution of the remote sensing image.
Wherein the model width m=9 pixels, w is the number of pixels calculated according to the actual width of the road and the image resolution (0.8 m), i.e. w=w road X 0.8. Fig. 3 shows a schematic diagram of a road cross-section model.
In the above embodiment, in the method for constructing a road cross section model in the second step, if the road width value information is missing, a corresponding road width initial value may be set according to the road class to be inspected and the road construction engineering requirement, and the lane width and the lane number are specified by referring to the "highway engineering technical standard", and the initial reference values of the road widths at each level are set as follows: the four-level highway is 3.5m, the three-level highway is 4m, the two-level highway is 6m, the expressway is 10m, and the like.
In the above embodiment, the specific method for extracting the image block corresponding to the road segment structure in the third step is as follows: for the L 0 L 1 ...L n The road section structure formed by the points is formed by a line segment L 0 L n Taking geometric positioning deviation Dis of the remote sensing image and road data as a buffer area radius as a center, creating a flat-head buffer area, and extracting images in the buffer area to obtain a corresponding rectangular remote sensing image block. In order to make the buffer area contain the roads in the image as much as possible and screen out the road track with too large geometric positioning deviation, dis=25m is set.
In the above embodiment, the template matching is performed on the road cross section model and the image block in the fourth step, which specifically includes:
step 401: establishing an X-Y coordinate system: in vectorsThe direction is X axis, anticlockwise vertical vector +.>The direction is the Y axis, the origin of coordinates is the point L 0 The point O is translated a distance Dis in the opposite direction along the Y-axis. FIG. 4 shows a schematic diagram of a method of establishing an X-Y coordinate system.
Step 402): in the X-Y coordinate system, a certain distance d is set at intervals in the X direction, and each interval d i Template matching is carried out on the road cross section model and the remote sensing image block along the Y direction, and certain interval X in the X axis direction is carried out i Obtaining a matching point with the largest correlation coefficient in the Y-axis direction, if the correlation coefficient value is larger than a set threshold value TH corr If the matching point is considered valid, the coordinates (x i ,y i ) Sequentially processing each interval in the X-axis direction to obtain a series of initial matching point sets with the largest correlation coefficients: (x) 0 ,y 0 ),(x 1 ,y 1 ),(x 2 ,y 2 )…(x n ,y n ). In order to increase the number of matching points in the X-Y coordinate system, in this example, d=1 pixels are selected as intervals in the X direction, and template matching is performed on the road cross-section model and the remote sensing image block along the Y direction. In order to improve the template matching efficiency, a rapid normalization correlation coefficient calculation is adopted, a specific calculation method is shown in the following formula, wherein m and n represent the length and width of the template, and g ij For pixel values in the template, g' i,j The pixel value is the corresponding pixel value in the remote sensing image block.
In the above embodiment, the road cross-section model in the fourth step is template-matched with the image block, and the correlation coefficient threshold Th is set corr =0.6 as a reference value to screen valid template matching points. FIG. 5 is a template webSchematic of the formulation.
In the above embodiment, the curve equation adopted for the curve fitting in the fifth step is:
y=a 0 +a 1 x+a 2 x 2
the matching error points are removed by adopting a RANSAC and least squares curve fitting method, so that an actual road in the image is obtained, and the specific flow is as follows:
step 501) model threshold setting: let N be the iteration number of RANSAC, the number proportion of the initial matching points meeting the curve model is RTh num The coordinate difference threshold is Th y . The parameter threshold iteration number N is 500, and the matching point set meets the point number proportion RTh of the curve model num =0.8;
Step 502) randomly selecting n points (n is more than or equal to 3) from the initial matching point set, and performing least square curve equation fitting to obtain initial parameter value a of parabola 0 ,a 1 ,a 2 . The least squares curve equation fit selected point number reference value n=6 is given here.
Step 503) sequentially determining the coordinates (x) of each store in the initial matching point set i ,y i ) And points (x) i ,a 0 +a 1 x i +a 2 x i 2 ) Comparing if the Y-axis coordinate difference is |y i -a 0 +a 1 x i +a 2 x i 2 I, less than threshold Th y The point is counted into a set of points that satisfy the curve model. Th is given here y Is 3.
Step 504) versus step 503), if the number of points satisfying the curve model is greater than the threshold RTh num The RANSAC iteration loop is ended, and all points meeting the model are applied to least square fitting to obtain a final curve parameter A 0 ,A 1 ,A 2 。
Step 505) versus step 503), if the number of points satisfying the model is less than the proportional threshold RTh num The RANSAC iterative loop continues and steps 502-504 are repeated until the end.
In the above embodiment, the specific process of road data similarity matching in the sixth step is:
step 601) for the curve model parameters A calculated in step five 0 、A 1 、A 2 Under the X-Y local coordinate system, sequentially calculating coordinate points (X curve0 ,y curve0 )...(x curvei ,y curvei ) Wherein y is curvei =A 0 +A 1 ×x curvei +A 2 ×x curvei 2
Step 602) converting the track of the road section structure into a series of coordinate points (X) in an X-Y coordinate system by adopting a vector rasterization mode road0 ,y road0 )...(x roadi ,y roadi )。
Step 603) comparing their Y-axis coordinate differences |y in sequence under an X-Y coordinate system curvei -y roadi I and find out the maximum coordinate difference Dis max And a minimum coordinate difference Dis min If Dis max -Dis min ≤Th dis The road vector to be checked is considered to be consistent with the road track in the image in the local range, and the road vector to be checked is considered to be a correct track; otherwise, the two trajectories are not consistent and are considered as 'error trajectories'. To improve the adaptability of the model and reduce excessive parameter adjustment, an adaptive coordinate difference threshold is set as Th y =H block 5, wherein H block The number of pixels is the height of the remote sensing image block corresponding to the road section structure. In order to locate the position of the error track and facilitate subsequent modification, the section of the 'error track' is output as a vector file as part of the problem in the data quality inspection.
In the above embodiment, the comprehensive evaluation in each road in the seventh step is specifically: taking the Ratio of the correct track part in each road to the whole road mileage, which is obtained in the step six, as a quality index for quantitatively evaluating certain road data, wherein the calculation formula is ratio=length correct /Length total Wherein Length is correct Length for mileage of all correct track parts in a certain road total The road is wholeBody mileage length. If Ratio is less than or equal to RTh false (RTh is set here false =0.4), i.e. the proportion of the correct track portion in the road is smaller, the road is determined to be a "problem road" for identification, and the road shape attribute field is marked for further manual review and subsequent data modification and editing.
To sum up: according to the road network data automatic quality inspection method based on the high-resolution remote sensing image, road data with larger problems in data quality can be clearly positioned according to the problem road identified by the method; according to the Ratio field value marked in the road shape attribute field, the quality of the road track can be quantitatively evaluated, and the track position with problems is positioned by combining with the output error track file, so as to assist the further modification and improvement of the road data.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should be covered by the protection scope of the present invention by making equivalents and modifications to the technical solution and the inventive concept thereof.
Claims (5)
1. The road network data automatic quality inspection method based on the high-resolution remote sensing image is characterized by comprising the following steps of:
step 1): breaking at characteristic points of road data according to road morphological characteristics to form a road section structure with simple characteristics;
step 2): constructing a general road cross section model according to road section structural features;
the specific method for constructing the universal road cross section model in the step 2) comprises the following steps: according to different characteristics of roads of different grades in the image, the establishment of the road cross section model is divided into the following two modes:
1) Constructing a model with the width of less than 10 meters, namely setting the model to be m pixels wide and 3w pixels high, taking 1 as a middle value and taking 0 as two sides;
2) For roads with the width of more than 10 meters, an m-w model is constructed, and all values take 1, w as the number of pixels calculated according to the road width value and the spatial resolution of the remote sensing image;
in the method for constructing the road cross section model in the step 2), if the road width value information cannot be obtained, setting a corresponding road width initial value according to the road grade to be checked and the road construction engineering requirement;
step 3): combining the geometric positioning offset error of the remote sensing image and the road data, and extracting an image data block within the range of the geometric positioning offset error Dis of the road section structure;
step 4): an X-Y coordinate system is established for the remote sensing image block in a rotating mode, and a road cross section model and the image block are subjected to template matching along the Y direction by taking a distance d as an interval in the X direction, so that a series of road matching points are obtained;
the road cross section model in the step 4) is matched with the image block in a template mode, and the specific method comprises the following steps:
step 401): establishing an X-Y coordinate system: in vectorsThe direction is X axis, anticlockwise vertical vector +.>The direction is the Y axis, the origin of coordinates is the point L 0 Translating a point O a distance Dis in the opposite direction of the Y-axis;
step 402): in the X-Y coordinate system, the distance d is set at intervals in the X direction, and each interval d i Template matching is carried out on the road cross section model and the remote sensing image block along the Y direction, and certain interval X in the X axis direction is carried out i Obtaining a matching point with the largest correlation coefficient in the Y-axis direction, if the correlation coefficient value is larger than a set threshold value TH corr If the matching point is considered valid, the coordinates (x i ,y i ) Sequentially processing the intervals in the X-axis direction to obtain a correlation coefficient greater than a set threshold value TH corr Is set of initial matching points (x 0 ,y 0 ),(x 1 ,y 1 ),(x 2 ,y 2 )…(x n ,y n ) The method comprises the steps of carrying out a first treatment on the surface of the In the X-Y coordinate system, d=1 pixels are selected in the X direction to increase the number of matching points, and template matching is carried out on the road cross section model and the remote sensing image block along the Y direction; in order to improve the template matching efficiency, a rapid normalization correlation coefficient calculation is adopted, a specific calculation method is shown in the following formula, wherein m and n represent the length and width of the template, and g ij For pixel values in the template, g' i,j The pixel value is the corresponding pixel value in the remote sensing image block;
step 5): calculating to obtain an actual road extraction result in the image block by using a RANSAC least squares curve fitting method;
step 6): combining the spatial relationship between the actual road extraction result and the road section structure in the image, and carrying out quick similarity matching, wherein the road section structure with lower similarity is used as a part with problems in data quality inspection; the road section structure with lower similarity refers to: dis within the extracted image data block max –Dis min >Th dis Is the case in (2); the specific process of road data similarity matching is as follows:
under the X-Y coordinates, calculating and obtaining each coordinate point (X) of the actual road in the image block according to the curve model parameters curve0 ,y curve0 )...(x curvei ,y curvei ) Coordinates of points in the road section structure (x road0 ,y road0 )…(x roadi ,y roadi ) Sequentially comparing their Y-axis coordinate differences |y curvei -y roadi I and find out the maximum coordinate difference Dis max And a minimum coordinate difference Dis min If Dis max -Dis min ≤Th dis The matching degree of the road vector to be checked and the road track in the image is considered to be high in the range of the extracted image data block, and the road vector to be checked and the road track in the image are considered to be correct tracks; otherwise, the error track is considered to be the error track, and the error track part is output as a part with problems in data quality inspection;
step 7): and comprehensively evaluating the structural similarity matching results of all the road sections in each road, screening out a problem road, and analyzing the road data quality evaluation and further modifying and updating the data.
2. The method for automatically inspecting road network data quality based on high-resolution remote sensing images according to claim 1, wherein the morphological features in the step 1) comprise 4 types: node bending angle, track bending direction, constituent unit length, track bending degree, for a curve composed of P 0 P 1 …P i P i+1 …P n Road track data composed of points, if point P i If either of the following conditions is not satisfied, then at P i Breaking at the point to form a new constituent unit:
condition 1) the turning and bending angle of the road is too large and not smooth enough, namely +.P i-1 P i P i+1 ≤Th angle Wherein Th is angle Is a bend angle threshold;
condition 2) the point bending tendency is inconsistent with the track bending direction, namely the judgment vectorIn vector->Is the left side or the right side of the road section structure and is consistent with the bending trend of all points in the same road section structure;
condition 3):therein LTh curve A threshold value for the length of the constituent units, the threshold value representing the minimum road segment error granularity allowed by the quality check;
condition 4) the degree of track curvature is measured by the ratio of the actual length of the curve to the length of the straight line between the ends of the curve, the greater the ratio, the greater the track complexity, i.e
3. The automatic road network data quality inspection method based on high-resolution remote sensing images as claimed in claim 1, wherein the specific method for extracting the road segment structure corresponding image blocks in the step 3) is as follows: for the L 0 L 1 …L n The road section structure formed by the points is formed by a line segment L 0 L n Taking geometric positioning deviation Dis of the remote sensing image and road data as a buffer area radius as a center, creating a flat-head buffer area, and extracting images in the buffer area to obtain a corresponding rectangular remote sensing image block.
4. The method for automatically checking the quality of road network data based on high-resolution remote sensing images according to claim 1, wherein the curve equation adopted by the curve fitting in the step 5) is as follows:
y=a 0 +a 1 x+a 2 x 2
the matching error points are removed by adopting a RANSAC and least squares curve fitting method, so that an actual road in the image is obtained, and the specific flow is as follows:
step 501) model threshold setting: let N be the iteration number of RANSAC, the point number proportion threshold value meeting the curve model in the initial matching point set be RTh num The coordinate difference threshold is Th y ;
Step 502) randomly selecting n points n is more than or equal to 3 from the initial matching point set, and performing least square curve equation fitting to obtain initial parameter value a of parabola 0 ,a 1 ,a 2 ;
Step 503) sequentially for the coordinates (x) of each point in the initial set of matching points i ,y i ) And points (x) i ,a 0 +a 1 x i +a 2 x i 2 ) Comparing if the Y-axis coordinate difference is |y i -a 0 +a 1 x i +a 2 x i 2 I, less than threshold Th y The point is counted into a point set meeting the curve model;
step 504) versus step 503), if the curve model is satisfiedThe number of points is greater than the proportional threshold RTh num The RANSAC iteration loop is ended, and all points meeting the model are applied to least square fitting to obtain a final curve parameter A 0 ,A 1 ,A 2 ;
Step 505) versus step 503), if the number of points satisfying the model is less than the proportional threshold RTh num The RANSAC iterative loop continues and steps 502-504 are repeated until the end.
5. The automatic road network data quality inspection method based on high-resolution remote sensing images as claimed in claim 1, wherein the comprehensive evaluation in each road in the step 7) is performed by the following specific steps: taking the Ratio of the correct track part in the whole road mileage in each road obtained in the step 6) as a quality index for quantitatively evaluating certain road data, wherein the calculation formula is ratio=length correct /Length total Wherein Length is correct Length for mileage of all correct track parts in a certain road total The road overall mileage length; if Ratio is less than or equal to RTh false And judging the road as a problem road and marking.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911375647.1A CN111145157B (en) | 2019-12-27 | 2019-12-27 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911375647.1A CN111145157B (en) | 2019-12-27 | 2019-12-27 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111145157A CN111145157A (en) | 2020-05-12 |
CN111145157B true CN111145157B (en) | 2023-08-04 |
Family
ID=70520840
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911375647.1A Active CN111145157B (en) | 2019-12-27 | 2019-12-27 | Road network data automatic quality inspection method based on high-resolution remote sensing image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111145157B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111915669B (en) * | 2020-08-03 | 2024-04-05 | 北京吉威空间信息股份有限公司 | Total quantity control-based land survey linear ground object pattern spotting method |
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 |
CN112396612B (en) * | 2020-11-16 | 2021-05-28 | 自然资源部国土卫星遥感应用中心 | Vector information assisted remote sensing image road information automatic extraction method |
CN112733231B (en) * | 2020-12-31 | 2023-01-31 | 同济大学建筑设计研究院(集团)有限公司 | Road three-dimensional model generation method and device, computer equipment and storage medium |
CN114705148B (en) * | 2022-04-03 | 2023-10-24 | 国交空间信息技术(北京)有限公司 | Road bending point detection method and device based on secondary screening |
CN116295444B (en) * | 2023-05-17 | 2023-10-20 | 国网山东省电力公司日照供电公司 | Navigation method, system, terminal and storage medium for field operation |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271928A (en) * | 2018-09-14 | 2019-01-25 | 武汉大学 | 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 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
IT201600084942A1 (en) * | 2016-08-12 | 2018-02-12 | Paolo Andreucci | System of analysis, measurement and automatic classification of road routes and relative method of operation. |
-
2019
- 2019-12-27 CN CN201911375647.1A patent/CN111145157B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271928A (en) * | 2018-09-14 | 2019-01-25 | 武汉大学 | 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 |
Non-Patent Citations (5)
Title |
---|
吴涛 ; 向隆刚 ; 龚健雅 ; .路网更新的轨迹-地图匹配方法.测绘学报.2017,(04),第507-515页. * |
张勇 ; 谷正气 ; 刘水长 ; 罗伦 ; 李程 ; .基于高分辨率遥感影像的城市群交通路网监测技术研究.遥感技术与应用.2015,(01),第76-81页. * |
李程 ; .遥感技术在农村公路核查中的应用研究.测绘与空间地理信息.2018,(05),第48-49、52页. * |
胡传文 ; 冯媛媛 ; .基于贝叶斯网络的高分辨率遥感影像城区道路检测方法.测绘通报.2012,(09),第51-54页. * |
袁胜古 ; 阳柯 ; 熊国清 ; 盛光晓 ; 邓曾 ; 米素娟 ; 徐昊 ; .基于规则库的农村公路基础空间数据质检系统设计及实现.城市勘测.2018,(05),第11-15、19页. * |
Also Published As
Publication number | Publication date |
---|---|
CN111145157A (en) | 2020-05-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111145157B (en) | Road network data automatic quality inspection method based on high-resolution remote sensing image | |
Galantucci et al. | Advanced damage detection techniques in historical buildings using digital photogrammetry and 3D surface anlysis | |
CN110619258B (en) | Road track checking method based on high-resolution remote sensing image | |
CN112906694B (en) | Reading correction system and method for transformer substation inclined pointer instrument image | |
CN115424232B (en) | Method for identifying and evaluating pavement pit, electronic equipment and storage medium | |
CN109886939B (en) | Bridge crack detection method based on tensor voting | |
CN110133639B (en) | Dowel bar construction quality detection method | |
CN104408463B (en) | High-resolution construction land pattern spot identification method | |
CN104834806A (en) | Joint roughness coefficient size effect sampling representativeness evaluation method | |
CN105069395B (en) | Roadmarking automatic identifying method based on Three Dimensional Ground laser scanner technique | |
CN113487722B (en) | Automatic concrete member detection method based on three-dimensional laser scanning method | |
CN101915570A (en) | Vanishing point based method for automatically extracting and classifying ground movement measurement image line segments | |
CN115560690B (en) | Structure integral deformation analysis method based on three-dimensional laser scanning technology | |
CN113360587B (en) | Land surveying and mapping equipment and method based on GIS technology | |
CN108830317B (en) | Rapid and fine evaluation method for joint attitude of surface mine slope rock mass based on digital photogrammetry | |
CN111242223B (en) | Street space quality evaluation method based on streetscape image multi-feature fusion | |
Chen et al. | A novel image-based approach for interactive characterization of rock fracture spacing in a tunnel face | |
Li et al. | A deep learning-based indoor acceptance system for assessment on flatness and verticality quality of concrete surfaces | |
CN116486289A (en) | Gas pipeline high-consequence area identification method driven by multi-source data and knowledge | |
Lang et al. | Pavement cracking detection and classification based on 3d image using multiscale clustering model | |
CN113420670A (en) | Environment-friendly supervision method for changing power transmission and transformation line migration based on high-resolution remote sensing | |
CN114004950A (en) | Intelligent pavement disease identification and management method based on BIM and LiDAR technology | |
CN113421236A (en) | Building wall surface water leakage apparent development condition prediction method based on deep learning | |
CN112989453A (en) | BIM-based holographic deformation information extraction method | |
CN113344866A (en) | Point cloud comprehensive precision evaluation method |
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 |