A road recognition method, device, equipment and storage medium

Abstract

The application relates to the field of maps, and provides a road recognition method, device and equipment and a storage medium. The method comprises the following steps: collecting trajectory data of each vehicle driving on a target road in a set period; generating a target multi-channel color graph of each vehicle in the set period based on the obtained trajectory data; performing multi-time trajectory feature extraction on the target multi-channel color graph to obtain a predicted road network graph of the target road in the set period; and determining a closed road in the target road based on the predicted road network graph and an original road network graph corresponding to the target road. The trajectory data of the vehicle can be easily collected through a positioning device carried by the vehicle or a positioning application program and a positioning device used by a driver, so that the labor cost and material cost of laying the collection equipment are saved. Moreover, the method is not limited by the laying scale of the collection equipment, can be widely applied to large-scale data mining of the whole road network, has good application effect, and has high recognition accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of maps, and provides a road recognition method and device, equipment and a storage medium. BACKGROUND

[0002] With the development of information technology, the disadvantages of traditional static maps are constantly emerging. Due to the lack of dynamic and interactive, they cannot play their due function. In this background, the emergence and rapid development of dynamic maps provide new opportunities for the transformation of the map field.

[0003] In order to reflect the closed road which is temporarily not allowed for general vehicles to pass due to construction, accident and other reasons in real time in the dynamic map, a multi-layer statistical analysis model is usually called to analyze the traffic flow data, road parameters and environmental parameters of the current road, and to determine whether the current road belongs to the closed road.

[0004] However, the traditional road recognition method has the following disadvantages:

[0005] Disadvantage one, multiple data acquisition devices are needed to obtain multi-dimensional data such as road parameters, and the acquisition cost is high;

[0006] Disadvantage two, the traditional road recognition method can only be applied to cities and highways with good device laying conditions, and the application effect is not good for low-grade roads with poor device laying conditions, and the recognition accuracy is low. SUMMARY

[0007] The embodiments of the present application provide a road recognition method, device, equipment and storage medium to solve the problems of high acquisition cost, small application range and low recognition accuracy.

[0008] In a first aspect, the embodiments of the present application provide a road recognition method, comprising:

[0009] Respectively collecting trajectory data of each vehicle driving on the target road within a set period; wherein each set of trajectory data contains each trajectory point and attribute information corresponding to the corresponding vehicle;

[0010] Based on the obtained trajectory data, respectively generating a target route map corresponding to each vehicle within the set period and a candidate multi-channel color map;

[0011] Based on each candidate multi-channel color map, generating a corresponding target multi-channel color map, and obtaining a predicted road network map of the target road within the set period by performing multiple trajectory feature extractions on the target multi-channel color map;

[0012] The predicted road network map is matched with an original road network map corresponding to the target road, and a road section in the target road that is not matched is determined as a closed road.

[0013] In a second aspect, the embodiments of the present application further provide a road recognition device, comprising:

[0014] a data collection unit configured to collect trajectory data of each vehicle driving on the target road in a set period respectively, wherein each set of trajectory data comprises each trajectory point and attribute information corresponding to the corresponding vehicle;

[0015] a data processing unit configured to generate a candidate multi-channel color map corresponding to a target road map of each vehicle in the set period based on the obtained trajectory data of each vehicle;

[0016] generate a corresponding target multi-channel color map based on each candidate multi-channel color map, and obtain a predicted road network map of the target road in the set period by performing multi-trajectory feature extraction on the target multi-channel color map;

[0017] a road recognition unit configured to match the predicted road network map with an original road network map corresponding to the target road, and determine a road section in the target road that is not matched as a closed road.

[0018] Optionally, the data processing unit is configured to:

[0019] perform mean value processing on the average trajectory horizontal axis speed of each pixel in each candidate single-channel color map to obtain a first target single-channel color map, wherein the first target single-channel color map is used to represent the average trajectory horizontal axis speed of each vehicle in the set period; and

[0020] perform mean value processing on the average trajectory vertical axis speed of each pixel in each candidate single-channel color map to obtain a second target single-channel color map, wherein the second target single-channel color map is used to represent the average trajectory vertical axis speed of each vehicle in the set period; and

[0021] perform accumulation processing on the average traffic flow of each pixel in each candidate single-channel color map to obtain a third target single-channel color map, wherein the third target single-channel color map is used to represent the average traffic flow of each vehicle in the set period.

[0022] the first target single-channel color map, the second target single-channel color map, and the third target single-channel color map obtained are used as the target multi-channel color map.

[0023] Optionally, the data processing unit is configured to:

[0024] Perform the following operations on each of the candidate single-channel color maps to obtain the first target single-channel color map:

[0025] Obtain the average horizontal trajectory velocity of a group of pixels at the same position in each candidate single-channel color image;

[0026] The average horizontal axis velocity of a set of average trajectories is averaged to obtain the average horizontal axis velocity of pixels at the same position in the first target single-channel color map.

[0027] Optionally, the road recognition unit is used for:

[0028] Using a preset road network prediction model, the target multi-channel color map is subjected to initial trajectory feature extraction to obtain a candidate road network map;

[0029] Using the road network prediction model, secondary trajectory feature extraction is performed on the obtained candidate road network map to obtain the predicted road network map.

[0030] Optionally, the road recognition unit is used for:

[0031] Road network differencing is performed on the corresponding road segments of the target road drawn in the predicted road network map and the original road network map, respectively; wherein, for each road segment extracted from the original road network map, the following operations are performed:

[0032] When it is determined that the same road segment is drawn in both the predicted road network map and the original road network map, it is determined that the road segment matches.

[0033] When it is determined that a road segment is not drawn in the predicted road network map, it is determined that the road segment is mismatched.

[0034] Thirdly, embodiments of this application also provide a computer device, including a processor and a memory, wherein the memory stores program code, and when the program code is executed by the processor, the processor performs the steps of any of the above-described road recognition methods.

[0035] Fourthly, embodiments of this application also provide a computer-readable storage medium including program code, which, when the program product is run on a computer device, is used to cause the computer device to perform the steps of any of the above-described road recognition methods.

[0036] Fifthly, embodiments of this application also provide a computer program product, including computer instructions, which are executed by a processor using the steps of any of the road recognition methods described above.

[0037] The beneficial effects of this application are as follows:

[0038] The embodiment of the present application provides a road recognition method, device and equipment and a storage medium, and the method comprises the following steps: collecting trajectory data of each vehicle driving on a target road in a set period, wherein each trajectory data comprises each trajectory point and attribute information corresponding to the corresponding vehicle; based on the obtained trajectory data, a target route map of each vehicle in the set period is generated, and a corresponding candidate multi-channel color map is generated; then, based on the candidate multi-channel color map, a target multi-channel color map is generated, and the target multi-channel color map is subjected to multiple times of trajectory feature extraction, so as to obtain a predicted road network map of the target road in the set period; the predicted road network map is matched with an original road network map corresponding to the target road, and a road section in the target road that is not matched is determined as a closed road.

[0039] The trajectory data of the vehicle can be collected more easily through the positioning device carried by the vehicle or the positioning application program and the positioning device used by the driver, and the labor cost and material cost of laying the collection equipment are saved. Moreover, the data collection mode adopted in the embodiment of the present application is not limited to the laying scale of the collection equipment, and can be widely applied to large-scale data mining of the whole road network, has good application effect, and has high recognition accuracy.

[0040] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and obtained by the structure particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0041] The drawings described herein are used to provide further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their description are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings:

[0042] Figure 1 An optional schematic diagram of one application scenario in the embodiment of the present application;

[0043] Figure 2A A flowchart of a road recognition method provided by the embodiment of the present application;

[0044] Figure 2B A logic diagram of a road recognition method provided by the embodiment of the present application;

[0045] Figure 2C A flowchart of data preprocessing of a group of trajectory data provided by the embodiment of the present application;

[0046] Figure 2D A logic diagram of data preprocessing of a group of trajectory data provided by the embodiment of the present application;

[0047] Figure 2E A logic diagram for generating a predicted road network map of a target road using a road network prediction model is provided for the embodiments of the present application;

[0048] Figure 2F A model architecture diagram of a road network prediction model is provided for the embodiments of the present application;

[0049] Figure 2G A logic diagram for identifying a closed road based on a predicted road network map and an original road network map is provided for the embodiments of the present application;

[0050] Figure 3 A flowchart for identifying a closed road based on full road network data mining and road identification is provided for the embodiments of the present application;

[0051] Figure 4 A structural diagram of a road identification device is provided for the embodiments of the present application;

[0052] Figure 5 A structural diagram of a computer device is provided for the embodiments of the present application;

[0053] Figure 6 A structural diagram of a computing device is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described below in detail with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments described in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0055] Some terms in the embodiments of the present application are explained below to facilitate understanding by those skilled in the art.

[0056] 1. Road network (road network): refers to a road system composed of various roads in a certain area, which are interconnected and interwoven into a network distribution.

[0057] Among them, the road network composed of highways of various levels is also called highway network, and the road network composed of various functional trunk roads and regional roads within the jurisdiction of a town is also called urban road network, which can provide safe, rapid, economical and comfortable driving conditions for various transportation tools.

[0058] 2. Closed road: refers to the road that is temporarily not allowed for general vehicle communication due to construction, accident, control, etc.

[0059] 3. Elevation: elevation refers to the height of a point relative to the datum plane. There are four commonly used elevation systems, namely, orthometric height, normal height, geopotential height, and geodetic height. The definitions of elevation datum vary from country to country.

[0060] 4. Artificial intelligence (AI):

[0061] Artificial intelligence is the use of digital computers or digital computer-controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which aims to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is the design principle and implementation method of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0062] Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technology generally includes sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc.; artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0063] With the research and progress of artificial intelligence technology, artificial intelligence has been researched and applied in many fields, such as common smart home, intelligent customer service, virtual assistant, smart speaker, intelligent marketing, unmanned driving, autonomous driving, robot, intelligent medical treatment, etc. It is believed that with the development of technology, artificial intelligence will be applied in more and more fields and play an increasingly important value.

[0064] 5. Machine learning:

[0065] Machine learning is a multi-disciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a subject that studies how computers simulate or implement human learning behavior to acquire new knowledge or skills, and reorganize existing knowledge structure to continuously improve their performance.

[0066] Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent. Its applications are widespread in various fields of artificial intelligence, including deep learning, reinforcement learning, transfer learning, inductive learning, and rule-based learning.

[0067] The design concept of the embodiments of this application is briefly introduced below:

[0068] With the development of information technology, the drawbacks of traditional static maps have become increasingly apparent. Due to their lack of dynamism and interactivity, they are unable to fulfill their intended functions. Against this backdrop, the emergence and rapid development of dynamic maps have provided new opportunities for the transformation of the mapping field.

[0069] In order to reflect in real time in dynamic maps the closure of roads that are temporarily closed to general vehicles due to construction, accidents or other reasons, multi-layer statistical analysis models are usually called to analyze the traffic flow data, road parameters and environmental parameters of the current road to determine whether the current road is a closed road.

[0070] However, traditional road recognition methods have the following drawbacks:

[0071] One drawback is that multiple data acquisition devices are required to obtain multi-dimensional data such as road parameters, resulting in high acquisition costs.

[0072] The second drawback is that traditional road recognition methods can only be applied to cities and highways with well-maintained equipment. For low-grade roads with poor equipment conditions, the application effect is not good and the recognition accuracy is low.

[0073] In view of this, embodiments of this application provide a road recognition method, apparatus, device, and storage medium. The method includes: collecting trajectory data of each vehicle traveling on a target road within a set period, wherein each trajectory data includes: trajectory points and attribute information corresponding to the vehicle; generating candidate multi-channel color maps corresponding to the target route map of each vehicle within the set period based on the obtained trajectory data; generating a corresponding target multi-channel color map based on the candidate multi-channel color maps; obtaining a predicted road network map of the target road within the set period by performing multiple trajectory feature extractions on the target multi-channel color map; matching the predicted road network map with the original road network map corresponding to the target road; and determining road segments in the target road that do not match as closed roads.

[0074] Vehicle trajectory data can be more easily collected through vehicle-mounted positioning devices or driver-used positioning applications and devices, saving manpower and material costs associated with deploying data collection equipment. Furthermore, the data collection method used in this application is not limited by the scale of equipment deployment and can be widely applied to large-scale data mining across the entire road network, demonstrating good application results and high recognition accuracy.

[0075] Based on the obtained trajectory data, a target multi-channel color map reflecting the trajectory characteristics of each vehicle in a set period is obtained, and a road network prediction model based on deep learning is used to generate a predicted road network map. Based on the predicted road network map and the original road network map generated by the vehicle navigation system, the closed road in the target road is determined.

[0076] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings of the specification. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application, and the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0077] The embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, intelligent transportation, and assisted driving. Figure 1 A schematic diagram of one of the application scenarios is shown, which includes two physical terminal devices 110 and a server 130. The two physical terminal devices 110 are connected to the server 130 through wired or wireless networks.

[0078] The physical terminal device 110 in the embodiments of the present application is a computer device used by a user. The computer device includes but is not limited to a mobile phone, a computer, a smart voice interactive device, a smart home appliance, a vehicle-mounted terminal, an aircraft, etc.

[0079] The server 130 in the embodiments of the present application can be a standalone physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. basic cloud computing services, which are not limited in the present application.

[0080] The physical terminal device 110 responds to the navigation operation triggered by the target object, starts the pre-installed vehicle navigation application, and presents the vehicle navigation interface 120 on the screen of the physical terminal device 110. The navigation information input by the target object on the vehicle navigation interface 120 is sent to the vehicle navigation system deployed on the server 130, so that the vehicle navigation system forms a navigation route between the starting point and the destination based on the received navigation information. Finally, the physical terminal device 110 presents the corresponding navigation route in the vehicle navigation interface 120.

[0081] However, the closed road formed due to construction, accident, traffic control and other reasons becomes one of the important factors affecting the travel of the target object. Therefore, how to quickly and accurately find the closed road in the road network has gradually developed into a key concern of the vehicle navigation system.

[0082] Due to the large scale of the road network, if the existence of the closed road in each road in the road network is directly detected, it poses unprecedented challenges to the computing power and stability of the server 130. In order to reduce the computing power and load pressure of the server 130, the vehicle navigation system utilizes the time sequence characteristics and spatial characteristics of the trajectory data, combines the change trend of the total number of trajectory points with the high-precision road recognition method, and applies it in the large batch data mining scene of the whole road network, so as to quickly and accurately identify the closed road. Through timely updating of the dynamic road network, the vehicle navigation system helps to avoid the closed road when formulating the navigation route, and improves the use experience of the target object.

[0083] Specifically, the vehicle navigation system periodically collects the trajectory data of each vehicle running on each road of the original road network map, and based on the obtained trajectory data, when it is found that the total number of trajectory points of the same road in the current set period presents a sharp downward trend compared with the past, the vehicle navigation system considers that there may be a closed road in this road, and then marks it as a target road.

[0084] The vehicle navigation system further generates a target road map and a corresponding candidate multi-channel color map for each vehicle running on the target road in the set period based on the trajectory data of each vehicle running on the target road in the set period. Then, based on each candidate multi-channel color map, a corresponding target multi-channel color map is generated, and by performing multiple trajectory feature extractions on the target multi-channel color map, a predicted road network map of the target road in the set period is obtained. The predicted road network map is matched with the original road network map corresponding to the target road, and the road section in the target road that does not match is determined as a closed road.

[0085] Next, refer to the flowchart shown in Figure 2A and the logic diagram shown in Figure 2B to introduce the application process of the road recognition method in detail.

[0086] S201: Collect the trajectory data of each vehicle running on the target road in the set period respectively; wherein each set of trajectory data includes the attribute information of each trajectory point corresponding to the corresponding vehicle.

[0087] The trajectory data of each vehicle driving on the target road within a set period is collected by a positioning device such as a Globe Positioning System (GPS) or a Beidou satellite navigation system. Each set of trajectory data includes the corresponding trajectory points and attribute information (such as the trajectory vector speed, elevation, and geographical position of each trajectory point) of the corresponding vehicle.

[0088] The vehicle in the embodiments of the present application is not limited to a car equipped with a GPS, a Beidou satellite navigation system, or other positioning devices. It can also be a bicycle, an electric vehicle, a motorcycle, or other vehicles. When the vehicle is not equipped with a positioning device, the trajectory data of the corresponding vehicle can also be collected by a device such as a smart phone or a smart wearable device of the driver that is installed with a positioning application or a positioning device.

[0089] For example, the GPS trajectory data of cars a-c driving on road A within a day is collected by the GPS installed in the car, and the GPS trajectory data of a bicycle driving on road A within a day is collected by a smart watch worn by the driver.

[0090] S202: Based on the obtained trajectory data, a candidate multi-channel color map corresponding to the target route map of each vehicle within a set period is generated.

[0091] The data preprocessing is performed on the obtained trajectory data, and a candidate multi-channel color map corresponding to the target route map of each vehicle within a set period is generated. Referring to the flowchart shown in FIG. 2 and the logic diagram shown in FIG. 3, the following operations are performed for each vehicle: Figure 2C Figure 2D

[0092] S2021: Based on the attribute information of each trajectory point and a preset filtering rule, the abnormal trajectory points in a set of trajectory data corresponding to a vehicle a are removed, and the trajectory route of the vehicle a within a set period is generated based on the other trajectory points in the set of trajectory data.

[0093] Table 1 shows one of the filtering rules, which details the filtering conditions for abnormal trajectory points and normal trajectory points. The attribute information of each trajectory point in a set of trajectory data is obtained, and each time the attribute information of a trajectory point is obtained, it is matched with the filtering rule for processing, so as to remove the trajectory points in the set of trajectory data that have significant abnormal conditions or are likely to be non-vehicle driving.

[0094] Table 1

[0095]

[0096] ​​S2022: Smooth the trajectory to obtain the corresponding target route map; wherein, the target route map contains multiple pixels, and each pixel contains zero or at least one trajectory point.

[0097] Because a set of trajectory data may contain trajectory points from different data sources, and the reporting time intervals between each trajectory point are also different, the sparsity of the trajectory points in this set of trajectory data is inconsistent. In order to more easily generate the image processed by the road network prediction model, step 2022 needs to be performed first to smooth the trajectory route.

[0098] First, such as Figure 2D As shown in the left figure, the region containing trajectory data (also known as a block, with a region size of 1km×1km) is divided into multiple pixels (also known as piecel, totaling 1024×1024 pixels), with each pixel having a spatial resolution of 1m×1m.

[0099] Then, each trajectory point in the trajectory route is obtained sequentially. For each trajectory point a', the relative distance between the two trajectory points (e.g., Euclidean distance, spherical distance, Manhattan distance) is determined based on the trajectory point positions of the trajectory point a' and its adjacent trajectory points. The position of each trajectory point is determined based on the position of the pixel that the corresponding trajectory point falls into among all pixels in the region.

[0100] like Figure 2D As shown in the left figure, a Cartesian coordinate system is established with the upper left corner of the region as the origin. In this coordinate system, the position of trajectory point a' is (1,7), and the position of the adjacent trajectory point b' is (2,6). Then the Euclidean distance between the two points is...

[0101] When the relative distance exceeds a set distance, linear interpolation is performed on the trajectory segment between the two trajectory points to obtain new trajectory points. Based on the original trajectory points and the newly inserted trajectory points, a new trajectory is generated, such as... Figure 2D The left image shows the target route map.

[0102] S2023: Based on the attribute information of each trajectory point within each pixel and the total number of trajectory points, each pixel is mapped to multiple candidate single-channel color maps of the same size as the target route map.

[0103] First, based on the attribute information of each trajectory point within each pixel and the total number of trajectory points, the average horizontal axis velocity and the average vertical axis velocity of each pixel are determined respectively. Then, based on the total number of trajectory points within each pixel and the maximum total number of trajectory points, the average traffic flow of each pixel is determined.

[0104] The attribute information of each trajectory point includes a trajectory vector velocity of the corresponding trajectory point. The vector nature of the velocity means that the velocity has both a velocity magnitude and a velocity direction. Therefore, a trajectory vector velocity can be decomposed into a trajectory horizontal axis velocity and a trajectory vertical axis velocity.

[0105] A one-to-many mapping relationship exists between the pixels and the trajectory points, i.e., one pixel contains zero or at least one trajectory point, but one trajectory point can only fall into one pixel. Therefore, the trajectory horizontal axis velocity, the trajectory vertical axis velocity of each trajectory point in each pixel, and the total number of trajectory points contained in each pixel are substituted into the formulas 1-2 to obtain the average trajectory horizontal axis velocity and the average trajectory vertical axis velocity of each pixel.

[0106]

[0107]

[0108] When there is only one trajectory point in a pixel, SinSpeed represents the trajectory horizontal axis velocity of the trajectory point in the pixel, and CosSpeed represents the trajectory vertical axis velocity of the trajectory point in the pixel. When there are at least two trajectory points in a pixel, SinSpeed represents the average trajectory horizontal axis velocity of the at least two trajectory points in the pixel, and CosSpeed represents the average trajectory vertical axis velocity of the at least two trajectory points in the pixel.

[0109] MaxSinSpeed represents the maximum trajectory horizontal axis velocity appearing in all pixels in the region, and MaxCosSpeed represents the maximum trajectory vertical axis velocity appearing in all pixels in the region. SinSpeed' represents the average trajectory horizontal axis velocity of a pixel, and CosSpeed' represents the average trajectory vertical axis velocity of a pixel.

[0110] Similarly, the total number of trajectory points in each pixel and the maximum total number of trajectory points are substituted into the formula 3 to obtain the candidate traffic flow of each pixel. Then, the obtained candidate traffic flows are substituted into the formula 4, and the minimum value between the candidate traffic flow and 255 is taken as the average traffic flow of each pixel.

[0111] In the formula, PointNum represents the total number of trajectory points contained in a pixel, MaxPointNum represents the maximum total number of trajectory points appearing in all pixels in the region, Flow' represents the candidate traffic flow of a pixel, and Flow represents the average traffic flow of a pixel.

[0112]

[0113] Flow = min(Flow', 255) Formula 4

[0114] Secondly, the average trajectory horizontal axis speed, the average trajectory vertical axis speed and the average traffic volume of each pixel are respectively mapped into a plurality of candidate single-channel color maps with the same size as the target route map.

[0115] After obtaining the plurality of components such as the average trajectory horizontal axis speed, the average trajectory vertical axis speed and the average traffic volume of each pixel, the average trajectory horizontal axis speed of each pixel is mapped into a candidate single-channel color map with the same size as the target route map, to obtain a G channel map as shown in FIG. 6. Figure 2D Similarly, the B channel map and the R channel map as shown in FIGS. 7 and 8 are obtained by mapping the other two components. Figure 2D

[0116] S2024: The generated plurality of candidate single-channel color maps are taken as candidate multi-channel color maps.

[0117] S203: Based on each candidate multi-channel color map, a corresponding target multi-channel color map is generated, and by performing multiple trajectory feature extractions on the target multi-channel color map, a predicted road network map of the target road in a set period is obtained.

[0118] For the target route map of each vehicle in the set period, the steps 2021-2024 are performed to obtain a corresponding candidate multi-channel color map.

[0119] After obtaining each candidate multi-channel color map, firstly, the average trajectory horizontal axis speed of each pixel in each candidate single-channel color map is processed by mean value to obtain a first target single-channel color map, which is used to represent the average trajectory horizontal axis speed of each vehicle in the set period; and

[0120] The average trajectory vertical axis speed of each pixel in each candidate single-channel color map is processed by mean value to obtain a second target single-channel color map, which is used to represent the average trajectory vertical axis speed of each vehicle in the set period.

[0121] The generation manners of the first target single-channel color map and the second target single-channel color map are similar. Taking the candidate single-channel color map representing the average trajectory horizontal axis speed as an example, the process of generating the first target single-channel color map is as follows:

[0122] The average trajectory horizontal axis speeds of a group of pixels at the same position in each candidate single-channel color map are obtained, and the average trajectory horizontal axis speeds are processed by mean value to obtain the average trajectory horizontal axis speed at the same position in the first target single-channel color map.

[0123] ​In addition, the average traffic flow of each pixel in each candidate single-channel color map is accumulated to obtain the third target single-channel color map, which is used to characterize the average traffic flow of each vehicle within a set period.

[0124] Specifically, the average traffic flow of a group of pixels at the same position in each candidate single-channel color image is obtained, and then the average traffic flow of each group is accumulated. The accumulated values ​​are used as the average traffic flow of pixels at the same position in the third target single-channel color image.

[0125] Then, the obtained first target single-channel color map, second target single-channel color map, and third target single-channel color map are used as target multi-channel color maps. These target multi-channel color maps are then used as input to the road network prediction model.

[0126] like Figure 2E As shown, a preset road network prediction model is used to perform initial trajectory feature extraction on the target multi-channel color map to obtain a candidate road network map. However, due to interference factors such as unstable positioning signals and inherent accuracy issues (within the normal range) in the original input data, the trajectory points in the candidate road network map may drift as a whole. This is especially true when predicting complex road networks, where drifting trajectory points can easily cause trajectory matching errors.

[0127] Therefore, in order to improve the accuracy of feature extraction of the model, a road network prediction model is also used to perform secondary trajectory feature extraction on the candidate road network map, so that the candidate road network map converges quickly again, significantly enhancing the accuracy of feature extraction of the model and obtaining a more accurate predicted road network map.

[0128] The road network prediction model is built upon the D-LinkNet model. For example... Figure 2F As shown, the road network prediction model includes an encoder (corresponding to...) Figure 2F Part A), convolutional layer (corresponding to Figure 2F Part B of the decoder and the decoder (corresponding to Figure 2F Part C of the document.

[0129] A ResNet34 pre-trained on the ImageNet dataset is used as the encoder to extract low-level features from the target's multi-channel color map. Dilated convolutional layers with a serial-parallel structure are used as convolutional layers to significantly increase the receptive field size of the features without reducing the feature map resolution, and to fuse multi-scale features, enabling the convolutional layers to extract more abstract high-level trajectory features from low-level trajectory features. The decoder uses transposed convolutional layers to upsample the high-level trajectory features output from the hidden layers, restoring the feature map resolution from 32×32 to 1024×1024.

[0130] S204: Perform matching processing on the predicted road network map and the original road network map corresponding to the target road, and determine the road segment that does not match in the target road as a closed road.

[0131] Differences between the predicted road network map and the original road network map are respectively performed on the corresponding road segments of the target road drawn in the predicted road network map and the original road network map; wherein, each time a road segment is extracted from the original road network map, the following operations are performed: when it is determined that the road segment is drawn in both the predicted road network map and the original road network map, it is determined that the road segment matches consistently; when it is determined that the road segment is not drawn in the predicted road network map, it is determined that the road segment does not match.

[0132] Since the original road network map (corresponding to the left part of Figure 2G ) generated by the vehicle navigation system at the beginning also contains redundant information such as road running state, surrounding buildings, etc., in order to facilitate the matching processing of the two maps afterwards, the road information of the target road is extracted from the original road network map according to the road grade of the target road, and the original road network map (corresponding to the middle part of Figure 2G ) that only shows the shape of the target road is obtained.

[0133] In both the original road network map (corresponding to the middle part of Figure 2G ) and the predicted road network map (corresponding to the right part of Figure 2G ), if the pixel value is 0, it means that the corresponding pixel is located in the background area; if the pixel value is 1, it means that the corresponding pixel is located in the road area.

[0134] Since the original road network map (corresponding to the middle part of Figure 2G ) and the predicted road network map have the same size, each time a road segment is extracted from the original road network map, the pixel values of each pixel on the road segment are compared with the pixel values of each pixel at the same position in the predicted road network map; if all the pixel values are the same, it means that the road segment is drawn in both the predicted road network map and the original road network map, and it is determined that the road segment matches consistently; if the pixel values of continuous multiple pixels are different, it means that the road segment is not drawn in the predicted road network map, and it is determined that the road segment does not match, and the road segment that does not match is determined as a closed road.

[0135] As can be seen from the original road network map shown on the left, Figure 2G the trajectory points distributed on the middle road segment are obviously more sparse and discontinuous, so in the predicted road network map generated based on the trajectory data of each vehicle driving on the target road within a set period, it can also be seen that there is a missing and discontinuous situation at this road segment. By performing road network difference between the original road network map (corresponding to the middle part of Figure 2G ) and the predicted road network map, the closed road in the target road can be quickly locked, and by updating the dynamic road network in time, the vehicle navigation system can help to avoid the closed road when formulating the navigation route, and improve the use experience of the target object.

[0136] ReferenceFigure 3 The flowchart shows the process of identifying the closed road based on the whole road network data mining and road identification as follows:

[0137] S301: The vehicle navigation system generates a corresponding original road network graph based on the navigation information input by the target object;

[0138] S302: The vehicle navigation system respectively collects the trajectory data of each vehicle driving on each road of the original road network graph within a set period;

[0139] S303: The vehicle navigation system, based on the obtained trajectory data, when it is found that the total number of trajectory points of the same road within the current set period presents a sharp downward trend compared with the past, marks the road as a target road;

[0140] S304: The vehicle navigation system respectively collects the trajectory data of each vehicle driving on the target road within a set period;

[0141] S305: The vehicle navigation system respectively generates a corresponding candidate RGB channel graph based on the target route graph of each vehicle within the set period based on the obtained trajectory data;

[0142] S306: The vehicle navigation system generates a corresponding target RGB channel graph based on each candidate RGB channel graph, and uses a road network prediction model to perform multiple trajectory feature extractions on the target RGB channel graph to generate a corresponding predicted road network graph;

[0143] S307: The vehicle navigation system performs road network difference on the predicted road network graph and the original road network graph, and determines the road section that does not match in the target road as a closed road.

[0144] Based on the same inventive concept as the above method embodiment, the present embodiment also provides a road identification device. As shown in Figure 4 The road identification device 400 can include:

[0145] The data acquisition unit 401 is configured to respectively collect trajectory data of each vehicle driving on the target road within a set period; wherein each set of trajectory data contains: each trajectory point and attribute information corresponding to the corresponding vehicle;

[0146] The data processing unit 402 is configured to respectively generate a corresponding candidate multi-channel color graph based on the target route graph of each vehicle within the set period based on the obtained trajectory data;

[0147] Based on each candidate multi-channel color graph, a corresponding target multi-channel color graph is generated, and a predicted road network graph of the target road within the set period is obtained by performing multiple trajectory feature extractions on the target multi-channel color graph;

[0148] The road recognition unit 403 is configured to match the predicted road network map with the original road network map corresponding to the target road, and determine a road segment in the target road as a closed road if the matching is inconsistent.

[0149] Optionally, the data processing unit 402 is configured to:

[0150] For each vehicle, the following operations are performed:

[0151] Based on the attribute information of each trajectory point and a preset screening rule, an abnormal trajectory point in a group of trajectory data corresponding to a vehicle is removed, and a trajectory route of the vehicle in a set period is generated based on other trajectory points in the group of trajectory data;

[0152] The trajectory route is smoothed to obtain a corresponding target route map; wherein the target route map comprises a plurality of pixels, and each pixel contains zero or at least one trajectory point;

[0153] Based on the attribute information of the trajectory points in each pixel and the total number of the trajectory points, each pixel is mapped into a plurality of candidate single-channel color graphs with the same size as the target route map;

[0154] The plurality of generated candidate single-channel color graphs are used as candidate multi-channel color graphs.

[0155] Optionally, the data processing unit 402 is configured to:

[0156] The area containing the trajectory route is divided into a plurality of pixels;

[0157] Each trajectory point in the trajectory route is obtained in sequence, and based on the trajectory point position of each of a trajectory point and an adjacent trajectory point, the relative distance between the two trajectory points is determined; wherein each trajectory point position is determined based on the position of the corresponding trajectory point in all pixels in the area;

[0158] When the relative distance exceeds a set distance, linear interpolation processing is performed on the trajectory segment between the two trajectory points to obtain a new trajectory point;

[0159] Based on the original trajectory points and the newly inserted trajectory points, a corresponding target route map is generated.

[0160] Optionally, the data processing unit 402 is configured to:

[0161] Based on the attribute information of each trajectory point in each pixel and the total number of the trajectory points, the average trajectory horizontal axis speed and the average trajectory vertical axis speed of each pixel are determined, and based on the total number of the trajectory points in each pixel and the maximum total number of the trajectory points, the average traffic flow of each pixel is determined;

[0162] The average trajectory horizontal axis speed, the average trajectory vertical axis speed and the average traffic volume of each pixel are respectively mapped into a plurality of candidate single-channel color maps with the same size as the target route map.

[0163] Optionally, the data processing unit 402 is configured to:

[0164] The average trajectory horizontal axis speeds of the pixels in each candidate single-channel color map are processed by mean value to obtain a first target single-channel color map, and the first target single-channel color map is used to represent the average trajectory horizontal axis speeds of each vehicle in a set period; and

[0165] The average trajectory vertical axis speeds of the pixels in each candidate single-channel color map are processed by mean value to obtain a second target single-channel color map, and the second target single-channel color map is used to represent the average trajectory vertical axis speeds of each vehicle in a set period; and

[0166] The average traffic volumes of the pixels in each candidate single-channel color map are processed by accumulation to obtain a third target single-channel color map, and the third target single-channel color map is used to represent the average traffic volumes of each vehicle in a set period.

[0167] The first target single-channel color map, the second target single-channel color map and the third target single-channel color map are used as a target multi-channel color map.

[0168] Optionally, the data processing unit 402 is configured to:

[0169] The following operations are performed on each candidate single-channel color map to obtain a first target single-channel color map:

[0170] The average trajectory horizontal axis speeds of a group of pixels at the same position in each candidate single-channel color map are obtained;

[0171] The average trajectory horizontal axis speeds are processed by mean value to obtain the average trajectory horizontal axis speed of the pixel at the same position in the first target single-channel color map.

[0172] Optionally, the road recognition unit 403 is configured to:

[0173] A preset road network prediction model is used to perform first trajectory feature extraction on the target multi-channel color map to obtain a candidate road network map.

[0174] The road network prediction model is used to perform second trajectory feature extraction on the obtained candidate road network map to obtain a predicted road network map.

[0175] Optionally, the road recognition unit 403 is configured to:

[0176] respectively, the corresponding road segments of the target road drawn in the predicted road network map and the original road network map are subjected to road network difference; wherein, each time a road segment is extracted in the original road network map, the following operations are performed:

[0177] When it is determined that one road segment is drawn in both the predicted road network map and the original road network map, it is determined that the one road segment is matched consistently.

[0178] When it is determined that one road segment is not drawn in the predicted road network map, it is determined that the one road segment is matched inconsistently.

[0179] For the convenience of description, the above parts are divided into modules (or units) according to functions and are described respectively. Of course, in the implementation of the present application, the functions of the modules (or units) can be realized in the same or multiple software or hardware.

[0180] After introducing the access method and device of the service platform of the exemplary embodiment of the present application, next, the computer device according to another exemplary embodiment of the present application is introduced.

[0181] Those skilled in the art can understand that each aspect of the present application can be implemented as a system, a method or a program product. Therefore, each aspect of the present application can be specifically implemented as a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software, which can be collectively referred to as "circuitry", "module" or "system" here.

[0182] Based on the same inventive concept as the above method embodiment, the present embodiment also provides a computer device, which is shown in Figure 5 The computer device 500 can at least include a processor 501 and a memory 502. The memory 502 stores program code, which, when executed by the processor 501, causes the processor 501 to perform the steps of any one of the road identification methods described above.

[0183] In some possible embodiments, the computing device according to the present application can at least include at least one processor and at least one memory. The memory stores program code, which, when executed by the processor, causes the processor to perform the steps of the road identification method according to various exemplary embodiments of the present application described above in the specification. For example, the processor can perform the steps as shown in Figure 2A

[0184] The computing device 600 according to this embodiment of the present application will be described below with reference to Figure 6 The computing device 600 is only an example and should not impose any limitation on the functions and use range of the embodiments of the present application. Figure 6 The computing device 600 is only an example and should not impose any limitation on the functions and use range of the embodiments of the present application. ​

[0185] As Figure 6 shown, computing device 600 is in the form of a general-purpose computing device. The components of computing device 600 can include, but are not limited to, at least one processing unit 601, at least one memory 602, and a bus 603 that couples various system components including the memory 602 to the processing unit 601.

[0186] The bus 603 represents one or more of any of several bus structures, including a memory bus or memory controller, a peripheral bus, a processor or local bus, using any of a variety of bus architectures.

[0187] The memory 602 can include read-only memory (ROM) 6023 in the form of flash memory or other nonvolatile memory. The memory 602 can also include random access memory (RAM) 6021 in the form of synchronous dynamic RAM (SDRAM), static RAM (SRAM), or other types of RAM equivalent.

[0188] The memory 602 can also include a program / utility 6025 having a set of program modules 6024, including an operating system, one or more application programs, other program modules, and program data, each of which can give the computing device 600 its functionality, some or all of which can be implemented as instructions stored in the memory 602. A user can enter commands and information into the computing device 600 through one or more input devices, such as a keyboard and a pointing device, commonly utilized

[0189] Computing device 600 can also communicate with one or more external devices 604 such as a keyboard or pointing device, through an input / output (I / O) interface 605. Note that the computing device 600 can also communicate with one or more devices that enable a user to interact with the computing device 600 in a network environment. For example, the computing device 600 can communicate with an application on the same device or on a different device. The computing device 600 can communicate with any devices (such as a printer, a scanner, a

[0190] Based on the same inventive concept as the method embodiments described above, the various aspects of the method for accessing the service platform provided by the present application can also be implemented in the form of a program product, which includes program code for causing a computer device to perform the steps of the road identification method according to various exemplary embodiments of the present application described above in the specification when the program product is run on the computer device, for example, the computer device can perform the steps as shown in FIG. 13. Figure 2A

[0191] The program product can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium may, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0192] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all the preferred embodiments and all the changes and modifications falling within the scope of the present application.

[0193] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.​

Claims

1. A road recognition method characterized by comprising: The method comprises the following steps: Collecting trajectory data of each vehicle driving on a target road in a set period respectively; each set of trajectory data comprises respective trajectory points and attribute information of the corresponding vehicle; For each vehicle, the following operations are performed: based on the attribute information of each trajectory point and a preset screening rule, abnormal trajectory points in a set of trajectory data corresponding to one vehicle are removed, and a trajectory route of the one vehicle in the set period is generated based on other trajectory points in the set of trajectory data; the trajectory route is smoothed to obtain a corresponding target route map, wherein the target route map comprises a plurality of pixels, each pixel contains zero or at least one trajectory point; based on the attribute information and the total number of trajectory points in each pixel, each pixel is mapped into a plurality of candidate single-channel color graphs with the same size as the target route map; and the generated plurality of candidate single-channel color graphs are used as candidate multi-channel color graphs; Based on each candidate multi-channel color graph, a corresponding target multi-channel color graph is generated, and by performing multiple trajectory feature extractions on the target multi-channel color graph, a predicted road network graph of the target road in the set period is obtained; The predicted road network graph and an original road network graph corresponding to the target road are matched, and road sections in the target road that do not match are determined as closed roads.

2. The method of claim 1, wherein, The smoothing of the trajectory route to obtain the target route map comprises: Dividing an area containing the trajectory route into a plurality of pixels; Obtaining each trajectory point in the trajectory route in turn, and determining the relative distance between two trajectory points based on the trajectory point positions of the two trajectory points when one trajectory point is obtained; wherein each trajectory point position is determined based on the position of the corresponding trajectory point in all pixels in the area; When the relative distance exceeds a set distance, performing linear interpolation processing on the trajectory segment between the two trajectory points to obtain a new trajectory point; Based on the original trajectory points and the newly inserted trajectory points, a corresponding target route map is generated.

3. The method of claim 1, wherein, The mapping of each pixel into a plurality of candidate single-channel color graphs with the same size as the target route map based on the attribute information and the total number of trajectory points in each pixel comprises: Based on the attribute information and the total number of trajectory points in each pixel, the average trajectory horizontal axis speed and the average trajectory vertical axis speed of each pixel are determined, and based on the total number of trajectory points and the maximum number of trajectory points in each pixel, the average traffic flow of each pixel is determined; The average trajectory horizontal axis speed, the average trajectory vertical axis speed and the average traffic flow of each pixel are mapped into a plurality of candidate single-channel color graphs with the same size as the target route map.

4. The method of claim 3, wherein, The generation of a corresponding target multi-channel color graph based on each candidate multi-channel color graph comprises: perform mean value processing on average trajectory horizontal axis speeds of pixels in each candidate single-channel color graph to obtain a first target single-channel color graph, the first target single-channel color graph being used to represent average trajectory horizontal axis speeds of the vehicles in the set period; and perform mean value processing on average trajectory vertical axis speeds of pixels in each candidate single-channel color graph to obtain a second target single-channel color graph, the second target single-channel color graph being used to represent average trajectory vertical axis speeds of the vehicles in the set period; and perform accumulation processing on average traffic flows of pixels in each candidate single-channel color graph to obtain a third target single-channel color graph, the third target single-channel color graph being used to represent average traffic flows of the vehicles in the set period; obtain the first target single-channel color graph, the second target single-channel color graph and the third target single-channel color graph as the target multi-channel color graph.

5. The method of claim 4, wherein, The mean value processing on the average trajectory horizontal axis speeds of the pixels in each candidate single-channel color graph to obtain the first target single-channel color graph includes: perform the following operations on each candidate single-channel color graph respectively to obtain the first target single-channel color graph: obtain average trajectory horizontal axis speeds of a group of pixels at the same position in each candidate single-channel color graph; perform mean value processing on a group of average trajectory horizontal axis speeds to obtain average trajectory horizontal axis speeds of pixels at the same position in the first target single-channel color graph.

6. The method according to any one of claims 1 to 3, wherein The obtaining of the predicted road network graph of the target road in the set period by performing multiple times of trajectory feature extraction on the target multi-channel color graph includes: perform first-time trajectory feature extraction on the target multi-channel color graph using a preset road network prediction model to obtain a candidate road network graph; perform second-time trajectory feature extraction on the obtained candidate road network graph using the road network prediction model to obtain the predicted road network graph.

7. The method of any one of claims 1-3, wherein, The matching processing on the predicted road network graph and an original road network graph corresponding to the target road includes: perform road network difference on corresponding road segments of the target road drawn in the predicted road network graph and the original road network graph respectively; wherein, each time a road segment is extracted in the original road network graph, the following operations are performed: when it is determined that the one road segment is drawn in both the predicted road network graph and the original road network graph, it is determined that the one road segment is matched consistently; when it is determined that the one road segment is not drawn in the predicted road network graph, it is determined that the one road segment is matched inconsistently.

8. A road recognition apparatus characterized by comprising: The method includes: a data collection unit configured to collect trajectory data of each vehicle driving on the target road in a set period respectively; wherein, each group of trajectory data contains each trajectory point and attribute information corresponding to the corresponding vehicle; The data processing unit is configured to: for each vehicle, remove an abnormal trajectory point in a set of trajectory data corresponding to the vehicle based on attribute information of the trajectory point and a preset screening rule, and generate a trajectory route of the vehicle in a set period based on other trajectory points in the set of trajectory data; and perform smoothing processing on the trajectory route to obtain a corresponding target route map, wherein the target route map includes a plurality of pixels, and each pixel contains zero or at least one trajectory point. The data processing unit is configured to: map each pixel to a plurality of candidate single-channel color graphs with the same size as the target route map based on attribute information of a trajectory point in each pixel and a total number of trajectory points; and use the generated plurality of candidate single-channel color graphs as candidate multi-channel color graphs. The data processing unit is configured to: generate a corresponding target multi-channel color graph based on each candidate multi-channel color graph, and obtain a predicted road network graph of the target road in the set period by performing multiple times of trajectory feature extraction on the target multi-channel color graph.

9. The apparatus of claim 8, wherein, The road recognition unit is configured to: perform matching processing on the predicted road network graph and an original road network graph corresponding to the target road, and determine a road section in the target road that is inconsistent in matching as a closed road. The data processing unit is configured to: divide an area including the trajectory route into a plurality of pixels; sequentially obtain each trajectory point in the trajectory route, and determine a relative distance between two trajectory points based on trajectory point positions of the two trajectory points when each trajectory point is obtained; wherein each trajectory point position is determined based on a position of a corresponding trajectory point in all pixels in the area; perform linear interpolation processing on a trajectory line segment between the two trajectory points when the relative distance exceeds a set distance to obtain a new trajectory point; and generate a corresponding target route map based on original trajectory points and newly inserted trajectory points. The data processing unit is configured to: determine an average trajectory horizontal axis speed and an average trajectory vertical axis speed of each pixel based on attribute information of each trajectory point in each pixel and a total number of trajectory points, and determine an average traffic flow of each pixel based on a total number of trajectory points in each pixel and a maximum total number of trajectory points; and map the average trajectory horizontal axis speed, the average trajectory vertical axis speed, and the average traffic flow of each pixel to a plurality of candidate single-channel color graphs with the same size as the target route map. The processor and the memory, wherein the memory stores program code, and when the program code is executed by the processor, the processor executes the steps of the method in any one of claims 1-7. The program code, when executed on the computer device, causes the computer device to execute the steps of the method in any one of claims 1-7.

10. The apparatus of claim 8, wherein, The computer instructions, when executed by the processor, implement the steps of the method in any one of claims 1-7. ​ ​ 11. A computer device, comprising: ​ 12. A computer-readable storage medium, characterized in that, ​ 13. A computer program product, characterised in that, ​

Images