CN116543310B - Road line extraction method based on Voronoi diagram and kernel density - Google Patents
Road line extraction method based on Voronoi diagram and kernel density Download PDFInfo
- Publication number
- CN116543310B CN116543310B CN202310793429.XA CN202310793429A CN116543310B CN 116543310 B CN116543310 B CN 116543310B CN 202310793429 A CN202310793429 A CN 202310793429A CN 116543310 B CN116543310 B CN 116543310B
- Authority
- CN
- China
- Prior art keywords
- road
- points
- track
- density
- data
- 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
- 238000000605 extraction Methods 0.000 title claims abstract description 48
- 238000010586 diagram Methods 0.000 title claims abstract description 30
- 239000013598 vector Substances 0.000 claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims description 64
- 230000008569 process Effects 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 8
- 238000004140 cleaning Methods 0.000 claims description 8
- 238000009499 grossing Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 13
- 238000004364 calculation method Methods 0.000 abstract description 9
- 238000004458 analytical method Methods 0.000 description 14
- 238000009826 distribution Methods 0.000 description 9
- 230000001419 dependent effect Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000011166 aliquoting Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000012732 spatial analysis Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a road line extraction method based on Voronoi diagrams and kernel density, which adopts a deformation kernel density algorithm to assign a density value to vertexes, namely track points, in a vector triangular network instead of assigning values on grid pixel layers, so that the characteristics of the kernel density algorithm are maintained, most vector information is not lost after rasterization, the input and output in the whole extraction flow are vector data, only a rasterization calculation mode is used, the capability of processing interference data is enhanced, the algorithm difficulty is reduced, the problem that the existing road extraction method has poor adaptability to data, poor processing effect on track data deviation, internal detail structures of roads, different road data density differences and the like, and the problem that the accurate extraction of road boundaries and central lines is difficult is solved, the processing capability of the interference data is enhanced, the influence of parameter setting on the extraction result is reduced, and the algorithm difficulty of extracting the boundaries through the triangular network is simplified.
Description
Technical Field
The invention relates to the technical field of geographic information systems and spatial data analysis and processing, in particular to a road line extraction method based on a Voronoi diagram and kernel density.
Background
The existing numerous road line extraction methods can be divided into two extraction methods based on grid image vectorization and vector triangular network. Both methods have respective limitations and advantages, the former is capable of weakening the density influence of sparse distribution elements in a certain area, so that the influence of interference data is reduced, and the extraction result of an independent road boundary is smooth and accurate; the former is limited in that the treatment of the road area of the high and low density area cannot be considered at the same time. The method has the advantages that the relation among the single elements is enhanced, the density difference of the pavement areas has little influence on the result, and the topology of the result is good; the latter is limited in that due to the existence of interference data, accurate boundary line extraction cannot be performed through a simple algorithm, and complicated intervention conditions and decision trees are required to be manually formulated; the road surface extraction result is only a rough surface area, and small burrs exist.
In the process of extracting the road line, the track data has two characteristics, namely noise exists in the data, and the density difference of the track data between different road areas is large. The method is greatly limited by the quality and the quantity of data, the extraction result under the track data with high quality and low noise is obviously superior to the data with lower quality, the adaptability to the data is not strong, and particularly, the processing effect on the conditions such as track data deviation, the internal detail structure of a road, the difference of data density of different levels of roads and the like is not good, and the accurate extraction of the road boundary and the center line is difficult. The rasterization processing is relatively more dependent on the quantity of data, and the vector triangle network is more dependent on the quality of the data, so that the adaptability to the data is not strong, the processing effect on the conditions of track data deviation, the internal detail structure of a road, the density difference of different levels of road data and the like is not good, and the problem that the road boundary and the central line are difficult to accurately extract is solved.
Writing in a patent with the application publication number of CN108961403A, classifying Delaunay triangles by applying a main road identification model to original road route data in an open source street map, and extracting an open source street map main road by applying a Delaunay triangle network and a seed point region growing algorithm, wherein the method needs human intervention, and the complexity of the method is increased; it is written in the patent with application publication number CN112148812a that, by extracting the road center line from the double-line road data by using the Thiessen polygon and then predicting and correcting the road center line according to the road shape, the method needs to use the double-line road data, which cannot improve timeliness, and has the problems of poor adaptability to data, poor processing effect on the conditions such as track data deviation, internal detail structure of the road, different level road data density differences, and the like, and difficulty in accurately extracting the road boundary and the center line.
Disclosure of Invention
Based on the above, the invention provides a road line extraction method based on Voronoi diagrams and kernel density, which solves the problems that the existing road extraction method has poor adaptability to data, poor processing effects on the conditions such as track data deviation, internal detail structures of roads, different-grade road data density differences and the like, and is difficult to accurately extract road boundaries and central lines.
The technical scheme of the invention is as follows:
a road line extraction method based on Voronoi diagrams and kernel density comprises the following steps:
step one: preprocessing data;
step two: establishing a Delaunay triangle network, and identifying non-road edges by adopting a deformation kernel density algorithm;
the deformation kernel density algorithm formula is:
k is a kernel density function, h is a smoothing parameter, and h >0;
step three: cleaning the external points of the road;
step four: secondarily constructing a Delaunay triangle network, and extracting a road surface;
step five: further processing the obtained road surface profile and optimizing the road boundary;
step six: and extracting skeleton lines, namely extracting central lines by using a skeletonizing method.
Preferably, in the first step, the specific data preprocessing step is as follows:
removing the track points of the positioning information and the attribute information recording errors through logic judgment of the attribute information;
visualizing the track points to obtain a track line result, namely obtaining a track line element;
and identifying obvious vectors and topologies of the track line elements in the electronic map and errors which do not accord with the actual vehicle running track elements, and further cleaning data.
Preferably, in the first step, the track points for removing the positioning information and the recording error of the attribute information through the logic judgment of the attribute information include the following track points:
A. points where the attribute and positioning information are incomplete or empty;
B. locating points with discrete points with overlarge distances;
C. the distance between track points in adjacent time is overlarge and exceeds the theoretical speed per hour;
D. discrete points with overlarge distances and overlarge time intervals in a track point set of the same vehicle;
E. repetition time and repeatedly located repetition points within the same vehicle track.
Preferably, in the second step, a Delaunay triangle network is established, and the specific steps of marking the non-road edge by adopting a deformation kernel density algorithm are as follows:
combining all track points to establish a Delaunay triangle network;
calculating deformation kernel density estimation values of all vertexes, namely deformation kernel density estimation values of the track points, in a Delaunay triangle network;
and taking the ratio of the density values of the deformation kernels at the two side points of each side in the Delaunay triangular network as the density change rate, judging the size of the value of the density change rate and the set threshold, and marking the sides with the values larger than the threshold, namely the left side and the right side, as non-road sides.
Preferably, in the third step, the specific step of cleaning the external points of the road is as follows:
judging the number of non-road sides marked in adjacent sides of the track points, and identifying the number as an external point of the road if the number is greater than a set constraint threshold;
after all the road outer points are cleaned, an outer contour polygon formed by the road inner points, namely the road surface contour, is left.
Preferably, in the fourth step, a Delaunay triangle network is constructed secondarily, and the specific steps of extracting the road mask are as follows:
secondarily constructing a Delaunay triangle network by using the internal points of the road, and removing the long-side triangle by using an integral long-side constraint criterion to obtain a rough road contour consisting of a plurality of short-side triangles;
and merging the left short-side triangles into a polygon, namely the road surface.
Preferably, in the fifth step, the obtained road surface profile is further processed, and the specific steps of optimizing the road boundary are as follows:
namely, removing some tiny polygons which are independent of the outside of the road surface according to the size of the area;
and (3) smoothing the extracted boundary by using a vector processing tool to reduce small burrs, and obtaining a final road boundary result after optimization.
Preferably, in the sixth step, the skeleton line is extracted by using a skeletonizing method, that is, the specific step of extracting the center line is as follows:
and generating a binarization grid image by the extracted road polygon, and extracting a skeleton line, namely a road center line, by using a skeletonizing method.
The algorithm for extracting the central line of the road surface by rasterization is simple and convenient, but is very dependent on the extraction result of the former, and if the extracted road polygon has high coverage and each road is continuous, the central line result achieves the same effect.
Preferably, in the step six, the skeletonizing method refers to a process of searching for any thick line-shaped object with a width larger than one pixel on the raster image, deleting the contour pixel, retaining the skeleton pixel, and recording the position as a break point output vector line.
Compared with the prior art, the invention has the beneficial effects that:
the method adopts the deformation kernel density algorithm to assign the density value to the vertex, namely the track point, in the vector triangular network instead of assigning the value to the grid pixel layer, so that the characteristics of the kernel density algorithm are maintained, most vector information is not lost after rasterization, the input and output in the whole extraction process are vector data, only a rasterization calculation mode is used, the capability of processing disturbance data is enhanced, the algorithm difficulty is reduced, the problems that the adaptability to data is not strong, the processing effect is poor on the conditions such as track data deviation, the internal detail structure of a road, the density difference of different levels of road data and the like in the current road extraction method, the road boundary and the central line are difficult to accurately extract are solved, the processing capability of the disturbance data is enhanced, the influence of parameter setting on the extraction result is reduced, and the algorithm difficulty of extracting the boundary through the triangular network is simplified.
Drawings
FIG. 1 is a flow chart of a road line extraction method based on Voronoi diagrams and kernel density according to an embodiment of the invention;
FIG. 2 is a comparative chart showing the process flow comparison of three methods according to the embodiment of the present invention;
FIG. 3 is a schematic diagram of a core density estimation of a wire element in an embodiment of the invention;
FIG. 4 is a schematic diagram showing a comparison of a density estimation method based on grid pixels in an embodiment of the present invention;
FIG. 5 is a schematic diagram of skeletonization and contouring in an embodiment of the present invention;
FIG. 6 is a schematic diagram of a Delaunay triangulation network in an embodiment of the invention;
FIG. 7 is a schematic illustration of a Voronoi diagram in an embodiment of the invention;
FIG. 8 is a schematic representation of a Voronoi diagram density analysis in an embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The vehicle-mounted GPS track data is one of the data with highest occurrence frequency in the traffic geographic research, is a complete record of the running path of the vehicle, and contains rich road information. The track data recorded by the GPS equipment carried on the vehicle are widely distributed and large in quantity, almost all vehicles with the equipment can be used as data sources, and positioning data can be directly interacted with the ground through satellite equipment; the broad, real-time, inexpensive nature of its possession is advantageous over conventional data sources. For the acquired trace data of a large number of GPS records, which belongs to the most original data, a user needs to perform a series of processes such as denoising, sorting, aggregation and the like on the GPS trace data, so that the road distribution condition on the vector layer can be obtained, and the geometric characteristics of the road network and the actual running state and rule of the road traffic are presented.
The existing road line extraction mainly comprises two aspects: road surfaces, namely road boundaries and road centerlines, wherein road surface extraction methods are mainly divided into two categories:
1. generating a grid image from the track data based on a kernel density estimation theory, and extracting vector lines from the grid image by using a computer image processing technology to obtain a road surface;
2. the method comprises the steps of carrying out abrupt edge recognition in a Delaunay triangle network by judging the density change rate of vertexes at two sides of an edge, namely track points, in the triangle network, and carrying out long edge recognition by using an integral long edge constraint rule, thereby extracting road boundaries from a vehicle track line set by a boundary expansion method for the abrupt edge and the long edge.
The method for extracting the road center line mainly comprises the following steps:
1. the map is reversely updated by utilizing navigation data, and a K-Means clustering algorithm is utilized to dynamically generate a road line from vehicle track data by combining a Gaussian model;
2. short side identification is carried out by using a Delaunay triangle network model, the rough outline surface of the road is extracted after the small triangles are combined to obtain the road boundary, and the road center line is extracted by using a secondarily constructed encryption triangle network;
3. short side identification is carried out by using a Delaunay triangle network model, the rough outline surface of the road is extracted after the small triangles are combined to obtain the road boundary, and the road center line is extracted by using a secondarily constructed encryption triangle network;
4. and (5) carrying out vector extraction on the center line after rasterizing the track data.
The vast majority of the technical schemes have high requirements on original data, have strong dependence, are greatly affected by factors such as original track data quality, quantity, different area density differences, interference data and the like, the extracted result under the track data with high quality and low noise is obviously superior to the data with lower quality, the adaptability to the data is not strong, and particularly, the processing effect on the conditions such as track data deviation, the internal detail structure of a road, different-level road data density differences and the like is not good, and the road boundary and the center line are difficult to accurately extract. The rasterization process is relatively more dependent on the amount of data and the vector triangle network is more dependent on the quality of the data; the generation of grid image vectorization can greatly reduce human intervention and decision to be made in boundary line extraction, and a vector triangular network algorithm can well process areas with different densities.
Examples:
based on the two methods, the method takes two types of Voronoi diagrams and kernel density estimation theory as supports, and considers how to combine different advantages of the methods, a new road characteristic line extraction method is explored to reduce the dependence on data and the complexity of an algorithm and an extraction flow, and the problems that the existing road line extraction method is poor in adaptability to data, poor in processing effect on conditions such as track data deviation, road internal detail structures and different-level road data density differences, and difficult to accurately extract road boundaries and central lines are solved.
As shown in fig. 1 to 2, the present embodiment discloses a road line extraction method based on Voronoi diagrams and kernel density, which includes the following steps:
step one: preprocessing data;
the method comprises the following specific steps:
removing the track points of the positioning information and the attribute information recording errors through logic judgment of the attribute information;
visualizing the track points to obtain a track line result, namely obtaining a track line element;
and identifying obvious vectors, topologies and other errors of the track line elements in the electronic map, wherein the errors do not accord with the actual vehicle running track, and further cleaning data.
Step two: establishing a Delaunay triangle network, and identifying non-road edges by adopting a deformation kernel density algorithm;
the deformation kernel density algorithm formula is:
where K is a kernel density function, typically K is a symmetric probability density function, e.g. a normally distributed density function such as a gaussian kernel function, h is a smoothing parameter called bandwidth, h >0;
the method comprises the following specific steps:
combining all track points to establish a Delaunay triangle network;
calculating deformation kernel density estimation values of all vertexes, namely deformation kernel density estimation values of the track points, in a Delaunay triangle network;
and taking the ratio of the density values of the deformation kernels at the two side points of each side in the Delaunay triangular network as the density change rate, judging the size of the value of the density change rate and the set threshold, and marking the sides with the values larger than the threshold, namely the left side and the right side, as non-road sides.
Step three: cleaning the external points of the road;
the method comprises the following specific steps:
judging the number of non-road sides marked in adjacent sides of the track points, and identifying the number as an external point of the road if the number is greater than a set constraint threshold;
after all the road outer points are cleaned, an outer contour polygon formed by the road inner points, namely the road surface contour, is left.
Step four: secondarily constructing a Delaunay triangle network, and extracting a road surface;
the method comprises the following specific steps:
secondarily constructing a Delaunay triangle network by using the internal points of the road, and removing the long-side triangle by using an integral long-side constraint criterion to obtain a rough road contour consisting of a plurality of short-side triangles;
and merging the left short-side triangles into a polygon, namely the road surface.
Step five: further processing the obtained road surface profile and optimizing the road boundary;
the method comprises the following specific steps:
namely, removing some tiny polygons which are independent of the outside of the road surface according to the size of the area;
and (3) smoothing the extracted boundary by using a vector processing tool to reduce small burrs, and obtaining a final road boundary result after optimization.
Step six: extracting skeleton lines, namely extracting center lines by using a skeletonizing method;
the method comprises the following specific steps:
and generating a binarization grid image by the extracted road polygon, and extracting a skeleton line, namely a road center line, by using a skeletonizing method.
The algorithm for extracting the central line of the road surface by rasterization is simple and convenient, but is very dependent on the extraction result of the former, and if the extracted road polygon has high coverage and each road is continuous, the central line result achieves the same effect.
The vector method in the existing road extraction always takes vector data as a processing source and an output source in the extraction step; and converting the track points into raster data by using a raster method, and finally vectorizing to obtain a vector result. The method has the advantages that the integrity of data information is guaranteed to the greatest extent, the complex vector data is simplified into the processing of grid pixels, and the kernel density in the existing rasterization extraction method is estimated to be assigned to the grid pixels;
in the road line method provided by the invention, the deformation kernel density algorithm is adopted, the density value is assigned to the vertex, namely the track point, in the vector triangle network instead of assigning values on the grid pixel level, the characteristics of the kernel density algorithm are reserved, most vector information is not lost after rasterization, the input and output in the whole extraction flow are vector data, the capacity of processing interference data is enhanced by using a rasterization calculation mode only, the algorithm difficulty is reduced, the problems that the adaptability to data is not strong, the processing effect is poor, the processing capacity of accurately extracting road boundaries and central lines are difficult to realize, the processing capacity of the interference data is enhanced, the influence of parameter setting on the extraction result is reduced, and the algorithm difficulty of extracting the boundaries through the triangle network is simplified are solved.
Preferably, in the first step, the track points for removing the positioning information and the recording error of the attribute information through the logic judgment of the attribute information include the following track points:
A. points where the attribute and positioning information are incomplete or empty;
B. locating points with discrete points with overlarge distances;
C. the distance between track points in adjacent time is overlarge and exceeds the theoretical speed per hour;
D. discrete points with overlarge distances and overlarge time intervals in a track point set of the same vehicle;
E. repetition time and repeatedly located repetition points within the same vehicle track.
Preferably, in the step six, the skeletonizing method refers to a process of searching for any thick line-shaped object with a width larger than one pixel on the raster image, deleting its contour pixels, retaining its skeleton pixels, and recording its position as a break point output vector line.
The road line extraction method provided by the embodiment of the invention solves the problems that the existing road extraction method has poor adaptability to data, poor processing effect on the conditions such as track data deviation, road internal detail structure, different grade road data density difference and the like, and is difficult to accurately extract road boundaries and central lines, enhances the processing capacity of interference data, reduces the influence of parameter setting on extraction results, and simplifies the algorithm difficulty of extracting boundaries through a triangular network. By combining qualitative and quantitative analysis results, the method changes the nuclear density estimation algorithm supported by the prior grid pixels into vector point data density calculation in the technical process, absorbs the technical characteristics of the rasterization and vectorization methods under the combined support of the Voronoi diagram and the nuclear density theory, can keep the advantages of the original method in the extraction effect, and improves the comprehensive performance of the accuracy and the integrity of the result compared with the prior method.
The working principle of the invention is as follows:
the method adopts the deformation kernel density algorithm to assign the density value to the vertex, namely the track point, in the vector triangular network instead of assigning the value to the grid pixel layer, so that the characteristics of the kernel density algorithm are maintained, most vector information is not lost after rasterization, the input and the output in the whole extraction process are vector data, only the rasterization calculation mode is used, the capability of processing interference data is enhanced, and meanwhile, the algorithm difficulty is reduced.
The nuclear density estimation is a non-parametric density analysis method based on a grid network. Specifically, x1 … xn is n sample points independently distributed in F, and the probability density estimation function is F (x), and then the kernel density estimation function is:
where K is a kernel density function, typically K is a symmetric probability density function, e.g. a normally distributed density function such as a gaussian kernel function, h is a smoothing parameter, called bandwidth (h > 0).
Specifically, to nuclear density estimation of elements in a two-dimensional space, firstly, establishing a grid network for a space universe, taking each grid pixel as a circle center, and taking h as a radius to establish a neighborhood calculation density estimation value, wherein the density weight value of the elements is larger as the elements are closer to the circle center, and is reduced in a quadratic curve along with the increase of the distance, and the density distribution of the complete nuclear density of the whole area is calculated and obtained at the distance h.
As shown in fig. 3, for example, the kernel density estimation of line elements, in abstract terms, each line is covered with a smooth surface. The value of which is greatest at the location of the line and gradually decreases as the distance from the line increases, and is zero at the location of the distance from the line equal to the threshold search radius. The volume of the space surrounded by the curved surface and the plane below is equal to the product of the line length and the weighted value (the value is 1 if no weighting exists), and the density value of each output grid pixel is the sum of all values overlapped at the center of the grid pixel.
The density estimation method based on the grid pixels has great advantages compared with other density estimation methods. Since the grid side length of the grid pixels is pixel-level, the density distribution of the entire grid area can be regarded as "smooth" and continuous to some extent; the kernel density estimation is based on the density value calculation of the data, interference of human priori knowledge is avoided, and more essential distribution characteristics and content information of the data are displayed.
In the contrast linear density analysis and nuclear density estimation methods, although density value calculation is carried out on grid pixels, the linear density analysis does not consider the distance between the linear density analysis and the circle center when density counting is carried out on elements in a threshold radius search circle, and the density calculation weights are 1; in the kernel density estimation algorithm, the density count value of the element is the highest value 1 at the center of the circle and is the highest value 0 at the distance from the center of the circle, and the value and the center distance are reduced in a quadratic curve, so that the density distribution of the original data is restored more accurately and truly.
As shown in fig. 4, sub-graph (a) is the original graph; sub-panel (b) is a line density analysis plot; subgraph (c) is a kernel density estimation graph of line elements; the same piece of data is processed by two methods, the same searching bandwidth 100 is adopted for density analysis, and the comparison result shows that the widths of raster images of different areas in the linear density analysis result are equal, and the difference of the area densities is only represented by the pixel gray values, so that the low-density area is seriously widened; the nuclear density estimation result is more in line with the density distribution of the original data, and the width of the analysis result is in direct proportion to the original data.
As shown in fig. 5, sub-image (a) is a raster image; subgraph (b) is a skeleton line extraction graph; subgraph (c) is a contour extraction graph; grid vectorization technology is an important branch in computer image processing and is also a key technology in the science of geographic information. Vector representation in the geography information discipline mainly comprises three types of space elements of points, lines and planes, and raster data is array data with different gray values or color values which are presented in a standard grid form. Vectorization of a raster image can be divided into a refinement method and a non-refinement method from the result, wherein the refinement method is a process of searching for any thick linear object with a width larger than one pixel on the raster image, deleting contour pixels of the thick linear object, reserving skeleton pixels of the thick linear object, and recording positions of the rough linear object as break points to output vector lines, and is also called a skeletonization method; the contour line-based method is to carry out the sketching identification of the pixel edge on the raster image and generate the contour surface or contour boundary of the image.
As shown in fig. 6, delaunay triangulation, abbreviated as triangulation, is widely used in geographic information field research and is a powerful tool for spatial clustering and proximity relation analysis. Conceptually, a plurality of discrete points are given in space, and the circumscribed circle of each Delaunay triangle taking the discrete points as vertexes does not contain any other points, so that the network constructed by the triangles is called a triangle network. The Delaunay triangle network has the advantages of good structure, simple data structure, small data redundancy, adaptability to various distributed density data and the like; its limitation is that the algorithm is complex and difficult to implement.
As shown in fig. 7, the Voronoi diagram, also known as a tessellated polygon, is a subdivision of the space plane, creating a V-diagram for discrete points in space, each polygon having and containing only one discrete point, characterized by any location within the polygon being closest to the point of the polygon and distant from the point within an adjacent polygon. Due to the aliquoting characteristics of the Thiessen polygons on the spatial subdivision, the method can be used for solving the problems of the nearest point, the minimum closed circle and the like, and a plurality of spatial analysis problems such as adjacency, proximity and accessibility analysis and the like. The method is characterized in that the method is an even graph with a triangular net, which is a concomitant relationship, a V graph can be constructed by connecting the circle centers of circumscribed circles of triangles adjacent to discrete points in the triangular net, each V graph polygon is internally provided with only one point, and the connection line of the discrete points in the adjacent polygons in the V graph is perpendicular to the common edge of the polygons, namely the edge of the triangular net.
As shown in fig. 8, the characteristics of the triangle mesh and the V-diagram determine its strong role in analyzing discrete data, and also determine its role in geoscience research, which is often used when researching the distribution of discrete elements in two-dimensional space. The triangular side length of the triangular net and the area of the V diagram are closely related to the distribution density of the discrete points, and the triangular net and the V diagram can be built for density analysis by utilizing the characteristic, and the V diagram density analysis is also a common analysis method. Taking discrete element points in fig. 8 as a V-chart, calculating the polygon area si of each element, wherein the smaller the polygon area of the V-chart in the area with denser points is, the larger the density is reflected as the gray value in the chart, namely the density di is inversely proportional to the area:
the foregoing examples merely illustrate specific embodiments of the invention, which are described in greater detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention.
Claims (5)
1. The road line extraction method based on the Voronoi diagram and the kernel density is characterized by comprising the following steps of:
step one: preprocessing data;
the data preprocessing comprises the following specific steps:
removing the track points of the positioning information and the attribute information recording errors through logic judgment of the attribute information;
visualizing the track points to obtain a track line result, namely obtaining a track line element;
identifying obvious vectors and topologies of the track line elements in the electronic map and errors which do not accord with the actual vehicle running track elements, and performing further data cleaning;
the track points for removing the positioning information and the recording errors of the attribute information through the logic judgment of the attribute information comprise the following track points:
A. points where the attribute and positioning information are incomplete or empty;
B. locating points with discrete points with overlarge distances;
C. the distance between track points in adjacent time is overlarge and exceeds the theoretical speed per hour;
D. discrete points with overlarge distances and overlarge time intervals in a track point set of the same vehicle;
E. repetition time and repeatedly positioned repetition points within the same vehicle track;
step two: establishing a Delaunay triangle network, and identifying non-road edges by adopting a deformation kernel density algorithm;
the deformation kernel density algorithm formula is:
k is a kernel density function, h is a smoothing parameter, and h >0;
the Delaunay triangle network is established, and the specific steps of marking the non-road edge are as follows:
establishing a Delaunay triangle network by combining all track points, and connecting circle centers of circumscribed circles of triangles adjacent to discrete points in the Delaunay triangle network to construct a Voronoi diagram according to the associated relationship between the Voronoi diagram and the Delaunay triangle network and the two figures;
calculating deformation kernel density estimation values of all vertexes, namely deformation kernel density estimation values of the track points, in a Delaunay triangle network by using the density of the Voronoi diagram;
taking the ratio of the deformation nuclear density values of the two side points of each side in the Delaunay triangular net as the density change rate, judging the size of the value of the density change rate and the set threshold value, and marking the sides with the values larger than the threshold value, namely the left side and the right side with large density change as non-road sides;
step three: cleaning the external points of the road;
the specific steps of cleaning the external points of the road are as follows:
judging the number of non-road sides marked in adjacent sides of the track points, and identifying the number as an external point of the road if the number is greater than a set constraint threshold;
after all the external points of the road are cleaned, an external outline polygon formed by the internal points of the road is left, namely the outline of the road surface;
step four: secondarily constructing a Delaunay triangle network, and extracting a road surface;
step five: further processing the obtained road surface profile and optimizing the road boundary;
step six: and extracting skeleton lines, namely extracting central lines by using a skeletonizing method.
2. The road line extraction method based on the Voronoi diagram and the kernel density according to claim 1, wherein in the fourth step, a triangle network is constructed secondarily, and the specific steps of extracting the road surface are as follows:
secondarily constructing a Delaunay triangle network by using the internal points of the road, and removing the long-side triangle by using an integral long-side constraint criterion to obtain a rough road contour consisting of a plurality of short-side triangles;
and merging the left short-side triangles into a polygon, namely the road surface.
3. The road line extraction method based on Voronoi diagram and kernel density according to claim 1, wherein in step five, the obtained road surface profile is further processed, and the specific steps of optimizing the road boundary are as follows:
namely, removing some tiny polygons which are independent of the outside of the road surface according to the size of the area;
and (3) smoothing the extracted boundary by using a vector processing tool to reduce small burrs, and obtaining a final road boundary result after optimization.
4. The road line extraction method based on the Voronoi diagram and the kernel density according to claim 1, wherein in the sixth step, the skeleton line is extracted by using a skeletonizing method, namely, the specific steps of extracting the center line are as follows:
and generating a binarization grid image by the extracted road polygon, and extracting a skeleton line, namely a road center line, by using a skeletonizing method.
5. The method for extracting road lines based on Voronoi diagram and kernel density according to claim 4, wherein in the sixth step, the skeletonizing method is a process of searching for any thick line-like object having a width larger than one pixel on the raster image, deleting its contour pixels, retaining its skeleton pixels, and recording its position as a break point output vector line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310793429.XA CN116543310B (en) | 2023-06-30 | 2023-06-30 | Road line extraction method based on Voronoi diagram and kernel density |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310793429.XA CN116543310B (en) | 2023-06-30 | 2023-06-30 | Road line extraction method based on Voronoi diagram and kernel density |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116543310A CN116543310A (en) | 2023-08-04 |
| CN116543310B true CN116543310B (en) | 2023-10-31 |
Family
ID=87452796
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310793429.XA Active CN116543310B (en) | 2023-06-30 | 2023-06-30 | Road line extraction method based on Voronoi diagram and kernel density |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116543310B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116977480B (en) * | 2023-09-21 | 2023-12-12 | 湖北大学 | Automatic segmentation method and system for scale-related heterogeneity line elements |
| CN117191004B (en) * | 2023-11-06 | 2024-03-19 | 中南大学 | Outdoor three-dimensional walking navigation road network map generation method integrating crowd-sourced track data |
Citations (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2002029610A2 (en) * | 2000-10-05 | 2002-04-11 | Digitalmc Corporation | Method and system to classify music |
| KR20060084591A (en) * | 2005-01-20 | 2006-07-25 | (주)제이투엠소프트 | Method for generating three-dimensional graphic data of road information and apparatus thereof |
| CN103186787A (en) * | 2011-12-31 | 2013-07-03 | 廖志武 | Low-quality Chinese character primary skeleton extraction algorithm based on point cloud model |
| CN103336783A (en) * | 2012-05-11 | 2013-10-02 | 南京大学 | Voronoi and inverse distance weighting combined density map drawing method |
| WO2014129116A1 (en) * | 2013-02-22 | 2014-08-28 | 国立大学法人東京工業大学 | Information processing device, information processing method, and non-transitory computer-readable medium |
| CN104089619A (en) * | 2014-05-14 | 2014-10-08 | 北京联合大学 | GPS navigation map accurate matching system of pilotless automobile, and its operation method |
| CN107247938A (en) * | 2017-06-08 | 2017-10-13 | 中国科学院遥感与数字地球研究所 | A kind of method of high-resolution remote sensing image City Building function classification |
| CN109887024A (en) * | 2019-02-16 | 2019-06-14 | 西南科技大学 | A kind of cloud normal estimates new method |
| CN110188664A (en) * | 2019-05-28 | 2019-08-30 | 福州大学 | A kind of fine extracting method of vehicle-mounted laser point cloud vector road boundary based on Snake |
| CN110322694A (en) * | 2019-07-16 | 2019-10-11 | 青岛海信网络科技股份有限公司 | A kind of method and device of urban traffic control piece Division |
| CN111080501A (en) * | 2019-12-06 | 2020-04-28 | 中国科学院大学 | Real crowd density space-time distribution estimation method based on mobile phone signaling data |
| CN111613045A (en) * | 2019-02-25 | 2020-09-01 | 阿里巴巴集团控股有限公司 | Method and device for verifying road traffic condition |
| CN112148812A (en) * | 2019-06-26 | 2020-12-29 | 丰图科技(深圳)有限公司 | Method, device and equipment for extracting road center line and storage medium thereof |
| CN112364890A (en) * | 2020-10-20 | 2021-02-12 | 武汉大学 | Intersection guiding method for making urban navigable network by taxi track |
| CN112417233A (en) * | 2020-12-16 | 2021-02-26 | 湖北大学 | Thermodynamic diagram generation method and system considering space density difference |
| CN113190538A (en) * | 2021-03-31 | 2021-07-30 | 北京中交兴路信息科技有限公司 | Road construction method and device based on track data, storage medium and terminal |
| WO2021189363A1 (en) * | 2020-03-26 | 2021-09-30 | 深圳先进技术研究院 | Method and apparatus for determining service area of parking lot, device, and storage medium |
| CN113466481A (en) * | 2020-03-30 | 2021-10-01 | 深圳迈瑞生物医疗电子股份有限公司 | Filter state detection system and detection method |
| CN113505187A (en) * | 2021-07-07 | 2021-10-15 | 西安理工大学 | Vehicle classification track error correction method based on map matching |
| WO2021243516A1 (en) * | 2020-06-01 | 2021-12-09 | 深圳先进技术研究院 | Urban public transport passenger travel trajectory estimation method and system, terminal, and storage medium |
| CN113902830A (en) * | 2021-12-08 | 2022-01-07 | 腾讯科技（深圳）有限公司 | Method for generating track road network |
| WO2022063005A1 (en) * | 2020-09-22 | 2022-03-31 | 北京智行者科技有限公司 | Global path planning method and apparatus |
| WO2022089194A1 (en) * | 2020-10-29 | 2022-05-05 | 同济大学 | Millimeter-wave radar data-based lane line detection method |
| CN114898559A (en) * | 2022-05-26 | 2022-08-12 | 中国科学院深圳先进技术研究院 | Method for measuring moving perception capability of urban vehicle |
| CN115326055A (en) * | 2022-08-24 | 2022-11-11 | 长沙迪迈数码科技股份有限公司 | Method and device for rasterizing track line of surface mine road |
| CN115752459A (en) * | 2021-09-02 | 2023-03-07 | 中国人民解放军战略支援部队信息工程大学 | Trajectory rectification method based on indoor position network model |
| CN115878735A (en) * | 2021-09-28 | 2023-03-31 | 北京三快在线科技有限公司 | Road network generation method, road network generation device, electronic equipment and storage medium |
| CN116010885A (en) * | 2022-12-21 | 2023-04-25 | 重庆邮电大学 | Method and system for detecting abnormal space-time data of vehicle under long-sequence condition |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6364922B2 (en) * | 2014-04-23 | 2018-08-01 | 富士通株式会社 | Integration destination route extraction method, route graph creation method, device, and program |
| EP3279868A1 (en) * | 2016-08-01 | 2018-02-07 | Ecole Nationale de l'Aviation Civile | Random path generation upon functional decomposition |
| US11062108B2 (en) * | 2017-11-07 | 2021-07-13 | Digimarc Corporation | Generating and reading optical codes with variable density to adapt for visual quality and reliability |
-
2023
- 2023-06-30 CN CN202310793429.XA patent/CN116543310B/en active Active
Patent Citations (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2002029610A2 (en) * | 2000-10-05 | 2002-04-11 | Digitalmc Corporation | Method and system to classify music |
| KR20060084591A (en) * | 2005-01-20 | 2006-07-25 | (주)제이투엠소프트 | Method for generating three-dimensional graphic data of road information and apparatus thereof |
| CN103186787A (en) * | 2011-12-31 | 2013-07-03 | 廖志武 | Low-quality Chinese character primary skeleton extraction algorithm based on point cloud model |
| CN103336783A (en) * | 2012-05-11 | 2013-10-02 | 南京大学 | Voronoi and inverse distance weighting combined density map drawing method |
| WO2014129116A1 (en) * | 2013-02-22 | 2014-08-28 | 国立大学法人東京工業大学 | Information processing device, information processing method, and non-transitory computer-readable medium |
| CN104089619A (en) * | 2014-05-14 | 2014-10-08 | 北京联合大学 | GPS navigation map accurate matching system of pilotless automobile, and its operation method |
| CN107247938A (en) * | 2017-06-08 | 2017-10-13 | 中国科学院遥感与数字地球研究所 | A kind of method of high-resolution remote sensing image City Building function classification |
| CN109887024A (en) * | 2019-02-16 | 2019-06-14 | 西南科技大学 | A kind of cloud normal estimates new method |
| CN111613045A (en) * | 2019-02-25 | 2020-09-01 | 阿里巴巴集团控股有限公司 | Method and device for verifying road traffic condition |
| CN110188664A (en) * | 2019-05-28 | 2019-08-30 | 福州大学 | A kind of fine extracting method of vehicle-mounted laser point cloud vector road boundary based on Snake |
| CN112148812A (en) * | 2019-06-26 | 2020-12-29 | 丰图科技(深圳)有限公司 | Method, device and equipment for extracting road center line and storage medium thereof |
| CN110322694A (en) * | 2019-07-16 | 2019-10-11 | 青岛海信网络科技股份有限公司 | A kind of method and device of urban traffic control piece Division |
| CN111080501A (en) * | 2019-12-06 | 2020-04-28 | 中国科学院大学 | Real crowd density space-time distribution estimation method based on mobile phone signaling data |
| WO2021189363A1 (en) * | 2020-03-26 | 2021-09-30 | 深圳先进技术研究院 | Method and apparatus for determining service area of parking lot, device, and storage medium |
| CN113466481A (en) * | 2020-03-30 | 2021-10-01 | 深圳迈瑞生物医疗电子股份有限公司 | Filter state detection system and detection method |
| WO2021243516A1 (en) * | 2020-06-01 | 2021-12-09 | 深圳先进技术研究院 | Urban public transport passenger travel trajectory estimation method and system, terminal, and storage medium |
| WO2022063005A1 (en) * | 2020-09-22 | 2022-03-31 | 北京智行者科技有限公司 | Global path planning method and apparatus |
| CN112364890A (en) * | 2020-10-20 | 2021-02-12 | 武汉大学 | Intersection guiding method for making urban navigable network by taxi track |
| WO2022089194A1 (en) * | 2020-10-29 | 2022-05-05 | 同济大学 | Millimeter-wave radar data-based lane line detection method |
| CN112417233A (en) * | 2020-12-16 | 2021-02-26 | 湖北大学 | Thermodynamic diagram generation method and system considering space density difference |
| CN113190538A (en) * | 2021-03-31 | 2021-07-30 | 北京中交兴路信息科技有限公司 | Road construction method and device based on track data, storage medium and terminal |
| CN113505187A (en) * | 2021-07-07 | 2021-10-15 | 西安理工大学 | Vehicle classification track error correction method based on map matching |
| CN115752459A (en) * | 2021-09-02 | 2023-03-07 | 中国人民解放军战略支援部队信息工程大学 | Trajectory rectification method based on indoor position network model |
| CN115878735A (en) * | 2021-09-28 | 2023-03-31 | 北京三快在线科技有限公司 | Road network generation method, road network generation device, electronic equipment and storage medium |
| CN113902830A (en) * | 2021-12-08 | 2022-01-07 | 腾讯科技（深圳）有限公司 | Method for generating track road network |
| CN114898559A (en) * | 2022-05-26 | 2022-08-12 | 中国科学院深圳先进技术研究院 | Method for measuring moving perception capability of urban vehicle |
| CN115326055A (en) * | 2022-08-24 | 2022-11-11 | 长沙迪迈数码科技股份有限公司 | Method and device for rasterizing track line of surface mine road |
| CN116010885A (en) * | 2022-12-21 | 2023-04-25 | 重庆邮电大学 | Method and system for detecting abnormal space-time data of vehicle under long-sequence condition |
Non-Patent Citations (4)
| Title |
|---|
| "基于众源轨迹数据的道路中心线提取";杨伟等;《地理与地理信息科学》;第32卷(第3期);第2页第2.1节 * |
| "基于圈层结构的游客活动空间边界提取新方法";吴朝宁等;《地理学报》;第76卷(第6期);第1540页第2.4节 * |
| "基于步行轨迹的复杂道路中心线提取方法";李俊杰等;《华中师范大学学报（自然科学版）》;第54卷(第1期);第91页第1.3节 * |
| "运用约束Delaunay三角网从众源轨迹线提取路道边界";杨伟等;"运用约束Delaunay三角网从众源轨迹线提取路道边界";第46卷(第2期);第238-240页第2节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116543310A (en) | 2023-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116543310B (en) | Road line extraction method based on Voronoi diagram and kernel density | |
| CN112070779B (en) | Remote sensing image road segmentation method based on convolutional neural network weak supervised learning | |
| CN108519094B (en) | Local path planning method and cloud processing terminal | |
| CN106780524B (en) | Automatic extraction method for three-dimensional point cloud road boundary | |
| CN101551905B (en) | Method for segmenting multi-dimensional texture image on basis of fuzzy C-means clustering and spatial information | |
| CN113065594B (en) | Road network extraction method and device based on Beidou data and remote sensing image fusion | |
| Falcidieno et al. | A new method for the characterization of topographic surfaces | |
| Sohn et al. | An implicit regularization for 3D building rooftop modeling using airborne lidar data | |
| CN113724279B (en) | System, method, equipment and storage medium for automatically dividing traffic cells into road networks | |
| CN111652241B (en) | Building contour extraction method integrating image features and densely matched point cloud features | |
| CN115222625A (en) | Laser radar point cloud denoising method based on multi-scale noise | |
| CN114332291A (en) | Oblique photography model building outer contour rule extraction method | |
| CN111858810A (en) | Modeling elevation point screening method for road DEM construction | |
| CN112348867A (en) | Method and system for constructing city high-precision three-dimensional terrain based on LiDAR point cloud data | |
| CN110580388A (en) | Channel network extraction method based on crowd-sourced trajectory data | |
| Ortega et al. | Generating 3D city models from open LiDAR point clouds: Advancing towards smart city applications | |
| CN112462347A (en) | Laser radar point cloud rapid classification filtering algorithm based on density clustering | |
| Wang et al. | Adaptive extraction and refinement of marine lanes from crowdsourced trajectory data | |
| Deng et al. | A novel fast classification filtering algorithm for LiDAR point clouds based on small grid density clustering | |
| CN115620263B (en) | Intelligent vehicle obstacle detection method based on image fusion of camera and laser radar | |
| Li et al. | An area merging method in map generalization considering typical characteristics of structured geographic objects | |
| US20230384117A1 (en) | Method, data processing apparatus and computer program product for generating map data | |
| CN114283343A (en) | Map updating method, training method and equipment based on remote sensing satellite image | |
| CN114037948A (en) | Vehicle-mounted road point cloud element vectorization method and device based on migration active learning | |
| Widyaningrum | Automatic object extraction from airborne laser scanning point clouds for digital base map production |
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 | ||
| CP03 | Change of name, title or address |
Address after: 620564 2F, Building B, COFCO California Smart City, Shigao Street, Renshou County, Meishan City, Sichuan Province Patentee after: Huantian Smart Technology Co.,Ltd. Country or region after: China Address before: 620564 2F, Building B, COFCO California Smart City, Shigao Street, Renshou County, Meishan City, Sichuan Province Patentee before: Meishan HUANTIAN Intelligent Technology Co.,Ltd. Country or region before: China |
|
| CP03 | Change of name, title or address |