CN113094422A - Urban road traffic flow chart generation method, system and equipment - Google Patents

Abstract

The invention discloses a method, system and equipment for generating a traffic flow map of an urban road. The present invention generates a third traffic flow feature map by inputting the traffic road 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, and based on the first traffic flow feature map and a third traffic flow feature map to generate a fine-grained traffic flow map. By using the traffic road network feature map generated by the first encoder as the prior knowledge for generating the fine-grained traffic flow map, and explicitly encoding the prior knowledge in the decoder, the present invention gives full play to the prior knowledge in generating the fine-grained traffic flow. The guiding role of the map fully explores the urban traffic flow distribution pattern, and can achieve superior performance and accuracy in the task of generating high-definition urban traffic flow maps.

Description

Method, system and device for generating urban road traffic flow map

technical field

The invention relates to the field of traffic, and in particular to a method, system and device for generating a traffic flow map of an urban road.

Background technique

Urban traffic flow monitoring is an important part in the construction of smart cities, which can provide an important information basis for early warning of public safety, urban planning and other tasks. In the modern information society, vehicles and infrastructure (such as traffic monitoring, air quality monitoring stations) are constantly generating massive amounts of urban data in heterogeneous formats, such as GPS track points, weather data, etc.

Deep learning and data mining make these data significant for exploring urban dynamic patterns and predicting performance. But generally, fine-grained actual city data is difficult to collect. For example, the monitoring of urban traffic flow requires the deployment of sufficient sensor devices throughout the city. Thousands of probes, loop sensors, surveillance cameras and other monitoring devices must be deployed in traffic sections to monitor traffic flow in real time and collect traffic data. The real-time road traffic monitoring system is highly dependent on these monitoring devices, and long-term stable operation requires a lot of power resources and reliable disaster recovery capabilities, which will bring huge expenses and high maintenance costs.

Today, the rapid development of smart cities around the world, and the corresponding exponential increase in the cost of monitoring traffic flow, may be a key factor limiting further development. To solve this problem and facilitate the development of smart city applications, a new approach is needed to reduce monitoring costs while ensuring the granularity and accuracy of traffic flow data. Therefore, how to infer the original fine-grained flow distribution from the easily obtained coarse-grained data becomes the focus of the problem.

In recent years, deep neural networks have been widely used for refined reasoning on urban traffic data. Existing studies mainly model this problem as a mapping problem of mapping data with low information entropy to data with high information entropy. These works typically divide the study area into grid maps based on longitude and latitude coordinates, and organize the collected traffic data into tensors to facilitate automatic representation learning that feeds them directly into convolutional networks for mapping. Some strategies also propose to establish spatial constraints between coarse-grained flow data and fine-grained flow data to facilitate learning representations of spatial correlations.

However, existing methods all ignore the importance of prior knowledge for this problem, such as the urban road network, which hardly changes significantly over a long period of time. Roads are an important part of urban transportation systems, and urban road networks are originally designed for vehicles, which means that most traffic flows will be generated on roads. For example, vehicles such as buses usually only travel on specific roads and do not appear in residential areas. The prior art does not fully consider the influence of prior knowledge in the process of generating a high-fine-grained urban traffic flow map, resulting in relatively poor accuracy of the generated high-fine-grained urban traffic flow map.

To sum up, the prior art does not fully consider the influence of prior knowledge in the process of generating a high-fine-grained urban traffic flow map, resulting in a technical problem of poor accuracy of the generated high-fine-grained urban traffic flow map. .

SUMMARY OF THE INVENTION

The present invention provides a method, system and device for generating an urban road traffic flow map. The present invention introduces a traffic road network feature map as a priori knowledge for generating a fine-grained traffic flow map, thereby improving the accuracy of the fine-grained traffic flow map.

In order to solve the above technical problems, an embodiment of the present invention provides a method for generating an urban road traffic flow map, including the following steps:

Obtain the coarse-grained traffic flow map of the target area, the environmental data of the target area, and the traffic map of the target area;

Inputting the coarse-grained 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 feature map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map;

Inputting the first traffic flow feature map into a second encoder for encoding, and generating a second traffic flow feature map;

Inputting the second traffic flow feature map and the traffic road network feature map into a decoder for decoding to generate a third traffic flow feature map;

Based on the first traffic flow feature map and the third traffic flow feature map, a fine-grained traffic flow map is generated.

Preferably, the coarse-grained traffic flow map and the traffic map are input into the first encoder for encoding, and the specific process of generating the traffic road network feature map is as follows:

based on the traffic map, generating a traffic road network map;

The traffic road network map and the coarse-grained traffic flow map are weighted to obtain a weighted first traffic road network map, and the first traffic road network map is input into the first encoder, so that the The first encoder encodes the road features in the first traffic road network map to obtain a traffic road network feature map.

Preferably, the first encoder performs feature extraction on the first traffic road network map, and the specific process of obtaining the traffic road network feature map is:

The first encoder performs horizontal convolution, vertical convolution, positive diagonal convolution and anti-diagonal convolution on the first traffic network map, respectively, to generate horizontal features, vertical features, positive diagonal features, and anti-diagonal features. feature, and a traffic road network feature map is obtained based on the horizontal feature, the vertical feature, the positive diagonal feature, and the opposite diagonal feature.

Preferably, based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map, the specific process of generating the first traffic flow feature map is as follows:

converting the environmental data into an environmental factor embedding vector;

performing feature extraction on the traffic road network feature map to generate a road network feature map;

Feature fusion is performed on the road network feature map, the external factor embedding vector and the coarse-grained traffic flow map to generate a first traffic flow feature map.

Preferably, the feature fusion of the road network feature map, the external factor embedding vector and the coarse-grained traffic flow map is performed, and the specific process of generating the first traffic flow feature map is as follows:

Perform feature cascading on the external factor embedding vector and the coarse-grained traffic flow map to obtain a first cascade feature, and fuse the first cascade feature to obtain a primary traffic flow feature map;

Feature cascading is performed on the road network feature map and the primary traffic flow feature map to obtain a second cascade feature, and the second cascade feature is fused to obtain the first traffic flow feature map.

Preferably, the first traffic flow feature map is input into the second encoder for encoding, and the specific process of generating the second traffic flow feature map is as follows:

The first traffic flow feature map is input into the second encoder, so that the second encoder encodes the first traffic flow feature map based on a multi-head attention mechanism to generate a second traffic flow feature map.

Preferably, the second traffic flow feature map and the traffic road network feature map are input into the decoder for decoding, and the specific process of generating the third traffic flow feature map is as follows:

The traffic road network feature map is used as an embedded code, and the embedded code and the second traffic flow feature map are input into the decoder, so that the decoder decodes based on the multi-head attention mechanism to obtain a third traffic flow. Traffic characteristic map.

Preferably, based on the first traffic flow feature map and the third traffic flow feature map, the specific process for generating a fine-grained traffic flow map is:

Feature fusion is performed on the first traffic flow feature map and the third traffic flow feature map to generate a fine-grained traffic flow map.

An embodiment of the present invention further provides a system for generating an urban road traffic flow map, including 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, and a third traffic flow feature map generating module. Traffic flow feature map generation module and fine-grained traffic flow map generation module;

The data acquisition module is used to acquire the coarse-grained traffic flow map of the target area, the environmental data of the target area and the traffic map of the target area;

The traffic road network feature map generation module is configured to input the coarse-grained traffic flow map and the traffic map into the first encoder for encoding, and generate a traffic road network feature map;

The first traffic flow feature map generating module is configured to generate a first traffic flow feature map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map;

The second traffic flow feature map generation module is configured to input the first traffic flow feature map into a second encoder for encoding, and generate a second traffic flow feature map;

The third traffic flow feature map generation module is configured to input the second traffic flow feature map and the traffic road network feature map into a decoder for decoding, and generate a third traffic flow feature map;

The fine-grained traffic flow map generating module is configured to generate a fine-grained traffic flow map based on the first traffic flow feature map and the third traffic flow feature map.

The embodiment of the present invention also provides a device for generating a city road traffic flow map, including a processor and a memory;

the memory is used to store program code and transmit the program code to the processor;

The processor is configured to execute the above-mentioned method for generating an urban road traffic flow map according to the instructions in the program code.

Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

In the embodiment of the present invention, the traffic road 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 to generate a third traffic flow feature map, and based on the first traffic flow feature map The flow feature map and the third traffic flow feature map generate a fine-grained traffic flow map. In the embodiment of the present invention, by using the traffic road network feature map generated by the first encoder as the prior knowledge for generating the fine-grained traffic flow map, and explicitly encoding the prior knowledge in the decoder, the prior knowledge is fully utilized in generating fine-grained traffic flow maps. The guiding role of the traffic flow map fully explores the urban traffic flow distribution pattern, and can achieve superior performance and accuracy in the task of generating high-definition urban traffic flow maps.

Description of drawings

FIG. 1 is a flowchart of a method for generating an urban road traffic flow map provided by an embodiment of the present invention.

FIG. 2 is a schematic structural diagram of an urban road traffic flow map generation model (RAFM) provided by an embodiment of the present invention.

FIG. 3 is a schematic structural diagram of a one-dimensional convolution layer in an urban road traffic flow map generation model provided by an embodiment of the present invention.

FIG. 4 is a schematic structural diagram of a one-dimensional residual module layer in an urban road traffic flow map generation model provided by an embodiment of the present invention.

FIG. 5 is a schematic structural diagram of a system for generating an urban road traffic flow map according to an embodiment of the present invention.

FIG. 6 is a schematic structural diagram of a device for generating an urban road traffic flow map according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Example 1

Please refer to FIG. 1 , which is a flowchart of a method for generating an urban road traffic flow map provided by an embodiment of the present invention, including the following steps:

S101: Obtain a coarse-grained traffic flow map of the target area, environmental data of the target area, and a traffic map of the target area.

Among them, it needs to be further explained that since environmental data such as weather, wind speed, temperature, time, etc. will have a complex and important impact on the distribution of urban traffic flow, in this embodiment, it is necessary to consider the environmental data of the target area in the fine-grained traffic flow. Influence during flow map generation.

S102: Input the coarse-grained traffic flow map and the traffic map into the first encoder for encoding, generate a traffic road network feature map, and use the traffic road network feature map as prior knowledge for generating a fine-grained traffic flow map.

S103: Based on the traffic road network feature map, the environmental data, and the coarse-grained traffic flow map, generate a first traffic flow feature map.

It should be further explained that the features of the traffic road network feature map, the environmental data, and the features of the coarse-grained traffic flow map are fused to generate the first traffic flow feature map.

S104: Input the first traffic flow feature map into the second encoder for encoding to generate a second traffic flow feature map.

S105: Input the second traffic flow feature map and the traffic road network feature map into the decoder for decoding, and explicitly encode the prior knowledge, thereby generating a third traffic flow feature map.

S106: Perform feature fusion on the first traffic flow feature map and the third traffic flow feature map to generate a fine-grained traffic flow map, that is, a high-definition urban road traffic flow map.

In the embodiment of the present invention, by using the traffic road network feature map generated by the first encoder as the prior knowledge for generating the fine-grained traffic flow map, and explicitly encoding the prior knowledge in the decoder, the prior knowledge is fully utilized in generating fine-grained traffic flow maps. The guiding role of the traffic flow map fully explores the urban traffic flow distribution pattern, and can achieve superior performance and accuracy in the task of generating high-definition urban traffic flow maps.

Embodiment 2

Before introducing the method of this embodiment in detail, some related symbols in this embodiment are first defined.

In this embodiment, the target area (such as a city) is evenly divided into an I×J grid map according to the geographic latitude and longitude coordinates. The data granularity of the traffic flow map is related to this area division, and a larger size grid map (the actual size of each grid is smaller) will result in a more fine-grained flow map.

为一个二元实数集合，代表流入/流出的车辆数。所以，在某个时间间隔内，整个研究区域的交通流量图可以表示为Denote the traffic flow data for grid (i,j) in time interval t as

is a set of binary real numbers representing the number of vehicles flowing in/out. So, in a certain time interval, the traffic flow map of the whole study area can be expressed as

In this embodiment, the coarse-grained represents the data granularity that can be obtained on the smallest-scale monitoring sensor, and the coarse-grained traffic flow map can be obtained by dividing the target area into an I×J grid map. In the same area, using the NI×NJ grid map and taking N as the scale factor, the corresponding fine-grained traffic flow map with high definition can be obtained. Each grid in the coarse-grained traffic flow map consists of N×N grids in the corresponding high-resolution fine-grained traffic flow map.

Let x ^(i,j) be the value of a grid of the coarse-grained traffic flow map, and y ^(i′,j′ ) represent the N×N small grids corresponding to the grid on the high-definition fine-grained traffic flow map. lattice, there are the following constraints:

where i∈[1,I],j∈[1,J]; i′∈[1,NI],j′∈[1,NJ],in,out are the number of incoming and outgoing vehicles, respectively.

和比例因子N∈Z+，本实施例的目标则是推理出高清晰度的细粒度交通流量图

其中，需要进一步说明的是，在本实施例中提供了一种城市道路交通流量图生成模型(RAFM)，结构如图2所示，模型由四个组件组成，包括主推理分支、路网分支(RNB)、转换器架构(TR E-D)和外部因素融合模块(EM)。使用训练好的RAFM模型来输出高清晰度的细粒度交通流量图

过程如下：For a given coarse-grained traffic flow map

and the scale factor N∈Z+, the goal of this embodiment is to infer a high-definition fine-grained traffic flow map

Among them, it should be further explained that an urban road traffic flow map generation model (RAFM) is provided in this embodiment. The structure is shown in Figure 2. The model consists of four components, including a main reasoning branch and a road network branch. (RNB), converter architecture (TR ED) and external factor fusion module (EM). Use the trained RAFM model to output high-definition fine-grained traffic flow maps

The process is as follows:

In this embodiment, a coarse-grained traffic flow map of the target area, environmental data of the target area, and a traffic map of the target area are first acquired. It should be further explained that in this embodiment, a real-world traffic map of the target area (eg, a city) is obtained first, and a detailed classification of each road in the target area can be obtained according to the urban traffic road labels. According to the actual driving scenarios of the studied vehicles (such as taxis), the roads that best meet the driving conditions of the research vehicles, such as highways, arterial roads, secondary roads, etc., are reserved, and the roads that are not frequently traveled or impossible to travel, such as trails, are discarded , railways, bicycle paths, etc. Since the traffic map contains a lot of information that is irrelevant to the task 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 the road to construct the traffic road network map, in which the road uses different straight lines or curves. To represent. The width and shape of the lines may vary in the resulting traffic network diagram. For example, roads with high traffic levels are often rendered as wider lines, and large roundabouts may be rendered as circles.

In order to facilitate subsequent processing, the coarse-grained traffic flow map obtained when dividing the target area is regarded as a square (for example, divided into a 16*16 grid map). Distortion, in order to make the traffic road network map match the coarse-grained traffic flow map, the traffic road network map needs to be resized to eliminate the scale distortion in latitude and longitude. For further processing, the rendered and adjusted traffic road network map is converted into A one-channel inverted grayscale image, where the numerical value on each pixel represents the presence of a road, so pixels that contain roads have a non-zero value, and pixels that don't have a value of zero.

In this embodiment, the road network branch (RNB) in the RAFM model is used to generate a traffic road network feature map, wherein the road network branch (RNB) includes a first encoder and a residual network, and the process is as follows:

表示交通路网图，为了消除交通路网图与实际交通流的分布差异，使用粗粒度交通流量图X对交通路网图进行加权，首先计算粗粒度交通流量图X在第0维上的平均值，然后用最近邻插值法将其调整为与G_o相同的大小:make

Represents the traffic road network map. In order to eliminate the distribution difference between the traffic road network map and the actual traffic flow, the coarse-grained traffic flow map X is used to weight the traffic road network map. First, the average of the coarse-grained traffic flow map X on the 0th dimension is calculated. value, and then resize it to the same size as G _o using nearest neighbor interpolation:

Calculate the weighted first traffic road network map

表示特征点乘，inflow,outflow分别表示流入的车辆数量，流出的车辆数量。in,

Represents feature point multiplication, inflow, outflow represent the number of inflowing vehicles and outflowing vehicles respectively.

After the first traffic road network map is obtained, the first encoder is used to fully encode the road features in the first traffic road network map as prior knowledge. The first encoder consists of a one-dimensional convolutional layer and two one-dimensional residual module layers, and takes the first traffic road network graph G as an input to construct a corresponding traffic road network feature map F _G .

In most cases, convolutional architectures use square filter kernels (e.g. 3×3), which have square receptive fields and are suitable for natural objects with sharp boundaries and shapes. However, since urban roads are very different from natural objects, most of the roads are thin and long, if we continue to use the traditional square kernel, then a large-sized filter kernel is needed to cover the long-length roads, which will lead to many Pixels not related to the road are extracted. The one-dimensional filter is more in line with the shape of the road, and can achieve road feature extraction more effectively than the traditional convolution.

表示尺寸为2r+1的一维卷积滤波器，输入为

一维滤波器为有四个不同的方向标识向量I＝(I_h,I_w)的κ，输出为

公式如下：As shown in Figure 3, the one-dimensional convolutional layer is composed of four groups of one-dimensional filter connections in four different directions, let

represents a 1D convolutional filter of size 2r+1, the input is

The one-dimensional filter is κ with four different direction identification vectors I=(I _h , I _w ), and the output is

The formula is as follows:

Among them, the direction identification vectors are respectively (0,1), (1,0), (1,1), (1,-1) for horizontal convolution, vertical convolution, positive diagonal convolution and anti diagonal convolution product. When r=4, each 1D filter has 9 parameters, the same as the 3x3 convolution filter. Therefore, the 3×3 convolution can be replaced in the feature extraction process, and the number of parameters of each set of one-dimensional filters is 1/4 of the number of parameters of the 3×3 filters, keeping the same number of parameters and computational overhead. After generating horizontal features, vertical features, positive diagonal features and anti-diagonal features, the four features are added and input into the one-dimensional residual module layer. As shown in Figure 4, the one-dimensional residual module layer contains two Each convolutional layer is followed by batch normalization. A ReLU function is used between the two convolutional layers to introduce nonlinearity, and the output of the one-dimensional residual module is obtained to obtain the traffic road network feature map F _G .

After obtaining the environmental data of the target area, the RAFM model uses the external factor fusion module (EM) to convert the external factor fusion module (EM) into the environmental factor embedding vector, as follows:

将不可分类因素被合并为向量

并将

以及

进行连接，得到环境因素嵌入向量：

Convert categorical factors such as weather, day of the week, and time into low-dimensional vectors, respectively, and feed them into different embedding layers and combine them into one vector

Combine unclassifiable factors into a vector

and will

as well as

Connect to get the environmental factor embedding vector:

Among them, the external factor fusion module (EM) consists of two dense layers followed by dropout and ReLU functions. After inputting e into the external factor fusion module, a feature map of the same size as the coarse-grained traffic flow map is obtained

After that, the RAFM model inputs the traffic road network feature map, the environmental factor embedding vector, and the coarse-grained traffic flow map into the main reasoning branch to generate the first traffic flow feature map, as follows:

然后将X_up和F_e级联后输入主推理分支中。主推理分支中首先使用一个卷积层提取低级特征，然后使用了16个具有相同结构的残差模块构造第一交通流量特征图F。具体为：将外部因素嵌入向量以及粗粒度交通流量图进行特征级联，得到第一级联特征，对第一级联特征进行融合，得到交通流量初级特征图；将路网特征图以及交通流量初级特征图进行特征级联，得到第二级联特征，对第二级联特征进行融合，得到第一交通流量特征图F。Upsampling the coarse-grained traffic flow map X using bilinear interpolation and a scale factor N to obtain a resized coarse-grained traffic flow map

Then X _up and Fe are _cascaded and fed into the main inference branch. In the main inference branch, a convolutional layer is first used to extract low-level features, and then 16 residual modules with the same structure are used to construct the first traffic flow feature map F. Specifically, the features are cascaded with the external factor embedding vector and the coarse-grained traffic flow map to obtain the first cascaded features, and the first cascaded features are fused to obtain the primary traffic flow feature map; the road network feature map and the traffic flow The primary feature map performs feature cascade to obtain the second cascade feature, and the second cascade feature is fused to obtain the first traffic flow feature map F.

After the first traffic flow feature map F is obtained, the first traffic flow feature map F and the traffic road network feature map F _G are input into the converter architecture (TR ED), wherein, it should be further explained that the converter architecture (TR ED) ED) is a transformer-based encoder-decoder architecture, including a transformer encoder (TRE) and a transformer decoder (TRD). Transducer Architecture (TR ED) infers high-resolution, fine-grained traffic flow maps based on prior knowledge and models the relationships between them. By using self-attention mechanism and encoder-decoder attention mechanism, this converter architecture is able to use the entire fine-grained traffic flow graph as context to perform global inference on features with pairwise relationships.

需要先通过平均池化层变为具有合适的分辨率大小的

交通路网特征图F_G则需要通过最大池化层进行同样的变换得到F′_G。The first traffic flow characteristic map

It needs to pass through the average pooling layer first to have a suitable resolution size

The traffic road network feature map F _G needs to undergo the same transformation through the maximum pooling layer to obtain F′ _G .

构成TRE的输入。由于TRE需要一个序列作为输入，因此将F_t的二维空间维度折叠为水平维度进行输入扁平化，从而得到形状为d×HW的特征图。由于TRE的架构是排列不变的，这意味着原始二维输入的信息会丢失，因此，为了保留每个像素点的相对位置信息，在TRE的每个多头自注意力层都加入了空间位置编码。TRE对F′的进行编码，得到第二交通流量特征图，将第二交通流量特征图传递到TRD。According to the definition of the converter structure, each encoder layer in TRE is set to a standard structure consisting of a multi-head self-attention module. First, 1×1 convolution reduces the channel dimension of F′ from C to d, generating a new feature map

The inputs that make up the TRE. Since TRE requires a sequence as input, the two-dimensional spatial dimension of F _t is collapsed into the horizontal dimension for input flattening, resulting in a feature map with shape d × HW. Since the architecture of TRE is invariant in arrangement, it means that the information of the original two-dimensional input will be lost. Therefore, in order to preserve the relative position information of each pixel, the spatial position is added to each multi-head self-attention layer of TRE. coding. TRE encodes F' to obtain a second traffic flow feature map, and transmits the second traffic flow feature map to TRD.

TRD also follows the standard architecture of the converter, but in this embodiment, the TRD adopts a parallel structure and decodes features in parallel at each decoder layer, while the original transformer adopts an autoregressive model, which transmits only one spatial position encoding and one time at the input. Output encoding to predict the output sequence one element at a time.

TRD receives the second traffic flow feature map sent by TRE, takes the traffic road network feature map _F′G as the embedded encoding, and adds it to the input of the attention layer of each decoder for perceptual learning, the encoding of TRE The result is passed into TRD as the context. After TRD transformation, the third traffic flow feature map is the traffic flow feature map guided by the urban road network.

Finally, the RAFM model inputs the first traffic flow feature map and the third traffic flow feature map into a convolutional layer for feature fusion to generate a fine-grained traffic flow map, that is, a high-definition urban road traffic flow map.

Among them, it needs to be further explained that in this embodiment, the RAFM model is implemented based on the Python and PyTorch deep learning framework, and the training of the RAFM model adopts the stochastic gradient descent optimizer, the momentum is 0.9, the initial learning rate is 3e-4, all layers The filter weights of all are initialized by Xavier, which optimizes the RAFM model end-to-end by minimizing the mean absolute error (MAPE) between the inference results and the corresponding ground truth.

Embodiment 3

As shown in FIG. 5 , an embodiment of the present invention further provides a system for generating an urban road traffic flow map. The system is configured to execute the above-mentioned method for generating an urban road traffic flow map, including a data acquisition module 201, a traffic road network a feature map generation module 202, a first traffic flow feature map generation module 203, a second traffic flow feature map generation module 204, a third traffic flow feature map generation module 205, and a fine-grained traffic flow map generation module 206;

The data acquisition module 201 is used to acquire the coarse-grained traffic flow map of the target area, the environmental data of the target area, and the traffic map of the target area;

The traffic road network feature map generation module 202 is configured to input the coarse-grained traffic flow map and the traffic map into the first encoder for encoding, and generate a traffic road network feature map;

The first traffic flow feature map generating module 203 is configured to generate a first traffic flow feature map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map;

The second traffic flow feature map generation module 204 is configured to input the first traffic flow feature map into the second encoder for encoding, and generate a second traffic flow feature map;

The third traffic flow feature map generation module 205 is configured to input the second traffic flow feature map and the traffic road network feature map into the decoder for decoding to generate a third traffic flow feature map;

The fine-grained traffic flow map generating module 206 is configured to generate a fine-grained 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 further provides a device 30 for generating an urban road traffic flow map, and the device includes a processor 300 and a memory 301;

The memory 301 is used for storing program codes 302 and transmitting the program codes 302 to the processor;

The processor 300 is configured to execute the steps in the above-mentioned embodiment of the method for generating an urban road traffic flow map according to the instructions in the program code 302 .

Exemplarily, the computer program 302 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 301 and executed by the processor 300 to complete the this application. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process 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, a palmtop computer, and a cloud server. The terminal device may include, but is not limited to, the processor 300 and the memory 301 . Those skilled in the art can understand that FIG. 6 is only an example of the terminal device 30, and does not constitute a limitation on the terminal device 30. It may include more or less components than the one shown, or combine some components, or different components For example, the terminal device may further include an input and output device, a network access device, a bus, and the like.

The so-called processor 300 may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and 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 memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) equipped on the terminal device 30. card, flash card (Flash Card) and so on. Further, the memory 301 may also include both an internal storage unit of the terminal device 30 and an external storage device. The memory 301 is used for storing the computer program and other programs and data required by the terminal device. The memory 301 can also be used to temporarily store data that has been output or will be output.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above may refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, removable hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes.

The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the protection scope of the present invention. . It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims (10)

1. an urban road traffic flow map generation method, is characterized in that, comprises the following steps: Obtain the coarse-grained traffic flow map of the target area, the environmental data of the target area, and the traffic map of the target area; Inputting the coarse-grained 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 feature map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map; Inputting the first traffic flow feature map into a second encoder for encoding, and generating a second traffic flow feature map; Inputting the second traffic flow feature map and the traffic road network feature map into a decoder for decoding to generate a third traffic flow feature map; Based on the first traffic flow feature map and the third traffic flow feature map, a fine-grained traffic flow map is generated. 2 . The method for generating an urban road traffic flow map according to claim 1 , wherein the coarse-grained traffic flow map and the traffic map are input into the first encoder for encoding to generate a traffic road network. 3 . The specific process of the feature map is: based on the traffic map, generating a traffic road network map; The traffic road network map and the coarse-grained traffic flow map are weighted to obtain a weighted first traffic road network map, and the first traffic road network map is input into the first encoder, so that the The first encoder encodes the road features in the first traffic road network map to obtain a traffic road network feature map. 3 . The method for generating an urban road traffic flow map according to claim 2 , wherein the first encoder performs feature extraction on the first traffic road network map to obtain the specific characteristics of the traffic road network feature map. 4 . The process is: The first encoder performs horizontal convolution, vertical convolution, positive diagonal convolution and anti-diagonal convolution on the first traffic network map, respectively, to generate horizontal features, vertical features, positive diagonal features, and anti-diagonal features. feature, and a traffic road network feature map is obtained based on the horizontal feature, the vertical feature, the positive diagonal feature, and the opposite diagonal feature. 4. The method for generating an urban road traffic flow map according to claim 1, wherein a first traffic flow feature map is generated based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map The specific process is: converting the environmental data into an environmental factor embedding vector; performing feature extraction on the traffic road network feature map to generate a road network feature map; Feature fusion is performed on the road network feature map, the external factor embedding vector and the coarse-grained traffic flow map to generate a first traffic flow feature map. 5. The method for generating an urban road traffic flow map according to claim 4, wherein the feature fusion of the road network feature map, the external factor embedding vector and the coarse-grained traffic flow map is performed to generate the first traffic flow The specific process of the feature map is: Perform feature cascading on the external factor embedding vector and the coarse-grained traffic flow map to obtain a first cascade feature, and fuse the first cascade feature to obtain a primary traffic flow feature map; Feature cascading is performed on the road network feature map and the primary traffic flow feature map to obtain a second cascade feature, and the second cascade feature is fused to obtain the first traffic flow feature map. 6. The method for generating an urban road traffic flow map according to claim 1, wherein the first traffic flow characteristic map is input into the second encoder for encoding, and the second traffic flow characteristic map is generated. The specific process is: The first traffic flow feature map is input into the second encoder, so that the second encoder encodes the first traffic flow feature map based on a multi-head attention mechanism to generate a second traffic flow feature map. 7 . The method for generating an urban road traffic flow map according to claim 1 , wherein the second traffic flow feature map and the traffic road network feature map are input into a decoder for decoding to generate the first traffic flow map. 8 . The specific process of the three traffic flow feature maps is as follows: The traffic road network feature map is used as an embedded code, and the embedded code and the second traffic flow feature map are input into the decoder, so that the decoder decodes based on the multi-head attention mechanism to obtain a third traffic flow. Traffic characteristic map. 8 . The method for generating an urban road traffic flow map according to claim 1 , wherein, based on the first traffic flow feature map and the third traffic flow feature map, a specific method for generating a fine-grained traffic flow map is generated. 9 . The process is: Feature fusion is performed on the first traffic flow feature map and the third traffic flow feature map to generate a fine-grained traffic flow map. 9. A system for generating an urban road traffic flow map, wherein the system is used to execute the method for generating an urban road traffic flow map according to any one of claims 1 to 8, comprising 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-grained traffic flow map generation module; The data acquisition module is used to acquire the coarse-grained traffic flow map of the target area, the environmental data of the target area and the traffic map of the target area; The traffic road network feature map generation module is configured to input the coarse-grained traffic flow map and the traffic map into the first encoder for encoding, and generate a traffic road network feature map; The first traffic flow feature map generating module is configured to generate a first traffic flow feature map based on the traffic road network feature map, the environmental data and the coarse-grained traffic flow map; The second traffic flow feature map generation module is configured to input the first traffic flow feature map into a second encoder for encoding, and generate a second traffic flow feature map; The third traffic flow feature map generation module is configured to input the second traffic flow feature map and the traffic road network feature map into the decoder for decoding, and generate a third traffic flow feature map; The fine-grained traffic flow map generating module is configured to generate a fine-grained traffic flow map based on the first traffic flow feature map and the third traffic flow feature map. 10. An urban road traffic flow map generating device, comprising a processor and a memory; the memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the method for generating an urban road traffic flow map according to any one of claims 1 to 8 according to the instructions in the program code.

Images