CN107229690B - High-precision dynamic map data processing system and method based on roadside sensors - Google Patents

Abstract

The invention discloses a kind of Dynamic High-accuracy map datum processing system and method based on trackside sensor, which includes: map data collecting end, generates server-side for high-precision map and provides magnanimity, the map road initial data of diversification；It includes: map datum processing module that high-precision map, which generates server-side, map road initial data is subjected to road feature extraction, hierarchical design is used to all roadway characteristics, same road information is spliced, first roadway characteristic data fusion is completed in roadway characteristic level, image mosaic finally is completed in image level, the result of splicing generates the downward projection figure of road；Map generates visualization model, marks road information on downward projection figure, and the road information and downward projection figure that have marked collectively constitute the map datum of high-precision map, and map datum is carried out visual edit, generates Dynamic High-accuracy map.The present invention can reduce accurately map generalization cost；Realize the quick update of Dynamic High-accuracy map.

Description

High-precision dynamic map data processing system and method based on roadside sensors

technical field

The invention relates to the technical field of computers, in particular to a high-precision dynamic map data processing system and method based on roadside sensors.

Background technique

At present, with the rapid development of autonomous driving technology, the importance of high-precision maps has become increasingly prominent, and it has become an indispensable part of the realization of driverless and intelligent transportation. The accuracy of the existing navigation maps is generally not high, and the entire road is used as the object to provide road information data or to issue navigation instructions. This type of navigation map is called a road-level map, which greatly simplifies the actual traffic environment and can provide The information content is low, the accuracy is low, and the driver's assistance ability is low.

The maps required for autonomous driving must not only have high precision, but also have a large number of rich road surrounding details. The navigation accuracy of ordinary maps can only reach the level of meters, and the high-precision maps can be accurate to the level of 10cm, which not only increases the data related to lane attributes, Various types of data such as overhead objects, guardrails, obstacles, road edge types, roadside landmarks, etc. have also been added. Multivariate and heterogeneous massive map data takes up a lot of storage space, and a single-layer high-precision map cannot meet the needs of real-time update.

The application of technologies such as deep learning and image recognition in the field of high-precision maps can greatly improve the efficiency of map data collection and processing. The continuous improvement of autonomous driving technology and user needs has put forward higher requirements for the data capacity, accuracy and update frequency of high-precision maps. There are many technical bottlenecks in the traditional map data collection and drawing methods. Image recognition, big data processing, etc. , deep learning and other artificial intelligence technologies can automatically identify traffic signs, ground signs, lane lines, signal lights, etc., realize automatic extraction of road and POI information from panoramic images, improve data processing efficiency and update frequency, and ensure data accuracy.

At present, most of the high-precision maps are produced by professional staff to re-collect all road information, and plan to periodically re-update most areas after the collection is completed. The acquisition equipment of this method is often a collection vehicle equipped with special equipment such as lidar. Japanese automakers such as Mitsubishi and Toyota have teamed up with Japanese map maker Zenrin to make three-dimensional dynamic maps. The plan is to profile roads with specialized vehicles equipped with high-end sensors, the first step being to cover 300 kilometers of Japan's main highways. Here, TomTom, and Google are doing 3D maps in a similar way. The domestic traditional map dealer AutoNavi meets the required 10cm-level accuracy by assembling 2 lidars and 4 cameras. Tencent, Baidu, NavInfo and other companies are also making high-precision maps in a similar way.

The above-mentioned original map information collected by special on-board sensors is very accurate, but there are the following problems:

1) The cost of in-vehicle equipment remains high. The use of lidar to collect information has high accuracy and good overallity, but the cost is high, the amount of data is large, and the generated image is a reflectivity image, which is different from the real scene;

2) The data processing efficiency is low, the period from map data collection to map update is long, and the actual road condition feature attribute state has already changed when the map is updated, which cannot effectively reflect the dynamic feature information of the actual road in a timely manner, hindering location services. The rapid development of autonomous driving has reduced the safety and reliability of unmanned vehicles;

3) The collected data is a dense point cloud, the data density is extremely large, consumes a lot of computing resources, and the later map traffic is high;

4) The collected road feature information content is limited. For some specific road features, specific sensors (such as temperature and humidity, road water and other weather-related dynamic feature data) are required to complete data collection. Therefore, the vehicle-mounted collection method cannot meet the requirements of automatic driving. High-precision map content requirements.

Therefore, how to produce or update high-precision maps with low cost, high efficiency and accuracy is an urgent problem to be solved.

SUMMARY OF THE INVENTION

In view of this, in order to solve the problems of large data volume, difficult processing, high cost, and long map update cycle in the prior art map collection method, the present invention proposes a high-precision dynamic map data processing system and method based on roadside sensors, High-precision map generation is realized through roadside sensors and AI technology, with low cost, and the generated results can effectively feed back the current state of road features in real time, providing accurate driving assistance information for autonomous driving.

The present invention solves the above-mentioned problems through the following technical means:

A high-precision dynamic map data processing system based on roadside sensors, comprising:

The map data collection terminal is used to provide massive and diversified raw map road data for the high-precision map generation server;

The high-precision map generation server is used to generate high-precision dynamic maps according to the massive and diversified raw data of map roads provided by the map data collection terminal;

The high-precision map generation server includes:

The map data processing module is used to extract road features from the original data of map roads, adopt layered design for all road features, splicing the same road information, first complete the road feature data fusion at the road feature level, and finally complete it at the image level. Image stitching, the result of stitching generates the overhead projection map of the road;

The map generation visualization module is used to mark road information on the overhead projection map. The marked road information and the overhead projection map together form the map data of the high-precision map, and the map data is visually edited to generate a high-precision dynamic map.

Further, the map data collection terminal includes:

The image data collection module is used to collect a large amount of original road image data of the map;

The road condition data collection module is used to collect massive amounts of original road condition data on maps.

Further, the map data processing module includes:

The image data preprocessing unit is used to perform image correction, image coordinate transformation, and image projection transformation on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after the preprocessing can be aligned;

The road feature extraction unit is used to identify the road dynamic features from the original road image data of the map; carry out refined feature modeling according to the existing navigation map to obtain the static features of the road; through the data screening and verification of the original road condition data of the map, extract the corresponding road semi-static characteristics and road semi-dynamic characteristics;

The road feature design unit is used for hierarchical design of all road features in terms of content. The first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road. dynamic features;

The multi-dimensional road information fusion and splicing unit is used to splicing the same road information. First, the road feature data fusion is completed at the road feature level, and finally the preprocessed image is completed at the image level. .

Further, the road feature extraction unit includes:

The dynamic feature extraction sub-unit is used to establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map;

The static feature extraction subunit is used to perform refined feature modeling according to the existing navigation map to obtain road static features;

The semi-static feature extraction sub-unit is used to extract the corresponding semi-static features of the road by performing data screening and verification on the original road condition data of the map;

The semi-dynamic feature extraction sub-unit is used to extract the corresponding semi-dynamic features of the road by performing data screening and verification on the original road condition data of the map.

Further, the multi-dimensional road information fusion and splicing unit includes:

The road feature fusion and splicing sub-unit is used to splicing all the extracted road features according to feature layers, first splicing static features, and finally merging dynamic features;

The image stitching subunit is used to fuse and stitch the preprocessed images into a basic overhead projection image.

Further, the map generation visualization module includes:

The geographic information labeling unit is used to label road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map;

Geographic information calibration and verification unit, used to calibrate and verify the marked road information;

The map generation unit is used to visually edit the map data to generate a high-precision dynamic map.

Further, the image data acquisition module is a camera, and the road condition data acquisition module includes a GPS, a temperature and humidity sensor, and a water accumulation sensor.

Further, the camera is a security monitoring camera used in urban traffic and a camera mounted on a smart street light pole of the Internet of Things; GPS is a GPS base station in the current smart city, and the temperature and humidity sensor and the water accumulation sensor are on the smart street light pole of the Internet of Things. The integrated temperature and humidity sensor and water accumulation sensor; the original road image data of the map and the original road condition data of the map are transmitted to the high-precision map generation server through the public gateway module configured on the IoT smart street light pole.

A high-precision dynamic map data processing method based on roadside sensors, comprising:

S1. Collect massive amounts of map original road image data and map original road condition data;

S2. Perform image correction, image coordinate transformation, and image projection transformation preprocessing on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after preprocessing can be aligned;

S3. Establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map;

Perform refined feature modeling according to the existing navigation map to obtain road static features;

Extract the corresponding semi-static and semi-dynamic features of the road by filtering and verifying the original road condition data of the map;

S4. Carry out hierarchical design for all road features in terms of content, the first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road dynamic feature;

S5, splicing the same road information, first completing the road feature data fusion at the road feature level, and finally completing the image splicing of the preprocessed image at the image level, and the result of the splicing generates an overhead projection map of the road;

S6, marking the road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map;

S7, calibrating and verifying the marked road information;

S8. Visually edit the map data to generate a high-precision dynamic map.

Further, in step S1, a camera is used to collect a large amount of original road image data of the map, and GPS, a temperature and humidity sensor, and a water accumulation sensor are used to collect the original road condition data of the map;

The cameras are the security monitoring cameras used in urban traffic and the cameras mounted on the IoT smart street light poles; GPS is the GPS base station in the current smart city, and the temperature and humidity sensors and the water accumulation sensors are the temperature and humidity sensors integrated on the IoT smart street light poles. Humidity sensor, water accumulation sensor; map original road image data and map original road condition data are transmitted to the high-precision map generation server through the public gateway module configured on the IoT smart street light pole.

Compared with the prior art, the beneficial effects of the present invention are as follows:

1), the camera that collects roadside images of the present invention is deployed on roadside infrastructure (street light poles, elevated poles), and is more effective in real time than the map data collected by the vehicle-mounted camera;

2), the present invention can collect multi-dimensional dynamic information of high-precision maps in real time (road area water, humidity, weather, street signs, etc.) through various sensor terminal equipment mounted on the integrated light pole of the Internet of Things, and these information only need the background After a simple verification process, it can be published to the application platform of high-precision maps and serve the field of intelligent transportation in real time;

3), the map data collection in the present invention makes full use of the current smart city infrastructure (smart street lights, roadside cameras, GPS base stations), by sharing the roadside sensor method of the smart city, and the automatic collection and uploading method of map data, It can reduce the cost of generating high-precision maps in many ways;

4) The present invention can realize the rapid update of high-precision dynamic maps through the layered design of road features, the fusion and splicing of feature layers, and the storage of data formats that support incremental update.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

1 is a schematic structural diagram of a high-precision dynamic map data processing system based on roadside sensors of the present invention;

2 is a topology diagram of a high-precision dynamic map data processing system based on roadside sensors of the present invention;

Fig. 3 is the working flow chart of the high-precision dynamic map data processing system based on the roadside sensor of the present invention;

Fig. 4 is the camera deployment diagram of the present invention to collect map;

FIG. 5 is a flow chart of the method for processing high-precision dynamic map data based on roadside sensors according to the present invention.

Detailed ways

In order to make the above objects, features and advantages of the present invention more clearly understood, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be pointed out that the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, those of ordinary skill in the art can obtain all the Other embodiments fall within the protection scope of the present invention.

Example 1

As shown in Figure 1, the present invention provides a high-precision dynamic map data processing system based on roadside sensors, including:

The map data collection terminal is used to provide massive and diversified raw map road data for the high-precision map generation server;

The high-precision map generation server is used to generate high-precision dynamic maps according to the massive and diversified raw data of map roads provided by the map data collection terminal;

The high-precision map generation server includes:

The map data processing module is used to extract road features from the original data of map roads, adopt layered design for all road features, splicing the same road information, first complete the road feature data fusion at the road feature level, and finally complete it at the image level. Image stitching, the result of stitching generates the overhead projection map of the road;

The map generation visualization module is used to mark road information on the overhead projection map. The marked road information and the overhead projection map together form the map data of the high-precision map, and the map data is visually edited to generate a high-precision dynamic map.

The map data collection terminal includes:

The image data collection module is used to collect a large amount of original road image data of the map;

The road condition data collection module is used to collect massive amounts of original road condition data on maps.

The image data acquisition module is a camera, and the road condition data acquisition module includes GPS, a temperature and humidity sensor, a water accumulation sensor, and the like.

The cameras are the security monitoring cameras used in urban traffic and the cameras mounted on the IoT smart street light poles; GPS is the GPS base station in the current smart city, and the temperature and humidity sensors and the water accumulation sensors are the temperature and humidity sensors integrated on the IoT smart street light poles. Humidity sensor, water sensor.

The map data processing module includes:

The image data preprocessing unit is used to perform image correction, image coordinate transformation, and image projection transformation on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after the preprocessing can be aligned;

The road feature extraction unit is used to identify the road dynamic features from the original road image data of the map; carry out refined feature modeling according to the existing navigation map to obtain the static features of the road; through the data screening and verification of the original road condition data of the map, extract the corresponding road semi-static characteristics and road semi-dynamic characteristics;

The road feature design unit is used for hierarchical design of all road features in terms of content. The first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road. dynamic features;

The multi-dimensional road information fusion and splicing unit is used to splicing the same road information. First, the road feature data fusion is completed at the road feature level, and finally the preprocessed image is completed at the image level. .

The road feature extraction unit includes:

The dynamic feature extraction sub-unit is used to establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map;

The static feature extraction subunit is used to perform refined feature modeling according to the existing navigation map to obtain road static features;

The semi-static feature extraction sub-unit is used to extract the corresponding semi-static features of the road by performing data screening and verification on the original road condition data of the map;

The semi-dynamic feature extraction sub-unit is used to extract the corresponding semi-dynamic features of the road by performing data screening and verification on the original road condition data of the map.

The multi-dimensional road information fusion and splicing unit includes:

The road feature fusion and splicing sub-unit is used to splicing all the extracted road features according to feature layers, first splicing static features, and finally merging dynamic features;

The image stitching subunit is used to fuse and stitch the preprocessed images into a basic overhead projection image.

The map generation visualization module includes:

The geographic information labeling unit is used to label road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map;

Geographic information calibration and verification unit, used to calibrate and verify the marked road information;

The map generation unit is used to visually edit the map data to generate a high-precision dynamic map.

As shown in Figure 2, in the image acquisition in the map data acquisition terminal, the cameras used are the security surveillance cameras used in urban traffic and the cameras mounted on the smart street light poles. As the construction of smart cities becomes more and more mature, the coverage of cameras becomes more and more dense. Each camera collects road attribute images within the monitoring range corresponding to the area where it is located, and uploads it to the high-precision map generation server. As shown in Figure 4, the cameras that collect road images are mounted on roadside facilities such as street light poles and overhead poles. For other road conditions attributes that cannot be obtained by the camera (such as data such as temperature and humidity, road area water, etc.), the temperature and humidity sensors and water accumulation sensors integrated on the IoT integrated light pole are used to obtain the data. The sensing equipment terminal can be flexibly selected according to the situation, and the data is transmitted to the high-precision map generation server through the public gateway module configured on the light pole.

The road feature design unit, according to the requirements of automatic driving for high-precision maps, models and designs the road features from the content, and designs the features of the road network in layers. The first layer is the static feature of the road, and the static feature information includes: the position of the road line lane line, the position of the traffic signal and the traffic sign; the basic information such as the road ID number, shape, slope, width, etc. The second layer is the semi-static characteristics of the road: traffic rule information (such as tidal road sections, etc.), road construction information, weather information in a wide range of areas (rain and snow weather), etc.; the third layer is the semi-dynamic information of road characteristics: traffic accident location, The location of traffic congestion, the location of traffic water, the location of road potholes, the location of road obstacles, etc.; the fourth layer of road features is dynamic information: the current coordinates and motion trajectories of pedestrians, cars, bicycles, motorcycles and other objects. The result of the road feature design unit is to abstract all the designed road feature information into data entities and objects in the high-precision map. The principle adopted in the design of road features is the demand for high-precision map content and accuracy for different levels of autonomous driving. With the continuous improvement of autonomous driving levels, road features need to be continuously refined and enriched.

The image data preprocessing unit first needs to perform a series of preprocessing such as image correction, image coordinate transformation, and image projection transformation for the uploaded original road image data of the map; for the processed image data in the same coordinate system, deep learning, Image recognition and other related AI technologies extract road dynamic features, such as using pre-trained deep learning models to identify dynamic road features (signal lights, pedestrians, cars, bicycles, motorcycles) in high-precision maps.

Multi-dimensional road information fusion stitching is to stitch the same road information in the high-precision map. First, the road feature data fusion analysis is completed at the road feature level, and finally the image stitching is completed at the image level. The visual editing of annotation and road feature database generates high-precision dynamic maps.

The high-precision dynamic map data processing system based on roadside sensors proposed by the invention integrates infrastructures such as smart street lamps, roadside cameras, communication base stations, etc., and can effectively reduce the cost of generating high-precision maps. In the present invention, roadside sensing data processing, AI-based road feature extraction and analysis, and multi-dimensional road feature analysis are integrated on the server or cloud, and the sensing device terminal only needs to be responsible for data collection and uploading data to the service backend. The sensor device terminal does not have excessive storage and computing resource requirements, which realizes the feasibility of the entire map generation scheme.

As shown in Figure 3, the workflow of the high-precision dynamic map data processing system based on roadside sensors of the present invention is as follows:

1) Coverage of roadside cameras. The camera used for collecting roadside image data in the present invention is installed and deployed on the street lamp poles on both sides of the road. With the intelligent development of the city, the infrastructure construction related to the street lamps on both sides of the municipal road is more and more perfect, and the distance between the street lamps on the municipal road is generally around 30 meters. As shown in Figure 4, assuming that the angle of view of the camera used to collect roadside images is θ degrees, the height of the street light pole is m meters, and the range of images that can be collected by the camera is Meter. Through calculation, as long as the camera angle of view θ for collecting roadside images is greater than The range l of the camera collection area is greater than 30 meters, and the road data collected based on the roadside camera will not be omitted.

2) The camera data for the road sections with stable road characteristics can be uploaded to the server for processing in a certain period. For road sections with complex and changeable road characteristics or high demand for active traffic safety (such as crossroads), it is necessary to shorten the time required for uploading images to the server. Period to ensure that the acquired characteristic data of dynamic road sections is accurate and effective.

3), urban IoT integrated light pole coverage. The IoT integrated light pole can integrate IoT charging piles, intelligent lighting, surveillance cameras, micro weather stations, electronic bulletin screens, alarm buttons and other functions. The present invention uses the sensing equipment on the integrated light pole to collect road condition data, and the data of each sensing equipment on the integrated light pole is transmitted to the map data processing server through the public gateway module configured on the light pole, and the map data processing server can Remote control, remote management, data acquisition, data analysis, message release, fault monitoring, etc. of each part are carried out through a unified management platform.

4), image data space transformation processing. Since the deployment of urban street lights is arranged in different positions, the deployment and arrangement of the cameras mounted on the street light poles are also different, and there is no guarantee that all the cameras that collect road image data are on the same plane. The original image is processed by coordinate transformation and projection transformation. After processing, the overlapping areas of images collected by cameras at adjacent positions can be aligned, and the aligned images are convenient for subsequent road feature extraction and splicing and map image fusion splicing.

5), road feature entity design. According to the content and accuracy requirements of high-precision maps for autonomous driving, a layered design is adopted for all road features. The first layer is the static features of the road. The static feature information includes: the position of the road line lane line, the position of the traffic signal and the traffic sign, the road ID number, the shape, the slope, the width and other basic information. The second layer is the semi-static characteristics of the road: traffic rule information (such as tidal road sections, etc.), road construction information, weather information in a wide range of areas (rain and snow weather), etc.; the third layer is the semi-dynamic information of road characteristics: traffic accident location, The location of traffic congestion, the location of traffic water, the location of road potholes, the location of road obstacles, etc.; the fourth layer of road features is dynamic information: the current coordinates and motion trajectories of pedestrians, cars, bicycles, motorcycles and other objects. The result of the road feature design module is to abstract all the designed road feature information into data entities and objects in the high-precision map, as shown in Figure 4 and Figure 5.

6), road feature extraction. The first layer of static road features can be refined and modeled according to the existing navigation map, and each feature entity can be obtained, and finally stored in the table of the database using an appropriate data structure. The road features of the second and third layers can be simply preprocessed (data screening and verification) on the road condition data reported by the map data collection terminal, and finally the corresponding semi-static and semi-dynamic features can be extracted.

7) The fourth layer of road dynamic features uses deep learning and image recognition related AI technologies to identify pedestrians, cars, bicycles, and motorcycles on the road from map image data. The map data processing server first performs deep learning training on a large-scale image library containing object categories such as pedestrians, cars, bicycles, and motorcycles, and then builds a model that can accurately identify road dynamic features. When the map data collection end uploads the current road image, it can quickly and effectively identify the characteristics of the traffic objects contained in the current road, and update these identified dynamic road characteristics to the database corresponding to the fourth-layer features of the high-precision map, which can quickly and truly feedback Current road conditions.

8) After the extraction of road features, all features are stored in a database in the form of a complete data entity, and the database supports incremental update of road features.

9) Multi-dimensional road feature fusion and splicing. Based on the accurate road feature database, a series of operations are performed to complete the fusion and splicing of road features. The fusion and splicing of road features splices all the road features extracted in the above 6) and 7) according to feature layers, first splicing static features, and finally merging dynamic features.

10) Fusion and splicing the roadside map after image space transformation in 5) into a basic overhead projection map, and finally label road information on this image, including: road edges, lane lines, intersection points, etc. The information together form the map data of the high-precision map;

11) Visually edit the data in the map database, and observe the lane-level high-precision road map and road dynamic characteristics, in which the data accuracy can reach centimeter-level. Once the map database is updated with dynamic road features, it can be quickly reflected on the visual map.

Example 2

As shown in FIG. 5 , the present invention also provides a high-precision dynamic map data processing method based on roadside sensors, including:

S1. Collect massive amounts of map original road image data and map original road condition data;

S2. Perform image correction, image coordinate transformation, and image projection transformation preprocessing on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after preprocessing can be aligned;

S3. Establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map;

Perform refined feature modeling according to the existing navigation map to obtain road static features;

Extract the corresponding semi-static and semi-dynamic features of the road by filtering and verifying the original road condition data of the map;

S4. Carry out hierarchical design for all road features in terms of content, the first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road dynamic feature;

S5, splicing the same road information, first completing the road feature data fusion at the road feature level, and finally completing the image splicing of the preprocessed image at the image level, and the result of the splicing generates an overhead projection map of the road;

S6, marking the road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map;

S7, calibrating and verifying the marked road information;

S8. Visually edit the map data to generate a high-precision dynamic map.

In step S1, a camera is used to collect a large amount of original road image data of the map, and GPS, a temperature and humidity sensor, and a water accumulation sensor are used to collect the original road condition data of the map;

The cameras are the security monitoring cameras used in urban traffic and the cameras mounted on the IoT smart street light poles; GPS is the GPS base station in the current smart city, and the temperature and humidity sensors and the water accumulation sensors are the temperature and humidity sensors integrated on the IoT smart street light poles. Humidity sensor, water sensor.

Compared with the prior art, the beneficial effects of the present invention are as follows:

1), the camera that collects roadside images of the present invention is deployed on roadside infrastructure (street light poles, elevated poles), and is more effective in real time than the map data collected by the vehicle-mounted camera;

2), the present invention can collect multi-dimensional dynamic information of high-precision maps in real time (road area water, humidity, weather, street signs, etc.) through various sensor terminal equipment mounted on the integrated light pole of the Internet of Things, and these information only need the background After a simple verification process, it can be published to the application platform of high-precision maps and serve the field of intelligent transportation in real time;

3), the map data collection in the present invention makes full use of the current smart city infrastructure (smart street lights, roadside cameras, GPS base stations), by sharing the roadside sensor method of the smart city, and the automatic collection and uploading method of map data, It can reduce the cost of generating high-precision maps in many ways;

4) The present invention can realize the rapid update of high-precision dynamic maps through the layered design of road features, the fusion and splicing of feature layers, and the storage of data formats that support incremental updating.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as limiting the scope of the patent of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can also be made, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be subject to the appended claims.

Claims (8)

1. a high-precision dynamic map data processing system based on roadside sensor, is characterized in that, comprises: The map data collection terminal is used to provide massive and diversified raw map road data for the high-precision map generation server; The high-precision map generation server is used to generate high-precision dynamic maps according to the massive and diversified raw data of map roads provided by the map data collection terminal; The high-precision map generation server includes: The map data processing module is used to extract road features from the original data of map roads, adopt layered design for all road features, splicing the same road information, first complete the road feature data fusion at the road feature level, and finally complete it at the image level. Image stitching, the result of stitching generates the overhead projection map of the road; The map generation visualization module is used to mark road information on the overhead projection map. The marked road information and the overhead projection map together form the map data of the high-precision map, and the map data is visually edited to generate a high-precision dynamic map; The map data processing module includes: The image data preprocessing unit is used to perform image correction, image coordinate transformation, and image projection transformation on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after the preprocessing can be aligned; The road feature extraction unit is used to identify the road dynamic features from the original road image data of the map; carry out refined feature modeling according to the existing navigation map to obtain the static features of the road; through the data screening and verification of the original road condition data of the map, extract the corresponding road semi-static characteristics and road semi-dynamic characteristics; The road feature design unit is used for hierarchical design of all road features in terms of content. The first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road. dynamic features; The multi-dimensional road information fusion and splicing unit is used to splicing the same road information. First, the road feature data fusion is completed at the road feature level, and finally the preprocessed image is completed at the image level. ; The road feature extraction unit includes: The dynamic feature extraction sub-unit is used to establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map; The static feature extraction subunit is used to perform refined feature modeling according to the existing navigation map to obtain road static features; The semi-static feature extraction sub-unit is used to extract the corresponding semi-static features of the road by performing data screening and verification on the original road condition data of the map; The semi-dynamic feature extraction sub-unit is used to extract the corresponding semi-dynamic features of the road by performing data screening and verification on the original road condition data of the map. 2. The high-precision dynamic map data processing system based on roadside sensors according to claim 1, wherein the map data collection terminal comprises: The image data collection module is used to collect a large amount of original road image data of the map; The road condition data collection module is used to collect massive amounts of original road condition data on maps. 3. The high-precision dynamic map data processing system based on roadside sensors according to claim 1, wherein the multi-dimensional road information fusion and splicing unit comprises: The road feature fusion and splicing sub-unit is used to splicing all the extracted road features according to feature layers, first splicing static features, and finally merging dynamic features; The image stitching subunit is used to fuse and stitch the preprocessed images into a basic overhead projection image. 4. The high-precision dynamic map data processing system based on roadside sensors according to claim 1, wherein the map generation visualization module comprises: The geographic information labeling unit is used to label road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map; Geographic information calibration and verification unit, used to calibrate and verify the marked road information; The map generation unit is used to visually edit the map data to generate a high-precision dynamic map. 5 . The high-precision dynamic map data processing system based on roadside sensors according to claim 2 , wherein the image data acquisition module is a camera, and the road condition data acquisition module comprises GPS, temperature and humidity sensors, stagnant water sensor. 6. The high-precision dynamic map data processing system based on roadside sensors according to claim 5, wherein the camera is a security monitoring camera used in urban traffic and a camera mounted on an Internet of Things smart street light pole; GPS It is the GPS base station in the current smart city. The temperature and humidity sensor and the water accumulation sensor are the temperature and humidity sensor and the water accumulation sensor integrated on the IoT smart street light pole; the original road image data of the map and the original road condition data of the map pass through the IoT smart street light pole. The public gateway module configured above is transmitted to the high-precision map generation server. 7. A high-precision dynamic map data processing method based on roadside sensors, characterized in that, comprising: S1. Collect massive amounts of map original road image data and map original road condition data; S2. Perform image correction, image coordinate transformation, and image projection transformation preprocessing on the original road image data of the map, and the overlapping areas of the images collected by the image data acquisition modules at adjacent positions after preprocessing can be aligned; S3. Establish a deep learning road dynamic feature recognition model by using AI technologies related to deep learning and image recognition, and use the deep learning road dynamic feature recognition model to identify road dynamic features from the original road image data of the map; Perform refined feature modeling according to the existing navigation map to obtain road static features; Extract the corresponding semi-static and semi-dynamic features of the road by filtering and verifying the original road condition data of the map; S4. Carry out hierarchical design for all road features in terms of content, the first layer is the road static feature, the second layer is the road semi-static feature, the third layer is the road semi-dynamic feature, and the fourth layer is the road dynamic feature; S5, splicing the same road information, first completing the road feature data fusion at the road feature level, and finally completing the image splicing of the preprocessed image at the image level, and the result of the splicing generates an overhead projection map of the road; S6, marking the road information on the overhead projection map, and the marked road information and the overhead projection map together form the map data of the high-precision map; S7, calibrating and verifying the marked road information; S8. Visually edit the map data to generate a high-precision dynamic map. 8. The high-precision dynamic map data processing method based on roadside sensors according to claim 7, characterized in that, in step S1, a camera is used to collect a large amount of original road image data of maps, GPS, temperature and humidity sensors, stagnant water are used The sensor collects the original road condition data of the map; The cameras are the security monitoring cameras used in urban traffic and the cameras mounted on the IoT smart street light poles; GPS is the GPS base station in the current smart city, and the temperature and humidity sensors and the water accumulation sensors are the temperature and humidity sensors integrated on the IoT smart street light poles. Humidity sensor, water accumulation sensor; map original road image data and map original road condition data are transmitted to the high-precision map generation server through the public gateway module configured on the IoT smart street light pole.