CN113094422B - Urban road traffic flow map generation method, system and equipment - Google Patents
Urban road traffic flow map generation method, system and equipment Download PDFInfo
- Publication number
- CN113094422B CN113094422B CN202110273734.7A CN202110273734A CN113094422B CN 113094422 B CN113094422 B CN 113094422B CN 202110273734 A CN202110273734 A CN 202110273734A CN 113094422 B CN113094422 B CN 113094422B
- Authority
- CN
- China
- Prior art keywords
- map
- traffic flow
- traffic
- feature
- generating
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000010586 diagram Methods 0.000 claims description 32
- 230000008569 process Effects 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 17
- 230000007613 environmental effect Effects 0.000 claims description 14
- 230000004927 fusion Effects 0.000 claims description 13
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/26—Visual data mining; Browsing structured data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- 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
- G06N3/042—Knowledge-based neural networks; Logical representations of neural 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
- G06N3/04—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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
- Y02A30/60—Planning or developing urban green infrastructure
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Tourism & Hospitality (AREA)
- Analytical Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Remote Sensing (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a method, a system and equipment for generating an urban road traffic flow map. The invention inputs the traffic network feature map generated by the first encoder and the second traffic flow feature map generated by the second encoder into the decoder for decoding to generate a third traffic flow feature map, and generates a fine-granularity traffic flow map based on the first traffic flow feature map and the third traffic flow feature map. According to the invention, the first encoder is used for generating the traffic road network characteristic map as the priori knowledge for generating the fine-granularity traffic flow map, and the priori knowledge is explicitly encoded in the decoder, so that the guidance function of the priori knowledge in the generation of the fine-granularity traffic flow map is fully exerted, the urban traffic flow distribution mode is fully explored, and excellent performance and accuracy can be obtained on the high-definition urban traffic flow map generation task.
Description
Technical Field
The present invention relates to the field of traffic, and in particular, to a method, a system, and an apparatus for generating an urban road traffic flow map.
Background
Urban traffic flow monitoring is an important component in smart city construction, and can provide an important information basis for public safety, urban planning and early warning of other tasks. In modern information society, vehicles and infrastructure (e.g., traffic monitoring, air quality monitoring stations) are constantly generating vast amounts of city data in heterogeneous formats, such as GPS track points, weather data, and the like.
Deep learning and data mining make these data significant in exploring urban dynamic patterns and performance predictions. But in general, fine-grained real city data is difficult to collect. For example, monitoring of urban traffic flow requires deployment of sufficient sensor equipment throughout the city, thousands of probes, loop sensors, surveillance cameras and other monitoring equipment must be deployed throughout the traffic segment to monitor traffic flow and collect traffic data in real-time. Real-time road traffic monitoring systems are highly dependent on these monitoring devices, requiring significant power resources and reliable disaster recovery capability for long-term stable operation, which can result in significant expense and high maintenance costs.
Today, the development of smart cities worldwide is changing, and the corresponding traffic flow monitoring costs are also increasing exponentially, which may become a key factor limiting further development. To address this problem and promote a better development of smart city applications, a new approach is needed to reduce monitoring costs while ensuring granularity and accuracy of traffic flow data. Therefore, how to infer the original fine-grained flow distribution from easily obtained coarse-grained data is an important issue.
In recent years, deep neural networks have been widely used for urban traffic data refinement reasoning. The existing research mainly models the problem as a mapping problem of mapping data with low information entropy to data with high information entropy. These efforts typically divide the area of investigation into grid maps according to longitude and latitude coordinates and organize the collected traffic data into tensors for direct input into a convolutional network for mapping automatic representation learning. Some strategies also propose to establish spatial constraints between coarse-granularity stream data and fine-granularity stream data in order to learn a representation of spatial correlation.
However, existing approaches ignore the importance of a priori knowledge to this problem, such as urban road networks, which have little significant change over a long period of time. Roads are an important component of urban traffic systems, and urban road networks are originally designed for vehicular travel, meaning that most traffic flows will occur on roads. For example, a vehicle such as a bus is usually driven only on a specific road and does not appear in a residential area. The influence of priori knowledge in the process of generating the high-granularity urban traffic flow map is not fully considered in the prior art, so that the accuracy of the generated high-granularity urban traffic flow map is relatively poor.
In summary, the influence of the priori knowledge in the process of generating the high-granularity urban traffic flow map is not fully considered in the prior art, so that the technical problem that the accuracy of the generated high-granularity urban traffic flow map is poor exists.
Disclosure of Invention
The invention provides a method, a system and equipment for generating an urban road traffic flow map.
In order to solve the technical problems, the embodiment of the invention provides a method for generating an urban road traffic flow map, which comprises the following steps:
acquiring a coarse-grained traffic flow map of a target area, environment data of the target area and a traffic map of the target area;
inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map;
generating a first traffic flow characteristic map based on the traffic road network characteristic map, the environmental data and the coarse-granularity traffic flow map;
inputting the first traffic flow characteristic map into a second encoder for encoding to generate a second traffic flow characteristic map;
inputting the second traffic flow characteristic diagram and the traffic road network characteristic diagram into a decoder for decoding to generate a third traffic flow characteristic diagram;
and generating a fine-granularity traffic flow map based on the first traffic flow characteristic map and the third traffic flow characteristic map.
Preferably, the specific process of inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map is as follows:
generating a traffic road network map based on the traffic map;
and weighting the traffic road network graph and the coarse-granularity traffic flow graph to obtain a weighted first traffic road network graph, and inputting the first traffic road network graph into a first encoder so that the first encoder encodes road features in the first traffic road network graph to obtain a traffic road network feature graph.
Preferably, the first encoder performs feature extraction on the first traffic network graph, and the specific process of obtaining the traffic network feature graph includes:
the first encoder carries out horizontal convolution, vertical convolution, opposite angle convolution and opposite angle convolution on the first traffic network graph respectively to generate horizontal features, vertical features, opposite angle features and opposite angle features, and a traffic network feature graph is obtained based on the horizontal features, the vertical features, the opposite angle features and the opposite angle features.
Preferably, the specific process of generating the first traffic flow feature map based on the traffic road network feature map, the environmental data and the coarse-granularity traffic flow map is as follows:
converting the environmental data into an environmental factor embedding vector;
extracting features of the traffic road network feature map to generate a road network feature map;
and carrying out feature fusion on the road network feature map, the external factor embedded vector and the coarse-granularity traffic flow map to generate a first traffic flow feature map.
Preferably, feature fusion is performed on the road network feature map, the external factor embedded vector and the coarse-granularity traffic flow map, and the specific process of generating the first traffic flow feature map is as follows:
feature cascading is carried out on the external factor embedded vector and the coarse-granularity traffic flow map to obtain a first cascade feature, and the first cascade feature is fused to obtain a traffic flow primary feature map;
and carrying out feature cascading on the road network feature map and the traffic flow primary feature map to obtain a second cascading feature, and fusing the second cascading feature to obtain the first traffic flow feature map.
Preferably, the specific process of inputting the first traffic flow characteristic map into the second encoder to encode and generating the second traffic flow characteristic map is as follows:
and inputting the first traffic flow characteristic map into a second encoder so that the second encoder encodes the first traffic flow characteristic map based on a multi-head attention mechanism to generate a second traffic flow characteristic map.
Preferably, the specific process of inputting the second traffic flow characteristic map and the traffic road network characteristic map into a decoder to decode and generating the third traffic flow characteristic map is as follows:
and taking the traffic road network feature map as an embedded code, and inputting the embedded code and the second traffic flow feature map into a decoder so that the decoder decodes based on a multi-head attention mechanism to obtain a third traffic flow feature map.
Preferably, based on the first traffic flow feature map and the third traffic flow feature map, the specific process of generating the fine-grained traffic flow map is as follows:
and carrying out feature fusion on the first traffic flow feature map and the third traffic flow feature map to generate a fine-granularity traffic flow map.
The embodiment of the invention also provides a system for generating the urban road traffic flow map, which comprises a data acquisition module, a traffic road network feature map generation module, a first traffic flow feature map generation module, a second traffic flow feature map generation module, a third traffic flow feature map generation module and a fine-granularity traffic flow map generation module;
the data acquisition module is used for acquiring a coarse-grained traffic flow map of the target area, environment data of the target area and a traffic map of the target area;
the traffic road network feature map generating module is used for inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map;
the first traffic flow characteristic map generating module is used for generating a first traffic flow characteristic map based on the traffic road network characteristic map, the environment data and the coarse-granularity traffic flow map;
the second traffic flow characteristic map generating module is used for inputting the first traffic flow characteristic map into a second encoder for encoding to generate a second traffic flow characteristic map;
the third traffic flow characteristic map generating module is used for inputting the second traffic flow characteristic map and the traffic road network characteristic map into a decoder for decoding to generate a third traffic flow characteristic map;
the fine-granularity traffic flow map generation module is used for generating a fine-granularity traffic flow map based on the first traffic flow characteristic map and the third traffic flow characteristic map.
The embodiment of the invention also provides urban road traffic flow map generating equipment, which comprises a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is used for executing the urban road traffic flow map generating method according to the instructions in the program codes.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
according to the embodiment of the invention, the traffic network feature map generated by the first encoder and the second traffic flow feature map generated by the second encoder are input into the decoder for decoding, the third traffic flow feature map is generated, and the fine-granularity traffic flow map is generated based on the first traffic flow feature map and the third traffic flow feature map. According to the embodiment of the invention, the first encoder is used for generating the traffic road network characteristic map as the priori knowledge for generating the fine-granularity traffic flow map, the priori knowledge is explicitly encoded in the decoder, the guiding function of the priori knowledge in the generation of the fine-granularity traffic flow map is fully exerted, the urban traffic flow distribution mode is fully explored, and excellent performance and accuracy can be obtained on the high-definition urban traffic flow map generation task.
Drawings
Fig. 1: the flow chart of the urban road traffic flow chart generation method is provided for the embodiment of the invention.
Fig. 2: the embodiment of the invention provides a structural schematic diagram of an urban road traffic flow map generation model (RAFM).
Fig. 3: the embodiment of the invention provides a structure schematic diagram of a one-dimensional convolution layer in an urban road traffic flow map generation model.
Fig. 4: the embodiment of the invention provides a structural schematic diagram of a one-dimensional residual error module layer in an urban road traffic flow map generation model.
Fig. 5: the embodiment of the invention provides a structural schematic diagram of an urban road traffic flow map generating system.
Fig. 6: the embodiment of the invention provides a structural schematic diagram of urban road traffic flow map generating equipment.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
Referring to fig. 1, a flowchart of a method for generating an urban road traffic flow map according to an embodiment of the invention includes the following steps:
s101, acquiring a coarse-grained traffic flow map of a target area, environment data of the target area and a traffic map of the target area.
It should be further noted that, since environmental data such as weather, wind speed, temperature, time, etc. have a complex and important influence on urban traffic flow distribution, in this embodiment, the influence of the environmental data of the target area in the generation process of the fine-grained traffic flow map needs to be considered.
S102, inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding, generating a traffic road network characteristic map, and taking the traffic road network characteristic map as priori knowledge for generating a fine-granularity traffic flow map.
And S103, generating a first traffic flow characteristic diagram based on the traffic network characteristic diagram, the environment data and the coarse-granularity traffic flow diagram.
The method further includes the step of fusing the characteristics of the traffic road network characteristic diagram, the environment data and the characteristics of the coarse-granularity traffic flow diagram to generate a first traffic flow characteristic diagram.
S104, inputting the first traffic flow characteristic diagram into a second encoder for encoding, and generating a second traffic flow characteristic diagram.
S105, inputting the second traffic flow characteristic diagram and the traffic road network characteristic diagram into a decoder for decoding, and explicitly encoding priori knowledge so as to generate a third traffic flow characteristic diagram.
And S106, carrying out feature fusion on the first traffic flow feature map and the third traffic flow feature map to generate a fine-granularity traffic flow map, namely, a city road traffic flow map with high definition.
According to the embodiment of the invention, the first encoder is used for generating the traffic road network characteristic map as the priori knowledge for generating the fine-granularity traffic flow map, the priori knowledge is explicitly encoded in the decoder, the guiding function of the priori knowledge in the generation of the fine-granularity traffic flow map is fully exerted, the urban traffic flow distribution mode is fully explored, and excellent performance and accuracy can be obtained on the high-definition urban traffic flow map generation task.
Example two
Before describing the method of the present embodiment in detail, some symbolic representations related to the present embodiment are first defined.
In this embodiment, the target area (e.g., city) is uniformly divided into grid patterns of i×j according to the longitude and latitude coordinates of the geography. The data granularity of the traffic flow map is related to this region division, and a larger size mesh map (smaller actual size of each mesh) will result in a finer granularity flow map.
Representing traffic flow data of grid (i, j) within t time interval as
Is a binary real number set representing the number of vehicles flowing in/out. Therefore, the traffic flow map of the entire investigation region can be expressed as
In the present embodiment, the coarse granularity indicates the granularity of data that can be obtained on the minimum-scale monitoring sensor, and the coarse granularity traffic flow map can be obtained by dividing the target area into the i×j mesh map processing. In the same area, an NI multiplied by NJ grid diagram is adopted, and N is used as a scale factor, so that a corresponding fine-grained traffic flow diagram with high definition can be obtained. Each grid in the coarse-grained traffic flow map is composed of N x N grids in the corresponding high-definition fine-grained traffic flow map.
Let x (i,j) As coarse-grained trafficThe value, y, of a grid of the flow map (i′,j′ ) Representing that the grid corresponds to N x N small grids on a fine-grained traffic flow map of high definition, then the following constraints exist:
wherein i e [1, I ], j e [1, J ]; i 'e [1, NI ], j' e [1, NJ ], in, out are the number of vehicles in and the number of vehicles out, respectively.
For a given coarse-grained traffic flow map
And the scaling factor N.epsilon.Z+, the goal of this embodiment is to infer a fine-grained traffic flow map of high definition +.>
It should be further noted that, in this embodiment, an urban road traffic flow map generating model (RAFM) is provided, and the structure is shown in fig. 2, where the model is composed of four components including a main reasoning branch, a Road Network Branch (RNB), a converter architecture (TR E-D), and an external factor fusion module (EM). Outputting a fine-grained traffic flow map with high definition using a trained RAFM model>
The process is as follows:
in this embodiment, first, a coarse-grained traffic flow map of a target area, environmental data of the target area, and a traffic map of the target area are acquired. It should be further noted that, in this embodiment, a real world traffic map of a target area (such as a city) is first obtained, and a detailed classification of each road in the target area may be obtained according to the city traffic road label. According to the actual driving scene of the vehicle under study (such as a taxi), roads which best meet the driving conditions of the vehicle under study, such as highways, trunks, secondary roads and the like, are reserved, and roads which are unlikely to run or can not run, such as minor diameters, railways, bicycle lanes and the like, are abandoned. Because the traffic map contains a lot of information which is irrelevant to tasks and needs to be further processed, geographic information software (such as Arcmap) is used as an auxiliary tool to draw and render the shape of a road to construct a traffic road network map, wherein the road is represented by different straight lines or curves. The width and shape of the lines may vary in the generated traffic road network graph. For example, roads with higher traffic classes are typically rendered as wider lines, and large circular roads may be rendered as circles.
For the convenience of subsequent processing, the coarse-granularity traffic flow map obtained when dividing the target area is regarded as square (for example, a grid map divided into 16×16), but a certain distortion exists when the coarse-granularity traffic flow map is matched with the real map, in order to make the traffic road network map coincide with the coarse-granularity traffic flow map, the traffic road network map needs to be subjected to size adjustment to eliminate proportional distortion in terms of longitude and latitude, and in order to further process, the rendered and adjusted traffic road network map is converted into an inverted gray image of a channel, wherein the value on each pixel represents the existence degree of the road, so that the value of the pixel containing the road is not zero, and the value of the pixel not containing the road is zero.
In this embodiment, a Road Network Branch (RNB) in the RAFM model is used to generate a traffic network feature map, where the Road Network Branch (RNB) includes a first encoder and a residual network, and the process is as follows:
order the
Representing traffic network map, weighting traffic network map by coarse-grained traffic flow map X in order to eliminate the distribution difference between traffic network map and actual traffic flow, calculating average value of coarse-grained traffic flow map X on 0 th dimension, and adjusting it to be in contact with G by nearest neighbor interpolation o The same size:
calculating a weighted first traffic network graph
Wherein, the liquid crystal display device comprises a liquid crystal display device,
feature point multiplication, in flow, and out flow respectively represent the number of vehicles flowing in and the number of vehicles flowing out.
After the first traffic network graph is obtained, the road features in the first traffic network graph are fully encoded by using the first encoder as priori knowledge. The first encoder consists of a one-dimensional convolution layer and two one-dimensional residual error module layers, and takes a first traffic road network graph G as input to construct a corresponding traffic road network characteristic graph F G 。
In most cases, convolution architectures use a square filter kernel (e.g., 3 x 3) that has a square receptive field that fits into a natural object with a sharp boundary and shape. However, since urban roads are very different from natural objects, most roads are thin and long, and if the conventional square kernel is continuously used, a long road is covered with a large-sized filter kernel, which results in many pixels not related to the road being extracted. The one-dimensional filter is more in line with the shape of the road, and the road feature extraction can be more effectively realized than the traditional convolution.
As shown in FIG. 3, the one-dimensional convolution layer is formed by connecting four groups of one-dimensional filters in four different directions, so that
One-dimensional convolution filter of size 2r+1, input +.>
The one-dimensional filter is a filter with four different direction identification vectors i= (I) h ,I w ) Kappa output of +.>
The formula is as follows:
the direction identification vectors are respectively (0, 1), (1, 0), (1, 1) and (1, -1) and are used for horizontal convolution, vertical convolution, dead angle convolution and anti-diagonal convolution. When r=4, there are 9 parameters per one-dimensional filter, which is the same as a 3×3 convolution filter. Therefore, the method can replace 3X 3 convolution in the feature extraction process, the parameter number of each group of one-dimensional filters is 1/4 of the parameter number of the 3X 3 filters, and the parameter number and the calculation cost are kept the same. After generating the horizontal feature, the vertical feature, the positive angle feature and the negative angle feature, adding the four features and inputting the four features into a one-dimensional residual error module layer, wherein the one-dimensional residual error module layer comprises two convolution layers, each convolution layer is respectively followed by batch normalization, nonlinearity is introduced between the two convolution layers by using a ReLU function, and a traffic network feature map F is obtained by obtaining the output of the one-dimensional residual error module G 。
After the environmental data of the target area is acquired, the RAFM model uses an external factor fusion module (EM) to convert the external factor fusion module (EM) into an environmental factor embedded vector, and the method specifically comprises the following steps:
converting the classifiable factors such as weather, day of week and time into low-dimensional vectors, respectively sending into different embedded layers, and combining into a vector
Combining non-sortable factors into a vector +.>
And will->
And +.>
Connecting to obtain an environment factor embedded vector: />
Wherein the external factor fusion module (EM) consists of two dense layers followed by dropout and ReLU functions. Inputting the e into an external factor fusion module to obtain a characteristic diagram with the same size as the coarse-granularity traffic flow diagram
Then, the RAFM model inputs the traffic network feature map, the environment factor embedded vector and the coarse-granularity traffic flow map into a main reasoning branch to generate a first traffic flow feature map, wherein the method comprises the following steps of:
upsampling the coarse-grain traffic flow map X by using bilinear interpolation and a scaling factor N to obtain a resized coarse-grain traffic flow map
Then X is taken up up And F e And inputting the cascade into a main reasoning branch. In the main reasoning branch, a convolution layer is firstly used for extracting low-level characteristics, and then 16 residual modules with the same structure are used for constructing a first traffic flow characteristic diagram F. The method comprises the following steps: carrying out feature cascading on the external factor embedded vector and the coarse-granularity traffic flow map to obtain a first cascade feature, and fusing the first cascade feature to obtain a traffic flow primary feature map; and carrying out feature cascading on the road network feature map and the traffic flow primary feature map to obtain a second cascading feature, and fusing the second cascading feature to obtain a first traffic flow feature map F.
After obtaining the first traffic flow characteristic map F, the first traffic flow characteristic map F and the traffic road network characteristic map F are obtained G In the input converter architecture (TR E-D), it is further explained that the converter architecture (TR E-D) is a conversion-based oneEncoder-decoder architecture of a decoder, comprising a converter encoder (TRE) and a converter decoder (TRD). The converter architecture (TR E-D) infers a fine-grained traffic flow map of high definition from a priori knowledge and models the relationship between them. The converter architecture enables global reasoning of features in pairs with the entire fine-grained traffic flow graph as context by using the self-attention mechanism and the codec attention mechanism.
First traffic flow characteristic map
It is necessary to first change to a right resolution size by averaging the pooling layer +.>
Traffic network feature map F G The same transformation is needed to be performed through the maximum pooling layer to obtain F' G 。
Each encoder layer in the TRE is set to a standard structure consisting of a multi-headed self-attention module, according to the definition of the converter structure. First, a 1×1 convolution reduces the channel dimension of F' from C to d, generating a new feature map
Constitutes the input of the TRE. Since the TRE requires a sequence as input, F will be t Is folded into the horizontal dimension and input flattened, thereby obtaining a feature map with the shape of d×hw. Since the architecture of the TRE is arranged unchanged, which means that the original two-dimensional input information is lost, spatial position coding is added to each multi-headed self-attention layer of the TRE in order to preserve the relative position information of each pixel. And the TRE encodes the F' to obtain a second traffic flow characteristic diagram, and the second traffic flow characteristic diagram is transmitted to the TRD.
The TRD also follows the standard architecture of the converter, but in this embodiment the TRD adopts a parallel structure to decode the features in parallel at each decoder layer, while the original transformer adopts an autoregressive model, delivering spatial position encoding and output encoding only once at the input, thus predicting the output sequence one element at a time.
The TRD receives the transmitted second traffic flow characteristic map of the TRE to communicate the road network characteristic map F' G As embedded coding and added to the input of the attention layer of each decoder for perceptual learning, the coding result of the TRE is passed into the TRD as context. And (3) converting the third traffic flow characteristic diagram after TRD conversion, namely, the traffic flow characteristic diagram under guidance of the urban road network.
And finally, inputting the first traffic flow characteristic map and the third traffic flow characteristic map into a convolution layer by the RAFM model to perform characteristic fusion, and generating a fine-granularity traffic flow map, namely a high-definition urban road traffic flow map.
It should be further noted that, in this embodiment, the RAFM model is implemented based on Python and PyTorch deep learning frameworks, the training of the RAFM model adopts a random gradient descent optimizer, the momentum is 0.9, the initial learning rate is 3e-4, the filter weights of all layers are initialized by Xavier, and the RAFM model is optimized end-to-end by minimizing the average absolute error (MAPE) between the reasoning result and the corresponding true value.
Example III
As shown in fig. 5, the embodiment of the present invention further provides an urban road traffic flow map generating system, where the system is configured to perform the above-mentioned urban road traffic flow map generating method, and the urban road traffic flow map generating system includes a data acquisition module 201, a traffic road network feature map generating module 202, a first traffic flow feature map generating module 203, a second traffic flow feature map generating module 204, a third traffic flow feature map generating module 205, and a fine-grained traffic flow map generating module 206;
the data acquisition module 201 is configured to acquire a coarse-granularity traffic flow map of a target area, environment data of the target area, and a traffic map of the target area;
the traffic road network feature map generating module 202 is configured to input the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding, and generate a traffic road network feature map;
the first traffic flow characteristic map generating module 203 is configured to generate a first traffic flow characteristic map based on the traffic road network characteristic map, the environmental data, and a coarse-granularity traffic flow map;
the second traffic flow characteristic map generating module 204 is configured to input the first traffic flow characteristic map to a second encoder for encoding, and generate a second traffic flow characteristic map;
the third traffic flow characteristic map generating module 205 is configured to input the second traffic flow characteristic map and the traffic road network characteristic map to a decoder for decoding, and generate a third traffic flow characteristic map;
the fine-granularity traffic flow map generation module 206 is configured to generate a fine-granularity traffic flow map based on the first traffic flow feature map and the third traffic flow feature map.
As shown in fig. 6, the present embodiment also provides an urban road traffic flow map generating apparatus 30, which includes a processor 300 and a memory 301;
the memory 301 is used for storing a program code 302 and transmitting the program code 302 to the processor;
the processor 300 is configured to perform the steps of one embodiment of the urban road traffic flow map generation method described above according to the instructions in the program code 302.
Illustratively, the computer program 302 may be partitioned into one or more modules/units that are stored in the memory 301 and executed by the processor 300 to complete the present application. The one or more modules/units may be a series of computer program instruction segments capable of performing the specified functions, which instruction segments are used for describing the execution of the computer program 302 in the terminal device 30.
The terminal device 30 may be a computing device such as a desktop computer, a notebook computer, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor 300, a memory 301. It will be appreciated by those skilled in the art that fig. 6 is merely an example of the terminal device 30 and is not meant to be limiting as to the terminal device 30, and may include more or fewer components than shown, or may combine certain components, or different components, e.g., the terminal device may also include input and output devices, network access devices, buses, etc.
The processor 300 may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 301 may be an internal storage unit of the terminal device 30, such as a hard disk or a memory of the terminal device 30. The memory 301 may also be an external storage device of the terminal device 30, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the terminal device 30. Further, the memory 301 may also include both an internal storage unit and an external storage device of the terminal device 30. The memory 301 is used for storing the computer program and other programs and data required by the terminal device. The memory 301 may also be used to temporarily store data that has been output or is to be output.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present invention, and are not to be construed as limiting the scope of the invention. It should be noted that any modifications, equivalent substitutions, improvements, etc. made by those skilled in the art without departing from the spirit and principles of the present invention are intended to be included in the scope of the present invention.
Claims (8)
1. The urban road traffic flow map generating method is characterized by comprising the following steps:
acquiring a coarse-grained traffic flow map of a target area, environment data of the target area and a traffic map of the target area;
inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map;
the specific process of inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map is as follows:
generating a traffic road network map based on the traffic map;
weighting the traffic road network graph and the coarse-granularity traffic flow graph to obtain a weighted first traffic road network graph, and inputting the first traffic road network graph into a first encoder so that the first encoder encodes road features in the first traffic road network graph to obtain a traffic road network feature graph;
the first encoder encodes the road characteristics in the first traffic network map, and the specific process of obtaining the traffic network characteristic map is as follows:
the first encoder carries out horizontal convolution, vertical convolution and opposite direction on the first traffic network diagram respectively
Generating horizontal features, vertical features, positive angle features and negative angle features by angle convolution and negative angle convolution, and obtaining a traffic network feature map based on the horizontal features, the vertical features, the positive angle features and the negative angle features;
generating a first traffic flow characteristic map based on the traffic road network characteristic map, the environmental data and the coarse-granularity traffic flow map;
inputting the first traffic flow characteristic map into a second encoder for encoding to generate a second traffic flow characteristic map;
inputting the second traffic flow characteristic diagram and the traffic road network characteristic diagram into a decoder for decoding to generate a third traffic flow characteristic diagram;
and generating a fine-granularity traffic flow map based on the first traffic flow characteristic map and the third traffic flow characteristic map.
2. The urban road traffic flow map generating method according to claim 1, wherein the specific process of generating the first traffic flow map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map is as follows:
converting the environmental data into an environmental factor embedding vector;
extracting features of the traffic road network feature map to generate a road network feature map;
and carrying out feature fusion on the road network feature map, the external factor embedded vector and the coarse-granularity traffic flow map to generate a first traffic flow feature map.
3. The method for generating an urban road traffic flow map according to claim 2, characterized in that,
the road network feature map, the external factor embedded vector and the coarse-granularity traffic flow map are subjected to feature fusion, and the specific process for generating the first traffic flow feature map comprises the following steps of:
feature cascading is carried out on the external factor embedded vector and the coarse-granularity traffic flow map to obtain a first cascade feature, and the first cascade feature is fused to obtain a traffic flow primary feature map;
and carrying out feature cascading on the road network feature map and the traffic flow primary feature map to obtain a second cascading feature, and fusing the second cascading feature to obtain the first traffic flow feature map.
4. The urban road traffic flow map generating method according to claim 1, wherein the specific process of inputting the first traffic flow map into a second encoder for encoding, and generating the second traffic flow map is as follows:
and inputting the first traffic flow characteristic map into a second encoder so that the second encoder encodes the first traffic flow characteristic map based on a multi-head attention mechanism to generate a second traffic flow characteristic map.
5. The urban road traffic flow map generating method according to claim 1, wherein the specific process of inputting the second traffic flow map and the traffic network map into a decoder for decoding, and generating a third traffic flow map is as follows:
and taking the traffic road network feature map as an embedded code, and inputting the embedded code and the second traffic flow feature map into a decoder so that the decoder decodes based on a multi-head attention mechanism to obtain a third traffic flow feature map.
6. The urban road traffic flow map generating method according to claim 1, wherein the specific process of generating the fine-grained traffic flow map based on the first traffic flow feature map and the third traffic flow feature map is as follows:
and carrying out feature fusion on the first traffic flow feature map and the third traffic flow feature map to generate a fine-granularity traffic flow map.
7. An urban road traffic flow map generating system, characterized in that the system is configured to perform the urban road traffic flow map generating method according to any one of claims 1 to 6, and comprises a data acquisition module, a traffic road network feature map generating module, a first traffic flow feature map generating module, a second traffic flow feature map generating module, a third traffic flow feature map generating module and a fine-grained traffic flow map generating module;
the data acquisition module is used for acquiring a coarse-grained traffic flow map of the target area, environment data of the target area and a traffic map of the target area;
the traffic road network feature map generating module is used for inputting the coarse-granularity traffic flow map and the traffic map into a first encoder for encoding to generate a traffic road network feature map;
the first traffic flow characteristic map generating module is used for generating a first traffic flow characteristic map based on the traffic road network characteristic map, the environment data and the coarse-granularity traffic flow map;
the second traffic flow characteristic map generating module is used for inputting the first traffic flow characteristic map into a second encoder for encoding to generate a second traffic flow characteristic map;
the third traffic flow characteristic map generating module is used for inputting the second traffic flow characteristic map and the traffic road network characteristic map into a decoder for decoding to generate a third traffic flow characteristic map;
the fine-granularity traffic flow map generation module is used for generating a fine-granularity traffic flow map based on the first traffic flow characteristic map and the third traffic flow characteristic map.
8. An urban road traffic flow map generating device is characterized by comprising a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the urban road traffic flow map generation method according to any one of claims 1 to 6 according to the instructions in the program code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110273734.7A CN113094422B (en) | 2021-03-12 | 2021-03-12 | Urban road traffic flow map generation method, system and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110273734.7A CN113094422B (en) | 2021-03-12 | 2021-03-12 | Urban road traffic flow map generation method, system and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113094422A CN113094422A (en) | 2021-07-09 |
| CN113094422B true CN113094422B (en) | 2023-07-07 |
Family
ID=76667160
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110273734.7A Active CN113094422B (en) | 2021-03-12 | 2021-03-12 | Urban road traffic flow map generation method, system and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113094422B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116206443B (en) * | 2023-02-03 | 2023-12-15 | 重庆邮电大学 | Traffic flow data interpolation method based on time-space road network pixelized representation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107170236A (en) * | 2017-06-14 | 2017-09-15 | 中山大学 | A kind of important intersection extracting method of road network based on floating car data |
| CN107958246A (en) * | 2018-01-17 | 2018-04-24 | 深圳市唯特视科技有限公司 | A kind of image alignment method based on new end-to-end human face super-resolution network |
| CN112288701A (en) * | 2020-10-23 | 2021-01-29 | 西安科锐盛创新科技有限公司 | Intelligent traffic image detection method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6542808B2 (en) * | 1999-03-08 | 2003-04-01 | Josef Mintz | Method and system for mapping traffic congestion |
-
2021
- 2021-03-12 CN CN202110273734.7A patent/CN113094422B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107170236A (en) * | 2017-06-14 | 2017-09-15 | 中山大学 | A kind of important intersection extracting method of road network based on floating car data |
| CN107958246A (en) * | 2018-01-17 | 2018-04-24 | 深圳市唯特视科技有限公司 | A kind of image alignment method based on new end-to-end human face super-resolution network |
| CN112288701A (en) * | 2020-10-23 | 2021-01-29 | 西安科锐盛创新科技有限公司 | Intelligent traffic image detection method |
Non-Patent Citations (1)
| Title |
|---|
| Deep Contrast Learning for Salient Object Detection;Guanbin Li etc;Computer Vision Founation;478-487 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113094422A (en) | 2021-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101836304B1 (en) | Method and device for detecting vehicle contour based on point cloud data | |
| CN112149547B (en) | Remote sensing image water body identification method based on image pyramid guidance and pixel pair matching | |
| CN111191654B (en) | Road data generation method and device, electronic equipment and storage medium | |
| CN110619605B (en) | Parallel meteorological element downscaling method and device, electronic equipment and storage medium | |
| Wu et al. | DeepDualMapper: A gated fusion network for automatic map extraction using aerial images and trajectories | |
| Varghese et al. | Deep learning in transport studies: A meta-analysis on the prediction accuracy | |
| CN115424059B (en) | Remote sensing land utilization classification method based on pixel level contrast learning | |
| CN112712138A (en) | Image processing method, device, equipment and storage medium | |
| CN112580453A (en) | Land use classification method and system based on remote sensing image and deep learning | |
| CN113094422B (en) | Urban road traffic flow map generation method, system and equipment | |
| US11651191B2 (en) | Methods, apparatuses, and computer program products using a repeated convolution-based attention module for improved neural network implementations | |
| CN115540893A (en) | Vehicle path planning method and device, electronic equipment and computer readable medium | |
| CN115965970A (en) | Method and system for realizing bird's-eye view semantic segmentation based on implicit set prediction | |
| Mukhopadhyay et al. | A hybrid lane detection model for wild road conditions | |
| Rahman et al. | Attention based deep hybrid networks for traffic flow prediction using google maps data | |
| CN117079148A (en) | Urban functional area identification method, device, equipment and medium | |
| CN113962460B (en) | Urban fine granularity flow prediction method and system based on space-time comparison self-supervision | |
| Feng et al. | Lightweight detection network for arbitrary-oriented vehicles in UAV imagery via precise positional information encoding and bidirectional feature fusion | |
| CN115773744A (en) | Model training and road network processing method, device, equipment, medium and product | |
| CN114170519A (en) | High-resolution remote sensing road extraction method based on deep learning and multidimensional attention | |
| Yuan et al. | SCTransNet: Spatial-channel Cross Transformer Network for Infrared Small Target Detection | |
| Tang et al. | Sparseocc: Rethinking sparse latent representation for vision-based semantic occupancy prediction | |
| Du et al. | IMG2HEIGHT: height estimation from single remote sensing image using a deep convolutional encoder-decoder network | |
| Yao et al. | TL-Detector: Lightweight Based Real-Time Traffic Light Detection Model for Intelligent Vehicles | |
| Zhao et al. | Effective Recognition of Word-Wheel Water Meter Readings for Smart Urban Infrastructure |
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 |