CN109583626B - Road network topology reconstruction method, medium and system - Google Patents
Road network topology reconstruction method, medium and system Download PDFInfo
- Publication number
- CN109583626B CN109583626B CN201811275175.8A CN201811275175A CN109583626B CN 109583626 B CN109583626 B CN 109583626B CN 201811275175 A CN201811275175 A CN 201811275175A CN 109583626 B CN109583626 B CN 109583626B
- Authority
- CN
- China
- Prior art keywords
- road network
- road
- similarity
- network topology
- source image
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
Abstract
The invention discloses a road network topology reconstruction method, medium and system, comprising the following steps: acquiring original road network data to generate a training data set; constructing a generator, wherein the generator generates first data information according to the original road network data; training according to the road reference image, the first road network source image and the first data information to construct a first discriminator; training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator; constructing a road network topology reconstruction to generate a confrontation network; acquiring a second road network source image, generating a second initial road network map, and inputting the second road network source image and the second initial road network map into a network to carry out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
Description
Technical Field
The invention relates to the technical field of remote sensing, in particular to a road network topology reconstruction method, medium and system.
Background
The remote sensing technology has the characteristics of high efficiency, real-time performance, information diversification and the like, and plays an important role in the field of urban traffic as an advanced earth observation method. Road network extraction has received much attention in the past decades as an important problem in remote sensing applications. The high-quality road network extraction result not only depends on the identification of road regions, but also depends on road network topology reconstruction.
Existing road extraction methods usually include a step of topology analysis to obtain a complete road network. These methods rely on specific features or a large number of samples of various road regions, and are effective for road extraction in urban areas. However, road conditions in rural areas are more complicated than in urban areas, spectral characteristics in remote sensing images are various, and vehicle trajectories that can be acquired in some sections of rural areas are often scarce. Therefore, the conventional method is difficult to be practically applied to road network topology reconstruction in rural areas.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the art described above. Therefore, an object of the present invention is to provide a road network topology reconstruction method, which can reconstruct the road network topology in rural areas, and the accuracy of the result is higher, and meanwhile, the original road network data is matched by a road network matching algorithm, so that the training data is easier to obtain, and the construction efficiency of the road network topology reconstruction for generating the confrontation network is improved.
A second object of the invention is to propose a computer-readable storage medium.
The third purpose of the invention is to provide a road network topology reconstruction system.
In order to achieve the above object, an embodiment of a first aspect of the present invention provides a road network topology rebuilding method, including the following steps: acquiring original road network data, and generating a training data set according to the original road network data based on a road network matching algorithm, wherein the original road network data comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data; training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; training according to the training data set to construct a road network topology reconstruction and generate a confrontation network; acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, and inputting the second road network source image and the second initial road network map into the road network topology reconstruction to generate a confrontation network for road network topology reconstruction.
According to the road network topology reconstruction method, firstly, original road network data are obtained, and based on a road network matching algorithm, a training data set is generated according to the original road network data, wherein the original road network data comprise a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data; then, training is carried out according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; secondly, training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; then, training is carried out according to the training data set to construct a road network topology reconstruction and generate a confrontation network; then, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
In addition, the road network topology reconstruction method proposed by the above embodiment of the present invention may further have the following additional technical features:
optionally, the generating a training data set according to the original road network data based on the road network matching algorithm specifically includes: preprocessing the original road network data to generate a vector road network map consisting of vector lines; selecting at least one vector line as a reference line, and selecting candidate line segments based on a buffer growth algorithm; and adopting an iterative expansion line algorithm to expand and judge the similarity of the selected candidate line segment and the reference line so as to generate a final matching pair.
Optionally, the selecting at least one vector line as a reference line and selecting candidate line segments based on a buffer growth algorithm specifically includes: selecting at least one vector line as a reference line, and arranging a buffer area around the reference line to obtain the vector line in the buffer area as a preselected object; calculating a similarity S between the reference line and the preselected object1And judging the similarity S1Whether the similarity is greater than a preset first similarity threshold value or not, and comparing the similarity S1And the preselected object corresponding to the similarity threshold value larger than the preset first similarity threshold value is used as a candidate line segment.
Optionally, the similarity S1Obtained by the following formula:
wherein, SimlenIs the Euclidean distance, Sim, of the two end points of the vector linedisIs a short-edge-based median Hausdorff distance, SimoriIs the angle between the straight line and the horizontal axis, SimshapeAs a similar term of shape, wlen，wdis，wori，wshapeAre each Simlen，Simdis，Simori，SimshapeThe weight of (c).
Optionally, the expanding and similarity judging the selected candidate line segment and the reference line by using an iterative expanded line algorithm specifically includes: respectively expanding each candidate line segment and the reference line to generate corresponding candidate broken lines and reference broken lines, and calculating the similarity S between the candidate broken lines and the reference broken lines2(ii) a Judging the similarity S2Whether the similarity is less than a preset second similarity threshold value or not, and determining the similarity S2And the candidate broken line and the reference broken line which correspond to the second similarity threshold value are expanded again to carry out the next iterative computation and similarity judgment until the similarity S is reached2And the candidate broken lines corresponding to the second similarity threshold value which is greater than or equal to the preset threshold value reach a preset number threshold value.
Optionally, the similarity S2Obtained by the following formula:
wherein, SimplenIs the sum of Euclidean distances of all vector lines contained in the candidate folding line or the reference folding line, SimpdisA median Hausdorff distance, Sim, based on the short edge for the candidate polyline or the reference polylinepshapeAs a similar term of shape, wplen，wpdis，wpshapeAre each Simplen，Simpdis，SimpshapeThe weight of (c).
Optionally, the generator is composed of an encoder and a decoder, wherein the encoder includes 3 convolutional layers, 4 residual blocks and 2 deconvolution layers, and the decoder adopts a symmetrical structure with the encoder so that the input and output of the generator have the same resolution.
Optionally, the training is performed according to the training data set to construct a road network topology reconstruction and generate a confrontation network, specifically: and performing combined training on the generator, the first discriminator, the second discriminator and a VGG network to construct the road network topology reconstruction and generate the confrontation network.
To achieve the above object, a second aspect of the present invention provides a computer-readable storage medium, on which a road network topology reconstruction program is stored, which, when executed, implements the road network topology reconstruction method as described above.
In order to achieve the above object, an embodiment of a third aspect of the present invention provides a road network topology rebuilding system, including: the system comprises an acquisition module, a comparison module and a comparison module, wherein the acquisition module is used for acquiring original road network data and generating a training data set according to the original road network data based on a road network matching algorithm, wherein the original road network data comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; a first construction module, configured to train the original road network data to construct a generator, where the generator generates first data information according to the original road network data; a second construction module, configured to train according to the road reference image, the first road network source image, and the first data information to construct a first discriminator, where the first discriminator is used to assist in generating a road network map; a third construction module, configured to train according to the road reference image, the first initial road network map, and the first data information to construct a second discriminator, where the second discriminator is configured to assist in reconstructing a road network topology; the generation module is used for training according to the training data set to construct road network topology reconstruction and generate a confrontation network; and the reconstruction module is used for acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, and inputting the second road network source image and the second initial road network map into the road network topology reconstruction to generate a confrontation network so as to carry out road network topology reconstruction.
The road network topology reconstruction system according to the embodiment of the invention comprises: the system comprises an acquisition module, a training module and a training module, wherein the acquisition module is used for acquiring original road network data and generating a training data set according to the original road network data based on a road network matching algorithm, and the original road network data comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; the system comprises a first construction module, a second construction module and a third construction module, wherein the first construction module is used for training original road network data to construct a generator, and the generator generates first data information according to the original road network data; the second construction module is used for training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; the third construction module is used for training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; the generation module is used for training according to the training data set to construct road network topology reconstruction and generate a confrontation network; the reconstruction module is used for acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
Drawings
Fig. 1 is a schematic flow chart of a road network topology reconstruction method according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a countermeasure network generated by reconstructing a road network topology according to an embodiment of the present invention;
FIG. 3 is a comparison graph of the effect of the method for reconstructing road network topology according to the embodiment of the present invention on the road network topology reconstruction compared with the method for generating an antagonistic network under the conventional conditions;
FIG. 4 is a comparison graph of detection effects of the road network topology reconstruction method and other related road extraction algorithms according to the embodiment of the present invention;
fig. 5 is a schematic flow chart of a road network topology reconstruction method according to another embodiment of the present invention;
fig. 6 is a block diagram of a road network topology reconstruction system according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
According to the road network topology reconstruction method provided by the embodiment of the invention, firstly, original road network data is obtained, and a training data set is generated according to the original road network data based on a road network matching algorithm, wherein the original road network data comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data; then, training is carried out according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; secondly, training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; then, training is carried out according to the training data set to construct a road network topology reconstruction and generate a confrontation network; then, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
In order to better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
In order to better understand the technical solution, the technical solution will be described in detail with reference to the drawings and the specific embodiments.
Fig. 1 is a schematic flow chart of a road network topology reconstruction method according to an embodiment of the present invention, as shown in fig. 1, the road network topology reconstruction method includes the following steps:
s101, original road network data are obtained, and a training data set is generated according to the original road network data based on a road network matching algorithm, wherein the original road network data comprise a first road network source image, a road reference image and a first original road network map corresponding to the first road network source image.
Based on the road network matching algorithm, there are various ways of generating the training data set according to the original road network data, for example: and generating a training data set according to the original road network data by using a semi-deterministic algorithm, generating a training data set according to the original road network data by using a probability statistical algorithm and the like.
As an example, based on the road network matching algorithm, the generating of the training data set from the original road network data may specifically include: a data preprocessing stage, a candidate line segment selection stage and a matching pair generation stage; specifically, firstly, preprocessing original road network data through a data preprocessing stage to generate a road network map composed of a group of vector lines, wherein each vector is represented by two end points, and each end point in the road network is supposed to be at a corner position; then, selecting candidate line segments through a buffer growth algorithm to obtain the candidate line segments with similarity greater than a threshold value with the selected reference line; then, generating a final matching pair by adopting an iterative extended line algorithm; therefore, through a simple matching algorithm, the existing road reference image and the first initial road network map can be matched to form a training data set, so that the generation of the training data set is easier, and the training rate of generating the countermeasure network through road network topology reconstruction is increased.
S102, training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data.
As an example, in building the generator, training of the generator is performed by taking the road network source image and the first initial road network map as input data.
As another example, the generator includes an encoder including 3 convolutional layers, 4 residual layers, and 2 deconvolution layers, and a decoder adopting a structure symmetrical to the encoder such that the input and output of the generator have the same resolution. It should be noted that 2-hop connections can be used between the encoder and the decoder to retain low-level features when expanding the resolution of the feature map.
S103, training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map.
That is, in the process of constructing the first discriminator, the first discriminator is trained by taking the road reference image, the first road network source image and the first data information as input, wherein the first data information is the output of the generator, and the first discriminator is used for assisting in generating the road network map.
As an example, the first discriminator comprises 4 two-dimensional convolutional neural network layers operating on N × N image blocks by a markov random field, as shown in fig. 2, wherein the N × N image blocks may be 70 pixels × 70 pixels.
And S104, training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology.
That is, in the construction process of the second discriminator, the second discriminator is trained by taking as input the road reference image, the first initial road network map and the first data information, wherein the first data information is the output of the generator, and the second discriminator is used for assisting in reconstructing the road network topology.
As an example, the second discriminator comprises 4 two-dimensional convolutional neural network layers operating on N × N image blocks by a markov random field, as shown in fig. 2, wherein the N × N image blocks may be 70 pixels × 70 pixels.
And S105, training according to the training data set to construct a road network topology reconstruction and generate a confrontation network.
There are various ways of generating the countermeasure network by training according to the training data set to construct the road network topology reconstruction.
As an example, as shown in fig. 2, the generator, the first discriminator, the second discriminator and one VGG network are jointly trained to construct a road network topology reconstruction and generate a countermeasure network.
S106, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction.
It should be noted that, this step is a test process; and the step of extracting the road from the second road network source image refers to generating a confrontation network or other road extraction methods by using the traditional conditions so as to generate a second initial road network map corresponding to the second road network source image.
In summary, according to the road network topology reconstruction method provided by the embodiment of the present invention, first, original road network data is obtained, and based on a road network matching algorithm, a training data set is generated according to the original road network data, wherein the original road network data includes a first road network source image, a road reference image, and a first initial road network map corresponding to the first road network source image; training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data; then, training is carried out according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; secondly, training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; then, training is carried out according to the training data set to construct a road network topology reconstruction and generate a confrontation network; then, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
FIG. 3 is a comparison of the effect of 3 sets of the method for reconstructing road network topology disclosed by the present invention on road network topology reconstruction with the method for generating confrontation network under conventional conditions, as shown in FIG. 4, wherein (a) column is road network source image, (b) column is effect image for generating confrontation network under conventional conditions, and (c) column is road network topology reconstruction method without L in the present inventionVGG(G) Effect graph at loss of term, (d) column is listed here for network without Lt(G, D) the effect graph when the item is lost, (e) the column is the effect graph adopting the text network, and (f) the column is the real reference road network graph. The result shows that the multi-condition generation countermeasure network provided by the invention has better performance in road network topology reconstruction than the traditional condition generation countermeasure network, and the design of the network structure is reasonable.
FIG. 4 is a comparison graph of the detection effects of 4 sets of road network topology reconstruction methods disclosed in the present invention and other related road extraction algorithms; the method comprises the following steps of (a) listing as a source image, (b) listing, (c) listing, (d) listing, (e) listing, (f) listing as effect graphs corresponding to other relevant continental extraction algorithms respectively, (g) listing as an effect graph adopting the method, and (h) listing as a real reference road network graph. The corresponding quantitative results are shown in the following table:
| method of producing a composite material | Recall rate | Accuracy of measurement | F1 score |
| b | 0.322 | 0.405 | 0.359 |
| C | 0.679 | 0.471 | 0.556 |
| d | 0.686 | 0.435 | 0.532 |
| e | 0.729 | 0.606 | 0.662 |
| f | 0.783 | 0.812 | 0.797 |
| g | 0.858 | 0.841 | 0.849 |
It can be seen from the table that, compared with the traditional road network extraction method, the method provided by the invention exceeds the existing method in recall rate, precision and F1 score, and in conclusion, the network performance provided by the invention is superior to the existing road network extraction method.
In some embodiments, the road network topology reconstruction method provided in the embodiments of the present invention performs generation of a training data set through a simple road network matching algorithm, so that training for generating a countermeasure network through road network topology reconstruction becomes simple and easy, as shown in fig. 5, the road network topology reconstruction method includes the following steps:
s201, original road network data is obtained.
S202, the original road network data are preprocessed to generate a vector road network map composed of vector lines.
S203, selecting at least one vector line as a reference line, and setting a buffer area around the reference line to obtain the vector line in the buffer area as a preselected object.
S204, calculating the similarity S between the reference line and the preselected object1And judging the similarity S1Whether the similarity is larger than a preset first similarity threshold value or not, and determining the similarity S1And the preselected object corresponding to the similarity threshold value larger than the preset first similarity threshold value is used as a candidate line segment.
Wherein the similarity S between the reference line and the preselected object is calculated1There are various methods.
As an example, similarity S1Obtained by the following formula:
wherein, SimlenIs the Euclidean distance, Sim, of the two end points of the vector linedisIs a short-edge-based median Hausdorff distance, SimoriIs the angle between the straight line and the horizontal axis, SimshapeAs a similar term of shape, wlen，wdis，wori，wshapeAre each Simlen，Simdis，Simori，SimshapeThe weight of (c).
S205, respectively expanding each candidate line segment and the reference line to generate corresponding candidate broken lines and reference broken lines, and calculating the similarity S between the candidate broken lines and the reference broken lines2。
Wherein, similarity S between the candidate broken line and the reference broken line is calculated2There are various methods.
As an example, similarity S2Obtained by the following formula:
wherein, SimplenIs the sum of Euclidean distances of all vector lines contained in the candidate folding line or the reference folding line, SimpdisA median Hausdorff distance, Sim, based on the short edge for the candidate polyline or the reference polylinepshapeAs a similar term of shape, wplen，wpdis，wpshapeAre each Simplen，Simpdis，SimpshapeThe weight of (c).
S206, judging similarity S2Whether the similarity is less than a preset second similarity threshold value or not, and determining the similarity S2The candidate broken line and the reference broken line which correspond to the second similarity threshold value are expanded again to carry out the next iterative computation and similarity judgment until the similarity S is reached2And the candidate broken lines corresponding to the second similarity threshold value which is greater than or equal to the preset threshold value reach a preset number threshold value.
That is, the similarity S between the candidate polyline and the reference polyline is calculated2Thereafter, the similarity S is retained2Continuing to expand the broken lines smaller than the preset second similarity threshold, and performing next iteration to know that the broken lines with the preset number of thresholds are found; the polyline pair with the preset number threshold may be a specified number limit, or may be a ratio between a qualified polyline and a candidate polyline.
S207, training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data.
And S208, training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map.
S209, training is carried out according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology.
And S210, training according to the training data set to construct a road network topology reconstruction and generate a confrontation network.
As an example, the generator, the first discriminator, the second discriminator and a VGG network are jointly trained to construct a road network topology reconstruction and generate a countermeasure network.
It should be noted that before training according to the training data set to construct the road network topology reconstruction and generate the countermeasure network, the design of the loss function may be further included.
As an example, the penalty function of the first discriminator is:
as an example, the penalty function of the second discriminator is:
wherein o, x and y represent the road network source image, the first initial road network map and the road reference image, respectively, G (-) represents the output of the generator, D (-) represents the output of the discriminator, P (-) represents the output of the discriminatordRepresenting the data distribution.
As an example, the loss function of the generator is:
LG(G)＝Lgr(G)+Lgt(G)+LVGG(G)
in the formula Lgr(G)，Lgt(G) Derived from the losses of the first and second discriminators, LVGG(G) The weighted sum of the pixel differences between the feature maps extracted by the VGG network under the L1 norm is expressed by the following calculation:
wherein D isr(·，·)，DtDenotes the outputs of the first and second discriminators, HKRepresents the output, λ, of the k-th layer of the pre-trained VGG networkkWeight, i, representing the k-th output of the VGG network1～imRepresenting m layers extracted from the VGG network.
As an example, the overall goal of the road network topology reconstruction to generate the countermeasure network is:
wherein L istotal＝Lr(G，D)+λtLt(G，D)+LVGG(G) I.e. the overall goal of the network is to minimize LG(G) Make the output of the generator and the reference data as similar as possible and maximize Lr(G, D) and Lt(G, D) enabling the discriminator to discriminate between the real data and the generated data.
Wherein L isVGG(G) Selecting the characteristic graphs of the 7 th, 12 th and 15 th layers of the VGG network for calculation, wherein the weight value lambda of the characteristic graphs iskAre all set as 1, LtotalMiddle Lt(G, D) weightλtIs 0.5.
S211, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction.
It should be noted that the above description about the road network topology reconstruction method in fig. 1 is also applicable to the road network topology reconstruction method, and is not repeated herein.
In summary, according to the road network topology reconstruction method of the embodiment of the present invention, firstly, original road network data is obtained; then, preprocessing the original road network data to generate a vector road network map consisting of vector lines; then, selecting at least one vector line as a reference line, and setting a buffer area around the reference line to obtain the vector line in the buffer area as a preselected object; next, the similarity S between the reference line and the preselected object is calculated1And judging the similarity S1Whether the similarity is larger than a preset first similarity threshold value or not, and determining the similarity S1The preselected object corresponding to the first similarity threshold value which is larger than the preset first similarity threshold value is used as a candidate line segment; then, each candidate line segment and the reference line are respectively expanded to generate corresponding candidate broken lines and reference broken lines, and the similarity S between the candidate broken lines and the reference broken lines is calculated2(ii) a Next, the similarity S is judged2Whether the similarity is less than a preset second similarity threshold value or not, and determining the similarity S2The candidate broken line and the reference broken line which correspond to the second similarity threshold value are expanded again to carry out the next iterative computation and similarity judgment until the similarity S is reached2The candidate broken lines corresponding to the second similarity threshold value which is greater than or equal to the preset threshold value reach a preset number threshold value; then, training the original road network data to construct a generator, wherein the generator generates first data information according to the original road network data; secondly, training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; then, according to the roadTraining the reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; then, training is carried out according to the training data set to construct a road network topology reconstruction and generate a confrontation network; then, acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; therefore, the road network topology of the rural environment is accurately reconstructed, the result is accurate, the training data is generated simply and easily through a simple road network matching algorithm, the training speed of generating the countermeasure network through road network topology reconstruction is increased, and the training difficulty of the network is reduced.
In order to implement the foregoing embodiments, the present invention further discloses a computer-readable storage medium, on which a road network topology reconstruction program is stored, which, when executed, implements the road network topology reconstruction method as described above.
In order to implement the foregoing embodiment, an embodiment of the present invention further discloses a road network topology rebuilding system, as shown in fig. 6, the road network topology rebuilding system includes: an acquisition module 10, a first building module 20, a second building module 30, a third building module 40, a generation module 50 and a reconstruction module 60.
The obtaining module 10 is configured to obtain original road network data, and generate a training data set according to the original road network data based on a road network matching algorithm, where the original road network data includes a first road network source image, a road reference image, and a first initial road network map corresponding to the first road network source image.
The first construction module 20 is used for training the original road network data to construct a generator, wherein the generator generates first data information from the original road network data.
The second construction module 30 is configured to train according to the road reference image, the first road network source image, and the first data information to construct a first discriminator, where the first discriminator is used to assist in generating a road network map.
The third building module 40 is configured to train according to the road reference image, the first initial road network map and the first data information to build a second discriminator, where the second discriminator is used to assist in reconstructing the road network topology.
The generating module 50 is configured to train according to the training data set to construct a road network topology reconstruction and generate a countermeasure network.
The reconstruction module 60 is configured to obtain a second road network source image, perform road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, and input the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, so as to perform the road network topology reconstruction.
It should be noted that the above description about the road network topology reconstruction method in fig. 1 is also applicable to the road network topology reconstruction system, and is not repeated herein.
In summary, the road network topology reconstruction system according to the embodiment of the present invention includes: the system comprises an acquisition module, a training module and a training module, wherein the acquisition module is used for acquiring original road network data and generating a training data set according to the original road network data based on a road network matching algorithm, and the original road network data comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image; the system comprises a first construction module, a second construction module and a third construction module, wherein the first construction module is used for training original road network data to construct a generator, and the generator generates first data information according to the original road network data; the second construction module is used for training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map; the third construction module is used for training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology; the generation module is used for training according to the training data set to construct road network topology reconstruction and generate a confrontation network; the reconstruction module is used for acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, inputting the second road network source image and the second initial road network map into a road network topology reconstruction to generate a confrontation network, and carrying out road network topology reconstruction; the method and the device have the advantages that the road network topology of the rural areas is reconstructed, the accuracy of the result is higher, meanwhile, the original road network data are matched through the road network matching algorithm, training data are easier to obtain, and the construction efficiency of the confrontation network generated by reconstructing the road network topology is improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
In the description of the present invention, it is to be understood that the terms "first", "second" and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intervening media, either internally or in any other relationship. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In the present invention, unless otherwise expressly stated or limited, the first feature "on" or "under" the second feature may be directly contacting the first and second features or indirectly contacting the first and second features through an intermediate. Also, a first feature "on," "over," and "above" a second feature may be directly or diagonally above the second feature, or may simply indicate that the first feature is at a higher level than the second feature. A first feature being "under," "below," and "beneath" a second feature may be directly under or obliquely under the first feature, or may simply mean that the first feature is at a lesser elevation than the second feature.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above should not be understood to necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A road network topology reconstruction method is characterized by comprising the following steps:
acquiring original road network data, and generating a training data set according to the original road network data based on a road network matching algorithm, wherein the training data set comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image;
training the training data set to construct a generator, wherein the generator generates first data information according to the original road network data;
training according to the road reference image, the first road network source image and the first data information to construct a first discriminator, wherein the first discriminator is used for assisting in generating a road network map;
training according to the road reference image, the first initial road network map and the first data information to construct a second discriminator, wherein the second discriminator is used for assisting in reconstructing road network topology;
training according to the training data set to construct a road network topology reconstruction and generate a confrontation network;
acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, and inputting the second road network source image and the second initial road network map into the road network topology reconstruction to generate a confrontation network for road network topology reconstruction.
2. The road network topology reconstruction method according to claim 1, wherein said generating a training data set from said original road network data based on a road network matching algorithm specifically comprises:
preprocessing the original road network data to generate a vector road network map consisting of vector lines;
selecting at least one vector line as a reference line, and selecting candidate line segments based on a buffer growth algorithm;
and adopting an iterative expansion line algorithm to expand and judge the similarity of the selected candidate line segment and the reference line so as to generate a final matching pair.
3. The road network topology reconstruction method according to claim 2, wherein said selecting at least one of said vector lines as a reference line and selecting candidate line segments based on a buffer growth algorithm specifically comprises:
selecting at least one vector line as a reference line, and arranging a buffer area around the reference line to obtain the vector line in the buffer area as a preselected object;
calculating a similarity S between the reference line and the preselected object1And judging the similarity S1Whether the similarity is greater than a preset first similarity threshold value or not, and comparing the similarity S1And the preselected object corresponding to the similarity threshold value larger than the preset first similarity threshold value is used as a candidate line segment.
4. The road network topology rebuilding method of claim 3, wherein said similarity S1Obtained by the following formula:
wherein, SimlenIs the ohm of two end points of the vector lineDistance of Kirschner, SimdisIs a short-edge-based median Hausdorff distance, SimoriIs the angle between the straight line and the horizontal axis, SimshapeAs a similar term of shape, wlen，wdis，wori，wshapeAre each Simlen，Simdis，Simori，SimshapeThe weight of (c).
5. The road network topology reconstruction method according to claim 2, wherein said expanding and similarity determining the selected candidate line segments and the reference line by using an iterative expanded line algorithm specifically comprises:
respectively expanding each candidate line segment and the reference line to generate corresponding candidate broken lines and reference broken lines, and calculating the similarity S between the candidate broken lines and the reference broken lines2；
Judging the similarity S2Whether the similarity is less than a preset second similarity threshold value or not, and determining the similarity S2And the candidate broken line and the reference broken line which correspond to the second similarity threshold value are expanded again to carry out the next iterative computation and similarity judgment until the similarity S is reached2And the candidate broken lines corresponding to the second similarity threshold value which is greater than or equal to the preset threshold value reach a preset number threshold value.
6. The road network topology rebuilding method of claim 5, wherein said similarity S2Obtained by the following formula:
wherein, SimplenIs the sum of Euclidean distances of all vector lines contained in the candidate folding line or the reference folding line, SimpdisA median Hausdorff distance, Sim, based on the short edge for the candidate polyline or the reference polylinepshapeAs a similar term of shape, wplen，wpdis，wpshapeAre each Simplen，Simpdis，SimpshapeThe weight of (c).
7. The road network topology reconstruction method according to claim 1, wherein said generator is composed of an encoder and a decoder, wherein said encoder comprises 3 convolutional layers, 4 residual blocks and 2 anti-convolutional layers, and said decoder adopts a symmetrical structure with said encoder, so that the input and output of said generator have the same resolution.
8. The road network topology reconstruction method according to claim 1, wherein said training is performed according to said training data set to construct a road network topology reconstruction and generate a countermeasure network, specifically:
and performing combined training on the generator, the first discriminator, the second discriminator and a VGG network to construct the road network topology reconstruction and generate the confrontation network.
9. A computer-readable storage medium, having stored thereon a road network topology reconstruction program, which when executed implements the road network topology reconstruction method according to any one of claims 1 to 8.
10. A road network topology reconstruction system, comprising:
the system comprises an acquisition module, a comparison module and a comparison module, wherein the acquisition module is used for acquiring original road network data and generating a training data set according to the original road network data based on a road network matching algorithm, wherein the training data set comprises a first road network source image, a road reference image and a first initial road network map corresponding to the first road network source image;
a first construction module, configured to train the training data set to construct a generator, where the generator generates first data information according to the original road network data;
a second construction module, configured to train according to the road reference image, the first road network source image, and the first data information to construct a first discriminator, where the first discriminator is used to assist in generating a road network map;
a third construction module, configured to train according to the road reference image, the first initial road network map, and the first data information to construct a second discriminator, where the second discriminator is configured to assist in reconstructing a road network topology;
the generation module is used for training according to the training data set to construct road network topology reconstruction and generate a confrontation network;
and the reconstruction module is used for acquiring a second road network source image, carrying out road extraction on the second road network source image to generate a second initial road network map corresponding to the second road network source image, and inputting the second road network source image and the second initial road network map into the road network topology reconstruction to generate a confrontation network so as to carry out road network topology reconstruction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811275175.8A CN109583626B (en) | 2018-10-30 | 2018-10-30 | Road network topology reconstruction method, medium and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811275175.8A CN109583626B (en) | 2018-10-30 | 2018-10-30 | Road network topology reconstruction method, medium and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109583626A CN109583626A (en) | 2019-04-05 |
| CN109583626B true CN109583626B (en) | 2020-12-01 |
Family
ID=65921297
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811275175.8A Active CN109583626B (en) | 2018-10-30 | 2018-10-30 | Road network topology reconstruction method, medium and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109583626B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11087170B2 (en) * | 2018-12-03 | 2021-08-10 | Advanced Micro Devices, Inc. | Deliberate conditional poison training for generative models |
| CN111145540B (en) * | 2019-12-18 | 2021-09-03 | 福建工程学院 | Method and system for discovering topological connecting edges of urban road network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295604A (en) * | 2016-08-19 | 2017-01-04 | 厦门大学 | Remote sensing image road network extractive technique based on Federated filter |
| CN106778605A (en) * | 2016-12-14 | 2017-05-31 | 武汉大学 | Remote sensing image road net extraction method under navigation data auxiliary |
| CN107958302A (en) * | 2017-11-17 | 2018-04-24 | 中南大学 | Empirical path planing method based on virtual topology transportation network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE202011110851U1 (en) * | 2010-12-07 | 2016-11-09 | Google Inc. | Device of route guidance |
-
2018
- 2018-10-30 CN CN201811275175.8A patent/CN109583626B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295604A (en) * | 2016-08-19 | 2017-01-04 | 厦门大学 | Remote sensing image road network extractive technique based on Federated filter |
| CN106778605A (en) * | 2016-12-14 | 2017-05-31 | 武汉大学 | Remote sensing image road net extraction method under navigation data auxiliary |
| CN107958302A (en) * | 2017-11-17 | 2018-04-24 | 中南大学 | Empirical path planing method based on virtual topology transportation network |
Non-Patent Citations (1)
| Title |
|---|
| 城市区域路网交通瓶颈识别与预测;邓瑞;《中国优秀硕士学位论文全文数据库》;20121015;第C034-743页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109583626A (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108510532B (en) | Optical and SAR image registration method based on deep convolution GAN | |
| Zhao et al. | Building extraction from satellite images using mask R-CNN with building boundary regularization | |
| JP6739517B2 (en) | Lane recognition modeling method, device, storage medium and device, and lane recognition method, device, storage medium and device | |
| CN104700099B (en) | The method and apparatus for recognizing traffic sign | |
| CN103034982B (en) | Image super-resolution rebuilding method based on variable focal length video sequence | |
| CN104574347B (en) | Satellite in orbit image geometry positioning accuracy evaluation method based on multi- source Remote Sensing Data data | |
| CN111860336B (en) | High-resolution remote sensing image inclined ship target detection method based on position sensing | |
| CN111723732A (en) | Optical remote sensing image change detection method, storage medium and computing device | |
| CN111127538B (en) | Multi-view image three-dimensional reconstruction method based on convolution cyclic coding-decoding structure | |
| Galvanin et al. | Extraction of building roof contours from LiDAR data using a Markov-random-field-based approach | |
| CN113689445B (en) | High-resolution remote sensing building extraction method combining semantic segmentation and edge detection | |
| CN114022729A (en) | Heterogeneous image matching positioning method and system based on twin network and supervised training | |
| CN109583626B (en) | Road network topology reconstruction method, medium and system | |
| CN116206185A (en) | Lightweight small target detection method based on improved YOLOv7 | |
| CN110956601B (en) | Infrared image fusion method and device based on multi-sensor mode coefficients and computer readable storage medium | |
| CN113610905A (en) | Deep learning remote sensing image registration method based on subimage matching and application | |
| CN109558801B (en) | Road network extraction method, medium, computer equipment and system | |
| CN107358625B (en) | SAR image change detection method based on SPP Net and region-of-interest detection | |
| CN107392211A (en) | The well-marked target detection method of the sparse cognition of view-based access control model | |
| Liu et al. | A comparison of deep learning architectures for semantic mapping of very high resolution images | |
| CN112287906B (en) | Template matching tracking method and system based on depth feature fusion | |
| CN116883679B (en) | Ground object target extraction method and device based on deep learning | |
| CN117274627A (en) | Multi-temporal snow remote sensing image matching method and system based on image conversion | |
| CN112001291A (en) | Method and system for quickly extracting main river channel in gravel distribution area of flood fan | |
| CN109886988A (en) | A kind of measure, system, device and the medium of Microwave Imager position error |
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 |