CN113420597A - Method and device for identifying roundabout, electronic equipment and storage medium - Google Patents
Method and device for identifying roundabout, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113420597A CN113420597A CN202110566935.6A CN202110566935A CN113420597A CN 113420597 A CN113420597 A CN 113420597A CN 202110566935 A CN202110566935 A CN 202110566935A CN 113420597 A CN113420597 A CN 113420597A
- Authority
- CN
- China
- Prior art keywords
- roundabout
- road
- image
- road area
- module
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a device for identifying a roundabout, electronic equipment and a storage medium, wherein the method comprises the following steps: intercepting a first road area from a road network by using a preset sliding window; inputting the first road area into a pre-trained roundabout recognition model; under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area; identifying a roundabout in the target road area; and mapping the image data of the identified roundabout into vector data. The method for identifying the roundabout can improve the identification accuracy of the roundabout in the road network.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to a method and an apparatus for identifying a roundabout, an electronic device, and a storage medium.
Background
Roundabouts are a very special structural mode in a road network, and the roundabouts can be divided into the following parts according to the diameter size: ordinary roundabouts, miniature roundabouts and miniature roundabouts. Due to the differences of drawing synthesis, process standards and the like, the expression modes of the roundabout in different road network data are different. Due to the fact that the expression modes of the roundabout in different road network data are different, the precision of the roundabout in identification is low. However, the identification of the roundabout is of great significance to drawing synthesis, road network matching, electronic navigation and the like, so that the identification of the roundabout in road network data accurately is a problem which needs to be solved urgently by technical personnel in the field.
The existing method for identifying the roundabout mainly comprises the following two modes:
the first method is as follows: performing depth-first search on road network data according to the visual consistency by using a spatial index, and judging that a roundabout possibly exists when a search path has a loop; and then further judging whether the ring intersection really exists according to preset conditions such as road length, area and the like. The method is complicated in identification and low in accuracy of identification results.
The second method comprises the following steps: roundabout in a road network are detected by identifying circles or ellipses. In an actual road network, as shown by the roundabout in the complex road network shown in fig. 1, the roundabout are not all standard circles and ellipses, so that the recognition result in this way has low accuracy.
Disclosure of Invention
In view of the above problems, embodiments of the present invention are proposed to provide a method and an apparatus for identifying a roundabout, an electronic device, and a storage medium, which overcome the above problems or at least partially solve the above problems.
In a first aspect, an embodiment of the present invention discloses a method for identifying a roundabout, including: intercepting a first road area from a road network by using a preset sliding window; inputting the first road area into a pre-trained roundabout recognition model; under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area; identifying a roundabout in the target road area; and mapping the image data of the identified roundabout into vector data.
Optionally, after the step of intercepting the first road region from the road network with the preset sliding window, the method further includes:
identifying the number of roads and the number of nodes contained in the first road area;
judging whether the first road area contains a roundabout or not according to the number of the roads and the number of the nodes;
if yes, the step of inputting the first road area into a pre-trained roundabout model is executed;
and if not, ignoring the first road area.
Optionally, before the step of intercepting the first road region from the road network with the preset sliding window, the method further includes:
generating a plurality of training sample pairs, wherein each training sample pair comprises: the method comprises the steps that a first image containing a roundabout and a mark image corresponding to the first image are included;
and training the roundabout identification model according to the training sample pairs until the roundabout identification model reaches the preset prediction accuracy.
Optionally, the step of generating a plurality of training sample pairs includes:
in the road network road data in the vector format, marks form the marks of all road arc sections of the annular intersection;
combining the marked road arc sections with the interconnection relationship into at least one roundabout to form a marked roundabout;
intercepting a second road area from the road network by a selection frame with a preset size;
identifying a third road area containing a roundabout from each second road area;
and for each third road area, converting the third road area into a first image and generating a marking image for the first image according to each marking roundabout.
Optionally, for any training sample pair, the step of training the roundabout recognition model according to the training sample comprises:
performing feature extraction on a first image in the training sample pair to obtain image features;
dividing the first image into a plurality of first sub-images with second preset sizes;
screening target sub-images from the first sub-images;
for each target sub-image, identifying whether the target sub-image contains a roundabout; under the condition that the target subimages contain the roundabout, performing frame expansion on the target subimages;
identifying a roundabout in the target sub-image after the frame is expanded;
matching the recognition result with a corresponding roundabout in the marked image contained in the training sample pair;
and adjusting preset model parameters of the identification model of the roundabout according to the matching result.
Optionally, the adjacent road areas intercepted by the preset sliding window partially coincide.
In a second aspect, an embodiment of the present invention discloses an intersection identification device, including: the first intercepting module is used for intercepting a first road area from a road network by using a preset sliding window; the input module is used for inputting the first road area into a pre-trained roundabout recognition model; the second intercepting module is used for expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, and intercepting to obtain a target road area; the first identification module is used for identifying a roundabout in the target road area; and the mapping module is used for mapping the identified image data of the roundabout into vector data.
Optionally, the apparatus further comprises:
the second identification module is used for identifying the number of roads and the number of nodes contained in the first road area after the first interception module intercepts the road area from the road network by using a preset sliding window;
the judging module is used for judging whether the first road area contains a roundabout or not according to the number of the roads and the number of the nodes;
the execution module is used for executing the step that the input module inputs the first road area into a pre-trained roundabout model if the first road area is in the preset road area; and if not, ignoring the first road area.
Optionally, the apparatus further comprises:
a generating module, configured to generate a plurality of training sample pairs before the first cutting module cuts the first road region from the road network with a preset sliding window, where each training sample pair includes: the method comprises the steps that a first image containing a roundabout and a mark image corresponding to the first image are included;
and the training module is used for training the roundabout identification model according to each training sample pair until the roundabout identification model reaches the preset prediction accuracy.
Optionally, the generating module includes:
the first sub-module is used for marking the marks of all road arc sections forming a ring intersection in the road network road data in the vector format;
the second sub-module is used for combining the marked road arc sections with the interconnection relationship into at least one roundabout to form a marked roundabout;
the third submodule is used for intercepting a second road area from the road network by a selection frame with a preset size;
a fourth sub-module for identifying a third road zone containing a roundabout from each of the second road zones;
and the fifth sub-module is used for converting the third road area into a first image aiming at each third road area and generating a marked image for the first image according to each marked roundabout.
Optionally, the training module comprises:
the sixth submodule is used for extracting the characteristics of the first image in the training sample pair to obtain image characteristics;
a seventh sub-module, configured to divide the first image into a plurality of first sub-images of a second preset size;
an eighth sub-module for screening target sub-images from each of the first sub-images;
the ninth sub-module is used for identifying whether the target sub-images contain roundabout or not aiming at each target sub-image;
a tenth submodule, configured to perform frame expansion on the target sub-image when the target sub-image includes a roundabout;
the eleventh submodule is used for identifying the roundabout in the target sub-image after the frame is expanded;
the twelfth submodule is used for matching the recognition result with a corresponding roundabout in the marked image contained in the training sample pair;
and the thirteenth submodule is used for adjusting the preset model parameters of the roundabout identification model according to the matching result.
Optionally, the adjacent road areas intercepted by the preset sliding window partially coincide.
In a third aspect, an embodiment of the present invention discloses an electronic device, including: one or more processors; and one or more machine-readable media having instructions stored thereon; the instructions, when executed by the one or more processors, cause the processors to perform a roundabout identification method as any one of above.
In a fourth aspect, an embodiment of the present invention discloses a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements a roundabout identification method as described in any one of the above.
In the embodiment of the invention, a first road area is intercepted from a road network by a preset sliding window; inputting the first road area into a pre-trained roundabout recognition model; under the condition that the output result of the roundabout identification model indicates that the road area comprises the roundabout, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area; identifying a roundabout in a target road area; and mapping the image data of the identified roundabout into vector data. According to the method for identifying the roundabout, the characteristics of the road network data are extracted based on the deep neural network, so that the roundabout is detected, on one hand, a worker does not need to manually write a complex roundabout identification rule, and a large amount of human resources can be saved; on the other hand, the roundabout in the intercepted road area is identified through the roundabout identification model, and the accuracy of the roundabout identification can be improved.
Drawings
FIG. 1 is a schematic diagram of a roundabout in a complex road network;
FIG. 2 is a flowchart illustrating steps of a method for identifying a roundabout according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the output result;
FIG. 4 is a flow chart of steps of another method of roundabout identification according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a training sample pair according to an embodiment of the present invention;
fig. 6 is a block diagram of a circular intersection recognition device according to an embodiment of the present invention;
fig. 7 is a block diagram showing a structure of another intersection identification device according to the embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may 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 disclosure to those skilled in the art.
Referring to fig. 2, a flowchart illustrating steps of a method for identifying a roundabout according to an embodiment of the present invention is shown.
The method for identifying the roundabout in the embodiment of the invention can comprise the following steps:
step 201: and intercepting a first road area from the road network by using a preset sliding window.
The road data in the road network is stored in a vector format, and each road is composed of arc segments formed by a series of coordinate points. Because the dimension limitation cannot identify the whole road network data at one time, a sliding window with a fixed size is required to intercept road regions from the road network, the road intersected with the sliding window is subjected to interception processing, only the inner part of the window is protected, and only the road regions in the sliding window are identified each time. The size of the sliding window can be set by those skilled in the art according to practical requirements, and is not particularly limited in the embodiments of the present application.
In an optional embodiment, each time the sliding window is moved to intercept the first road region, a certain region is overlapped, so that adjacent road regions intercepted by the sliding window are partially overlapped, and the road region interception mode can avoid the omission of road network data.
Step 202: and inputting the first road area into a pre-trained roundabout recognition model.
The identification model of the roundabout is a deep neural network model, and the model is trained in advance through a large number of training samples to reach the preset accuracy and then is put into use. The trained roundabout recognition model can identify roundabout of various expression forms in the road area.
The core algorithm of the roundabout identification model can be a Mask R-CNN algorithm, the idea of the Mask R-CNN algorithm is simple, and FCN (Full Connected Network) is added on the basis of the original fast-rcnn algorithm to generate corresponding MASK () branches. The whole Mask R-CNN algorithm is very flexible, can be used for completing various tasks including target classification, target detection and the like, and has good expansibility and usability. The roundabout identification model is obtained by training based on roundabout data marked in a large number. For the training method of the roundabout identification model, the relevant description in the subsequent embodiment is referred to.
Step 203: and under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain the target road area.
And after the first road area is input into the roundabout identification model as input data, the roundabout identification model outputs a result, the output result is used for indicating whether the first road area contains the roundabout or not, and the roundabout is marked in the output result under the condition that the roundabout is contained. An exemplary output result diagram is shown in fig. 3, where the border of the roundabout marked with a rectangular box in fig. 3.
And under the condition that the output result of the roundabout identification model indicates that the roundabout is not included in the first road area, ending the identification process of the first road area, returning to execute the step 201, and intercepting the next first road area from the road network again to identify the roundabout.
Under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, the roundabout frame marked in the output result is expanded to the outside by a first preset size, and subsequently, the roundabout identification is carried out on the target road area in the expanded frame again, so that the identification accuracy can be improved through dual identification.
The first preset size may be set by a person skilled in the art according to actual requirements, and is not particularly limited in the embodiments of the present invention.
Step 204: a roundabout is identified in the target road area.
When identifying a roundabout in a target road area, firstly identifying all roads contained in the target road area and the position relation among the roads; and secondly, identifying to obtain the roundabout according to the position relation between the roads, wherein the identification range is limited in the frame, so that the identification accuracy can be ensured.
Step 205: and mapping the image data of the identified roundabout into vector data.
And the image data of the roundabout is mapped into vector data, so that the data can be conveniently and uniformly managed subsequently.
According to the method for identifying the roundabout, provided by the embodiment of the invention, a first road area is intercepted from a road network by using a preset sliding window; inputting the first road area into a pre-trained roundabout recognition model; under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area; identifying a roundabout in a target road area; and mapping the image data of the identified roundabout into vector data. According to the method for identifying the roundabout, the characteristics of the road network data are extracted based on the deep neural network, so that the roundabout is detected, on one hand, a worker does not need to manually write a complex roundabout identification rule, and a large amount of human resources can be saved; on the other hand, the roundabout in the intercepted road area is identified through the roundabout identification model, and the accuracy of the roundabout identification can be improved.
Referring to fig. 4, a flowchart illustrating steps of a method for identifying a roundabout according to an embodiment of the present invention is shown.
The method for identifying the roundabout in the embodiment of the invention can comprise the following steps:
step 401: a plurality of training sample pairs are generated.
Each training sample pair includes: the first image containing the roundabout and the mark image corresponding to the first image are included.
A schematic diagram of a training sample pair is shown in fig. 5, where the left image is the first image and the right image is the labeled image.
The specific number of training sample pairs generated can be flexibly set by those skilled in the art, and is not particularly limited in the embodiment of the present invention.
In an alternative embodiment, the step of generating a plurality of training sample pairs comprises the sub-steps of:
the first substep: in the road network road data in the vector format, the marks form the marks of all road arc sections of the annular intersection.
The road data in the road network is stored in a vector format, each road is composed of arc sections formed by a series of coordinate points, firstly, the road marks of the road arc sections forming the circular intersection in the road network are marked manually, and then all the marked road arc sections which are mutually connected are connected in series to form the circular intersection.
And a second substep: and combining the marked road arc sections with the interconnection relationship into at least one roundabout to form the marked roundabout.
And a third substep: and cutting out the second road area from the road network by using a selection frame with a preset size.
The preset size can be set by those skilled in the art according to actual requirements, and is not particularly limited in the embodiments of the present invention.
Because the dimension limitation cannot identify the whole road network data at one time, road areas are cut from the road network by a selection frame with a fixed size, the road areas intersected with the selection frame are cut off, only the road areas in the selection frame are reserved, and only the roundabout in the selection frame is identified each time. When a road region is cut out from the road network by the selection box, the cut-out adjacent road regions may be set to partially overlap.
And a fourth substep: from among the second road areas, a third road area including the roundabout is identified.
And a fifth substep: and for each third road area, converting the third road area into a first image and generating a marking image for the first image according to each marked roundabout.
Since the input format of the roundabout model is image data, and the roads in the road network are vector data, it is necessary to rasterize the vector data into image data, i.e., convert the third road region into the first image, and generate a labeled image for the first image. And inputting the converted first image and the labeled image as a training sample pair into a roundabout recognition model for training.
Step 402: and training the roundabout recognition model according to the training sample pairs until the roundabout recognition model reaches the preset prediction accuracy.
When the roundabout recognition model is trained according to the training sample pairs, one training sample pair is input into the roundabout recognition model every time, and the preset model parameters of the roundabout recognition model are adjusted according to the recognition effect after the training sample pair is input every time.
In an optional embodiment, the step of training the roundabout recognition model according to the training samples comprises the following sub-steps for any training sample pair:
the first substep: and performing feature extraction on the first image in the training sample pair to obtain image features.
In practical implementation, any suitable algorithm may be used for feature extraction, for example: ResNet50 FPN.
And a second substep: the first image is divided into a plurality of first sub-images with a second preset size.
The second preset size can be set by a person skilled in the art according to practical requirements, and is not particularly limited in the embodiments of the present invention.
And a third substep: and screening the target sub-images from the first sub-images.
The target sub-images are regions of interest, and the target sub-images may be one or more.
And a fourth substep: and identifying whether the target sub-images contain roundabout intersections or not aiming at each target sub-image.
The target sub-image may be divided into: the method comprises a first type and a second type, wherein the first type comprises an object sub-image of the roundabout, and the second type does not comprise the object sub-image of the roundabout.
And a fifth substep: and under the condition that the target subimages contain the roundabout, performing frame expansion on the target subimages.
The frame expansion of the target sub-image can improve the integrity of the roundabout contained in the target sub-image.
And a sixth substep: and identifying a roundabout in the target sub-image after the frame is expanded.
And a seventh substep: and matching the recognition result with the corresponding roundabout in the marked image contained in the training sample pair.
And a substep eight: and adjusting preset model parameters of the identification model of the roundabout according to the matching result.
And determining the detection effect of the model according to the matching result, wherein the detection effect can be represented by AP. AP in the modelIOUWhen AP reaches 0.996 at 0.902 of 0.50:0.95 and 0.902 of IoU, the model may reach 98.5% + 85.8% of assessment.
Step 403: and intercepting a first road area from the road network by using a preset sliding window.
In an optional embodiment, after the step of cutting out the first road region from the road network with the preset sliding window, the following steps may be further included:
firstly, identifying the number of roads and the number of nodes contained in a first road area;
secondly, judging whether a first road area contains a roundabout or not according to the number of the roads and the number of the nodes;
when judging whether the first road area contains the roundabout or not by combining the number of the roads and the number of the nodes, the judgment can be carried out based on a preset roundabout identification rule.
Finally, if yes, executing step 404 to input the first road area into a pre-trained roundabout model; if not, ignoring the first road region.
In the optional mode, the roundabout in the first road area is roughly identified before the first road area is input into the roundabout identification model, and the roundabout is input into the roundabout model for rough identification after the first road area is determined to contain the roundabout, so that the calculation load can be reduced.
Step 404: and inputting the first road area into a pre-trained roundabout recognition model.
Step 405: and under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain the target road area.
Step 406: a roundabout is identified in the target road area.
Step 407: and mapping the image data of the identified roundabout into vector data.
The specific contents of step 404 to step 407 may refer to the related descriptions in step 201 to step 205, which are not described in detail in the embodiment of the present invention.
According to the method for identifying the roundabout, the characteristics of the road network data are extracted based on the deep neural network, so that the roundabout is detected, on one hand, a worker does not need to manually write a complex roundabout identification rule, and a large amount of human resources can be saved; on the other hand, the roundabout in the intercepted road area is identified through the roundabout identification model, and the accuracy of the roundabout identification can be improved. In addition, the method for identifying the roundabout roughly identifies the roundabout in the first road area before inputting the first road area into the roundabout identification model, and roughly identifies the roundabout after determining that the first road area comprises the roundabout and inputting the first road area into the roundabout model, so that the calculation load can be reduced.
Referring to fig. 6, a block diagram of a circular intersection recognition device according to an embodiment of the present invention is shown.
The roundabout identification device of the embodiment of the invention can comprise the following modules:
the first intercepting module 601 is configured to intercept a first road region from a road network by using a preset sliding window;
an input module 602, configured to input the first road area into a pre-trained roundabout recognition model;
a second intercepting module 603, configured to, when the output result of the roundabout identification model indicates that the road area includes a roundabout, expand a frame of the roundabout marked by the output result to an outside by a first preset size in the first road area, and intercept the frame to obtain a target road area;
a first identification module 604 for identifying a roundabout in the target road region;
a mapping module 605 for mapping the image data of the identified roundabout into vector data.
The roundabout identification device provided by the embodiment of the invention intercepts a first road area from a road network by using a preset sliding window; inputting the first road area into a pre-trained roundabout recognition model; under the condition that the output result of the roundabout identification model indicates that the road area comprises the roundabout, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area; identifying a roundabout in a target road area; and mapping the image data of the identified roundabout into vector data. According to the roundabout identification device provided by the embodiment of the invention, the characteristics of road network data are extracted and the roundabout is detected based on the deep neural network, so that on one hand, a worker does not need to manually write a complex roundabout identification rule, and a large amount of manpower resources can be saved; on the other hand, the roundabout in the intercepted road area is identified through the roundabout identification model, and the accuracy of the roundabout identification can be improved.
Referring to fig. 7, a block diagram of another roundabout identification device according to an embodiment of the present invention is shown.
The roundabout recognition device of the embodiment of the present invention is further optimized for the roundabout recognition device shown in fig. 6, and the optimized roundabout recognition device may include the following modules:
the first intercepting module 701 is used for intercepting a first road region from a road network by using a preset sliding window;
an input module 702, configured to input the first road area into a pre-trained roundabout recognition model;
a second intercepting module 703, configured to, when the output result of the roundabout identification model indicates that a roundabout is included in the first road area, extend a frame of the roundabout marked by the output result to an outside by a first preset size in the first road area, and intercept the frame to obtain a target road area;
a first identification module 704 for identifying a roundabout in the target road region;
a mapping module 705 for mapping the identified image data of the roundabout into vector data.
Optionally, the apparatus further comprises:
a second identifying module 706, configured to identify the number of roads and the number of nodes included in the first road region after the first intercepting module 701 intercepts the road region from the road network with a preset sliding window;
a judging module 707, configured to judge whether the first road area includes a roundabout according to the number of roads and the number of nodes;
an executing module 708, configured to execute the step of inputting the first road area into a pre-trained roundabout model by the input module if yes; and if not, ignoring the first road area.
Optionally, the apparatus further comprises:
a generating module 709, configured to generate a plurality of training sample pairs before the first intercepting module 701 intercepts a first road region from a road network with a preset sliding window, where each training sample pair includes: the method comprises the steps that a first image containing a roundabout and a mark image corresponding to the first image are included;
and the training module 710 is configured to train the roundabout identification model according to each training sample pair until the roundabout identification model reaches a preset prediction accuracy.
Optionally, the generating module 709 includes:
the first sub-module 7091 is configured to mark the identifiers of the road arc segments forming the ring intersection in the road network road data in the vector format;
the second sub-module 7092 is configured to combine the marked road arcs having an interconnection relationship into at least one roundabout to form a marked roundabout;
a third sub-module 7093, configured to intercept a second road region from the road network with a selection box of a preset size;
a fourth sub-module 7094 for identifying, from each of the second road areas, a third road area including a roundabout;
a fifth sub-module 7095 is configured to, for each of the third road areas, convert the third road area into a first image and generate a labeled image for the first image according to each labeled roundabout.
Optionally, the training module 710 includes:
a sixth sub-module 7101, configured to perform feature extraction on the first image in the training sample pair to obtain an image feature;
a seventh sub-module 7102 for dividing said first image into a plurality of first sub-images of a second preset size;
an eighth sub-module 7103 for screening a target sub-image from each of said first sub-images;
a ninth sub-module 7104, configured to, for each target sub-image, identify whether the target sub-image includes a roundabout;
a tenth submodule 7105, configured to perform frame expansion on the target sub-image when the target sub-image includes a roundabout;
an eleventh submodule 7106, configured to identify a roundabout in the target sub-image after the frame is expanded; a twelfth sub-module 7107 for matching the recognition result with the corresponding roundabout in the labeled image included in the training sample pair;
and a thirteenth sub-module 7108, configured to adjust preset model parameters of the roundabout identification model according to the matching result.
Optionally, the adjacent road areas intercepted by the preset sliding window partially coincide.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
In an embodiment of the invention, an electronic device is also provided. The electronic device may include one or more processors and one or more machine-readable media having instructions, such as an application program, stored thereon. The instructions, when executed by the one or more processors, cause the processors to perform the roundabout identification method described above.
In an embodiment of the present invention, there is also provided a non-transitory computer-readable storage medium having stored thereon a computer program executable by a processor of an electronic device to perform the roundabout identification method described above. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of 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, embodiments of 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.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (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 terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, 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 terminal 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 terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications of these 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 embodiments of the invention.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.
The method and the device for identifying the roundabout, the electronic device and the storage medium provided by the invention are described in detail, a specific example is applied in the text to explain the principle and the implementation mode of the invention, and the description of the embodiment is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (14)
1. A method for identifying a circular intersection is characterized by comprising the following steps:
intercepting a first road area from a road network by using a preset sliding window;
inputting the first road area into a pre-trained roundabout recognition model;
under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area, and intercepting to obtain a target road area;
identifying a roundabout in the target road area;
and mapping the image data of the identified roundabout into vector data.
2. The method according to claim 1, wherein after the step of intercepting the first road region from the road network with a preset sliding window, the method further comprises:
identifying the number of roads and the number of nodes contained in the first road area;
judging whether the first road area contains a roundabout or not according to the number of the roads and the number of the nodes;
if yes, the step of inputting the first road area into a pre-trained roundabout model is executed;
and if not, ignoring the first road area.
3. The method according to claim 1, wherein prior to the step of intercepting the first road region from the road network with a preset sliding window, the method further comprises:
generating a plurality of training sample pairs, wherein each training sample pair comprises: the method comprises the steps that a first image containing a roundabout and a mark image corresponding to the first image are included;
and training the roundabout identification model according to the training sample pairs until the roundabout identification model reaches the preset prediction accuracy.
4. The method of claim 3, wherein the step of generating a plurality of training sample pairs comprises:
in the road network road data in the vector format, marks form the marks of all road arc sections of the annular intersection;
combining the marked road arc sections with the interconnection relationship into at least one roundabout to form a marked roundabout;
intercepting a second road area from the road network by a selection frame with a preset size;
identifying a third road area containing a roundabout from each second road area;
and for each third road area, converting the third road area into a first image and generating a marking image for the first image according to each marking roundabout.
5. The method according to claim 1, wherein the step of training the roundabout recognition model according to the training samples comprises, for any training sample pair:
performing feature extraction on a first image in the training sample pair to obtain image features;
dividing the first image into a plurality of first sub-images with second preset sizes;
screening target sub-images from the first sub-images;
for each target sub-image, identifying whether the target sub-image contains a roundabout; under the condition that the target subimages contain the roundabout, performing frame expansion on the target subimages;
identifying a roundabout in the target sub-image after the frame is expanded;
matching the recognition result with a corresponding roundabout in the marked image contained in the training sample pair;
and adjusting preset model parameters of the identification model of the roundabout according to the matching result.
6. The method according to claim 1, characterized in that the adjacent road areas intercepted by the preset sliding window partially coincide.
7. An annular intersection recognition device, comprising:
the first intercepting module is used for intercepting a first road area from a road network by using a preset sliding window;
the input module is used for inputting the first road area into a pre-trained roundabout recognition model;
the second intercepting module is used for expanding the frame of the roundabout marked by the output result to the outside by a first preset size in the first road area under the condition that the output result of the roundabout identification model indicates that the roundabout is included in the first road area, and intercepting to obtain a target road area;
the first identification module is used for identifying a roundabout in the target road area;
and the mapping module is used for mapping the identified image data of the roundabout into vector data.
8. The apparatus of claim 7, further comprising:
the second identification module is used for identifying the number of roads and the number of nodes contained in a first road area after the first road area is intercepted from a road network by the first interception module through a preset sliding window;
the judging module is used for judging whether the first road area contains a roundabout or not according to the number of the roads and the number of the nodes;
the execution module is used for executing the step that the input module inputs the first road area into a pre-trained roundabout model if the first road area is in the preset road area; and if not, ignoring the first road area.
9. The apparatus of claim 7, further comprising:
a generating module, configured to generate a plurality of training sample pairs before the first cutting module cuts the first road region from the road network with a preset sliding window, where each training sample pair includes: the method comprises the steps that a first image containing a roundabout and a mark image corresponding to the first image are included;
and the training module is used for training the roundabout identification model according to each training sample pair until the roundabout identification model reaches the preset prediction accuracy.
10. The apparatus of claim 9, wherein the generating module comprises:
the first sub-module is used for marking the marks of all road arc sections forming a ring intersection in the road network road data in the vector format;
the second sub-module is used for combining the marked road arc sections with the interconnection relationship into at least one roundabout to form a marked roundabout;
the third submodule is used for intercepting a second road area from the road network by a selection frame with a preset size;
a fourth sub-module for identifying a third road zone containing a roundabout from each of the second road zones;
and the fifth sub-module is used for converting the third road area into a first image aiming at each third road area and generating a marked image for the first image according to each marked roundabout.
11. The apparatus of claim 9, wherein the training module comprises:
the sixth submodule is used for extracting the characteristics of the first image in the training sample pair to obtain image characteristics;
a seventh sub-module, configured to divide the first image into a plurality of first sub-images of a second preset size;
an eighth sub-module for screening target sub-images from each of the first sub-images;
the ninth sub-module is used for identifying whether the target sub-images contain roundabout or not aiming at each target sub-image;
a tenth submodule, configured to perform frame expansion on the target sub-image when the target sub-image includes a roundabout;
the eleventh submodule is used for identifying the roundabout in the target sub-image after the frame is expanded;
the twelfth submodule is used for matching the recognition result with a corresponding roundabout in the marked image contained in the training sample pair;
and the thirteenth submodule is used for adjusting the preset model parameters of the roundabout identification model according to the matching result.
12. The apparatus according to claim 7, wherein the preset sliding window intercepts the adjacent road area partially overlapping.
13. An electronic device, comprising:
one or more processors; and
one or more machine-readable media having instructions stored thereon;
the instructions, when executed by the one or more processors, cause the processors to perform the roundabout identification method of any one of claims 1 to 6.
14. A computer-readable storage medium, characterized in that a computer program is stored thereon, which program, when being executed by a processor, carries out the roundabout identification method according to any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110566935.6A CN113420597A (en) | 2021-05-24 | 2021-05-24 | Method and device for identifying roundabout, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110566935.6A CN113420597A (en) | 2021-05-24 | 2021-05-24 | Method and device for identifying roundabout, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113420597A true CN113420597A (en) | 2021-09-21 |
Family
ID=77712865
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110566935.6A Withdrawn CN113420597A (en) | 2021-05-24 | 2021-05-24 | Method and device for identifying roundabout, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113420597A (en) |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109919010A (en) * | 2019-01-24 | 2019-06-21 | 北京三快在线科技有限公司 | Image processing method and device |
CN110020651A (en) * | 2019-04-19 | 2019-07-16 | 福州大学 | Car plate detection localization method based on deep learning network |
CN110175574A (en) * | 2019-05-28 | 2019-08-27 | 中国人民解放军战略支援部队信息工程大学 | A kind of Road network extraction method and device |
CN110276346A (en) * | 2019-06-06 | 2019-09-24 | 北京字节跳动网络技术有限公司 | Target area identification model training method, device and computer readable storage medium |
CN110688958A (en) * | 2019-09-28 | 2020-01-14 | 中国测绘科学研究院 | Intersection identification method based on GoogleLeNet neural network |
CN110956147A (en) * | 2019-12-05 | 2020-04-03 | 京东数字科技控股有限公司 | Method and device for generating training data |
CN111382625A (en) * | 2018-12-28 | 2020-07-07 | 沈阳美行科技有限公司 | Road sign identification method and device and electronic equipment |
WO2020151329A1 (en) * | 2019-01-23 | 2020-07-30 | 平安科技(深圳)有限公司 | Target detection based identification box determining method and device and terminal equipment |
US20200257914A1 (en) * | 2017-11-20 | 2020-08-13 | Tencent Technology (Shenzhen) Company Limited | Living body recognition method, storage medium, and computer device |
CN111598091A (en) * | 2020-05-20 | 2020-08-28 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and computer readable storage medium |
-
2021
- 2021-05-24 CN CN202110566935.6A patent/CN113420597A/en not_active Withdrawn
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200257914A1 (en) * | 2017-11-20 | 2020-08-13 | Tencent Technology (Shenzhen) Company Limited | Living body recognition method, storage medium, and computer device |
CN111382625A (en) * | 2018-12-28 | 2020-07-07 | 沈阳美行科技有限公司 | Road sign identification method and device and electronic equipment |
WO2020151329A1 (en) * | 2019-01-23 | 2020-07-30 | 平安科技(深圳)有限公司 | Target detection based identification box determining method and device and terminal equipment |
CN109919010A (en) * | 2019-01-24 | 2019-06-21 | 北京三快在线科技有限公司 | Image processing method and device |
CN110020651A (en) * | 2019-04-19 | 2019-07-16 | 福州大学 | Car plate detection localization method based on deep learning network |
CN110175574A (en) * | 2019-05-28 | 2019-08-27 | 中国人民解放军战略支援部队信息工程大学 | A kind of Road network extraction method and device |
CN110276346A (en) * | 2019-06-06 | 2019-09-24 | 北京字节跳动网络技术有限公司 | Target area identification model training method, device and computer readable storage medium |
CN110688958A (en) * | 2019-09-28 | 2020-01-14 | 中国测绘科学研究院 | Intersection identification method based on GoogleLeNet neural network |
CN110956147A (en) * | 2019-12-05 | 2020-04-03 | 京东数字科技控股有限公司 | Method and device for generating training data |
CN111598091A (en) * | 2020-05-20 | 2020-08-28 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and computer readable storage medium |
Non-Patent Citations (3)
Title |
---|
崔晓杰等: "基于改进霍夫变换的环形交叉口识别方法", 《测绘学报》 * |
张鸿刚: "一种基于GoogLeNet神经网络的复杂交叉路口识别与简化方法", 《中国优秀硕士学位论文全文数据库(信息科技辑)》 * |
张鸿刚等: "GoogLeNet神经网络的复杂交叉路口识别方法", 《测绘科学》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11853883B2 (en) | System and method for instance-level lane detection for autonomous vehicle control | |
US10223585B2 (en) | Page segmentation of vector graphics documents | |
CN111597861A (en) | System and method for automatically interpreting ground object of remote sensing image | |
CN110188495B (en) | Method for generating three-dimensional house type graph based on two-dimensional house type graph of deep learning | |
CN109977337B (en) | Webpage design comparison method, device and equipment and readable storage medium | |
CN109447992B (en) | Image segmentation method and device | |
CN108197567B (en) | Method, apparatus and computer readable medium for image processing | |
CN113657409A (en) | Vehicle loss detection method, device, electronic device and storage medium | |
JP2023527615A (en) | Target object detection model training method, target object detection method, device, electronic device, storage medium and computer program | |
CN111649749A (en) | Navigation method based on BIM (building information modeling), electronic equipment and related product | |
CN115546809A (en) | Table structure identification method based on cell constraint and application thereof | |
CN111127582B (en) | Track overlapping section identification method, device, system and storage medium | |
CN115393625A (en) | Semi-supervised training of image segmentation from coarse markers | |
CN110874170A (en) | Image area correction method, image segmentation method and device | |
CN113420597A (en) | Method and device for identifying roundabout, electronic equipment and storage medium | |
CN112486384A (en) | Picture examination processing method and related device | |
KR101411893B1 (en) | Automatic Recognition Method of Direction Information in Road Sign Image | |
CN114512052B (en) | Method and device for generating diverging and converging intersections by fusing remote sensing images and track data | |
Palamakumbure et al. | ArcGIS V. 10 landslide susceptibility data mining add-in tool integrating data mining and GIS techniques to model landslide susceptibility | |
KR102301635B1 (en) | Method of inferring bounding box using artificial intelligence model and computer apparatus of inferring bounding box | |
CN112184742B (en) | Method, device, equipment and computer readable medium for detecting graph common edge | |
CN108961263A (en) | A kind of visualized graphs quickly apply the method and system with transmission | |
CN114140551A (en) | Expressway bifurcation merging point conjecture method and system based on track image | |
CN115512375A (en) | Training method of text error correction model, text recognition method and related equipment | |
CN115049899B (en) | Model training method, reference expression generation method and related equipment |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210921 |