JP2019105789A - Road structure data generator, road structure database - Google Patents

Abstract

To create a road network including facilities on the side of the road, to realize driving assistance including navigation and automatic driving with high accuracy and efficiency, and to predict accurately how a leading vehicle or an oncoming vehicle moves at the next moment by being used in a driving support device.SOLUTION: In addition to lane-level road network information, it automatically generates information on roadside areas where vehicles may enter and leave. As a database, places are stored where vehicle right turn, left turn and stop may occur at places other than intersections. Driving behavior is comprehensively extracted by aggregating and analyzing a large amount of vehicle data that has run on roads in lane units. For this reason, it is possible to extract connection information with the area accessible from the lane, and an appropriate entry/exit route.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a road structure data generation device that generates information on a roadside area in which a vehicle may enter or leave, and a road structure database used when executing processing representing a connection with a road network in the vicinity. .

  In recent years, a system that creates road information using probe data has been proposed as a driving support device.

  Patent Document 1 describes that information of onboard sensors is collected to automatically generate a lane level road network. In Patent Document 1, unlike the conventional navigation map composed of nodes and links, the center line is detected in units of lanes, and a connection relationship for each lane is automatically generated at an intersection. The autonomous driving system designs a vehicle route to a destination based on such information.

  Patent Document 2 describes that a new road is detected by extracting a vehicle trajectory that does not exist in existing map information for the purpose of determining a new road.

  In Non-Patent Document 1, it is described to use the concept of a potential map in order to make a route plan that is human-like. That is, Non-Patent Document 1 integrates an actualized risk potential generated based on map information, a preceding vehicle trajectory, etc. and a potential risk potential for a traffic participant who has acquired empirically, and a potential map around the vehicle. It is described to generate and based on that, to generate an optimal path.

JP, 2015-4814, A JP 2014-130529 A

“Chunzhao Guo, et al.,“ Human-like Behavior Generation for Intelligent Vehicles in Urban Environment based on Hybrid Potential Map, ”Proc. Of 2017 IEEE Intelligent Vehicles Symp. (IV), pp. 197-203, 2017.”

  However, there are many stores, parking lots and the like (hereinafter referred to as facilities) on the side of the road mainly on ordinary roads, and vehicles often use the facilities as a destination.

  The facilities beside the road are provided with an entrance (including a taxiway) connecting to the surrounding road. In some cases, a plurality of gateways exist. The vehicle decelerates and stops near the entrance, or turns to the left or right.

  For example, in patent document 2, since the positioning result of GPS (Global Positioning System) is used as a clue, to extract the vehicle position change in the vicinity of the entrance on the roadside precisely and make the position a database, the amount of information and control There is a lack of features.

  The present invention is capable of generating a road network including facilities to the side of a road, and obtaining a road structure data generating device capable of realizing driving assistance including navigation and automatic driving with high accuracy and efficiency. Is the purpose.

  In addition to the above object, it is an object of the present invention to obtain a road structure database which can accurately predict how a preceding vehicle or an oncoming vehicle moves at the next moment by being used in a driving assistance device.

  The road structure data generation device according to the present invention includes an acquisition unit for acquiring operation information related to the operation of the vehicle from an on-board sensor mounted on the vehicle, operation information acquired by the acquisition unit, and road structure information known in advance. On the basis of the basic information generation means for generating basic information to be a basis for recognizing the road structure, and analysis means for analyzing trajectory information at the time of traveling of the vehicle on the basis of the basic information generated by the basic information generation means. And classification means for classifying locus information analyzed by the analysis means in lane units of the traveling path, and statistical processing of the traveling locus of the vehicle, the end point of the road boundary, the road bifurcation point, and the starting point of the entrance to the facility A lane connectable to the attention point is selected based on extraction means for extracting the attention point to be included and trajectory information classified by the classification means, and a travel route connecting the selected lane and the attention point It has produced a road connection information generating means for, the.

  According to the present invention, the acquisition means acquires driving information related to the driving of the vehicle from the on-board sensor mounted on the vehicle. The base information generation means generates base information to be a base for recognizing the road structure based on the driving information acquired by the acquisition means and the road structure information known in advance.

  Based on the base information, trajectory information during traveling of the vehicle is analyzed, and the analyzed trajectory information is classified in lane units of the traveling path (analysis means, classification means).

  On the other hand, the extraction means extracts an attention point including the end point of the road boundary, the road bifurcation point, and the entrance point of the entrance to the facility by the statistical processing of the traveling locus of the vehicle.

  The road connection information generation means selects a lane connectable to the attention point based on the trajectory information classified by the classification means, and generates a travel route connecting the selected lane and the attention point.

  As a result, it is possible to generate a road network including facilities to the side of the road, and drive assistance including navigation and automatic driving can be realized with high accuracy and efficiency.

  In the present invention, the vehicle further includes an exceptional running judging means for judging whether or not there is an exceptional running including at least one of lane change of the running road and obstacle avoidance behavior for each of the locus information analyzed by the analyzing means. The traveling locus determined to be exceptional traveling is excluded from the objects to be classified by the classification means.

  It is determined whether or not there is exceptional traveling including at least one of lane change of the traveling road and obstacle avoidance behavior for each of the trajectory information, and the traveling trajectory determined as exceptional traveling is excluded from the classification targets. Thereby, more accurate road structure data can be generated.

  In the present invention, the base information generated by the base information generation means is based on the driving information and the road structure information, the traveling information of the vehicle, the vehicle position information, the road information of the traveling road, and the peripheral information of the traveling road. It is characterized by including.

  More detailed driving assistance is possible by generating, based on the driving information and the road structure information, the driving information of the vehicle, the position information of the vehicle, the road information of the traveling road, and the peripheral information of the traveling road as the base information. It becomes.

  In the present invention, the attention point is a facility having an area where the vehicle can travel and which is an entrance to the road, and in the facility, a route network through which the vehicle can travel, a parking space, and a boundary of a travelable range. And space information generating means for generating space information in the facility including the landmark.

  Driving support can be performed in a traveling space in a facility different from the road.

  In the present invention, the space information generation unit generates the space information by acquiring traveling guide information of a vehicle related to the facility and at least one piece of public information including a city map and an aerial photograph to which data is distributed. The information processing system further includes editing means for performing editing including verification, updating, and correction of the space information in the facility generated by the means.

  By sequentially editing space information that changes frequently, such as a parking space in the facility, optimal driving support can be performed in the facility.

  For example, when performing editing including verification, update, and correction of spatial information in a facility, it is necessary to acquire facility information. From the viewpoint of acquiring information of this facility, open map data such as Open Sreet Map can be used. In addition, if the facility is an outdoor parking lot, it is possible to image-process an aerial photograph and acquire information on a parking position, a travel route, and a doorway. That is, it is possible to create facility information from public information as well as travel guidance information distributed from the facility.

  In the present invention, the apparatus further includes moving object information detection means for detecting moving object information from an infrastructure sensor including a surveillance camera installed along the traveling path, and the moving object information detected by the moving object information detection means It is characterized in that it is added to the base information passing through the traveling path of the area where the infrastructure sensor is installed.

  By utilizing the infrastructure sensor, it is possible to increase the amount of information, which is an element for creating a road structure database, to each level.

  In the road structure database according to the present invention, the driving support device guides the connection relationship between the locus information of the vehicle traveling on the road and the attention point including the end point of the road boundary, the road bifurcation point, and the entrance point to the facility. And selecting a lane to which the point of interest and the locus information can be connected, and storing road structure data including a travel route connecting the selected lane and the point of attention .

  According to the present invention, the lanes to which the point of interest and the locus information can be connected are selected, and the road structure data including the travel route connecting the selected lane and the point of interest is stored as a road structure database. By using it, it is possible to accurately predict how the preceding vehicle or oncoming vehicle will move at the next moment.

  As described above, according to the present invention, it is possible to create a road network including facilities to the side of the road, and drive assistance including navigation and automatic driving can be realized with high accuracy and efficiency.

  Further, in addition to the above effects, by being used for the driving support device, it is possible to accurately predict how the preceding vehicle and the oncoming vehicle move at the next moment.

It is a schematic block diagram of the road structure database production | generation apparatus which concerns on 1st Embodiment. It is a perspective view showing a foundation information source concerning a 1st embodiment. It is a front view which shows the road boundary which concerns on 1st Embodiment. It is a front view which shows an example of the surrounding road image which concerns on 1st Embodiment. It is a top view which shows the target for estimating the own vehicle position which concerns on 1st Embodiment. It is a top view which shows an example of the exception data of the traveling locus which concerns on 1st Embodiment. It is the top view which visualized distribution of the path and speed concerning a 1st embodiment, (A) is a top view around the intersection, (B) visualizes distribution of the path and speed on the top view of the intersection It is a distribution chart displayed. It is a top view which shows the confluence | merging path | route and branch path | route which concern on 1st Embodiment, (A) is a one-lane one-lane road, (B) shows a one-side two-lane road, (C) a road with a central divider. It is a control flowchart which shows the road structure database production | generation procedure which concerns on 1st Embodiment. It is a schematic block diagram of the driving | operation action database production | generation apparatus which concerns on 2nd Embodiment. It is a perspective view showing a foundation information source concerning a 2nd embodiment. (A)-(D) are perspective views which show the installation state of the camera applied as an infrastructure sensor which concerns on 2nd Embodiment. It is a perspective view which concerns on 2nd Embodiment and shows the positional relationship of an own vehicle and a camera when estimating an own vehicle position by a camera based on triangulation. It is a top view of the road expressed with the road structure database finally generated concerning a 2nd embodiment. It is a control flowchart which shows the road structure database production | generation procedure which concerns on 2nd Embodiment.

  In the present invention, connection information to a roadside facility (store, parking lot, etc.), that is, roadside area other than a road where a vehicle can enter and exit is automatically added to a road network map at lane level (lane unit). Automatically generate a database of

  The present invention will be described in detail in the following first and second embodiments, but in the first and second embodiments, a plurality of vehicles traveling on the same road A large amount of data collected / detected at the time of traveling is collected, and statistical processing is performed to automatically generate connection information of road network and roadside area (connection relation to a roadside area and a connection to a lane).

  In the first embodiment and the second embodiment, data processing on the vehicle side and data processing on the center server side having functions as a base information generating unit 14 and a roadside area information generating unit 16 described later It does not distinguish clearly. The distribution of data processing on the vehicle side / center server side can be considered various patterns by the capacity of the on-board computer and the restriction of communication with the center server, but its functional layout greatly affects the processing content of the whole system. It is because there is not.

"First embodiment"
FIG. 1 is a schematic configuration diagram of a road structure data generation device 10 according to the first embodiment.

  As shown in FIG. 1, the road structure data generation device 10 includes an on-vehicle sensor group 12, a base information generation unit 14, and a roadside area information generation unit 16. Further, as a storage medium for storing information, a road structure storage unit 18 which is a road structure database is connected to the base information generation unit 14 and the roadside area information generation unit 16.

(On-board sensor group 12)
The in-vehicle sensor group 12 used in the first embodiment includes an environment recognition sensor 24, a vehicle motion sensor 26, and a positioning sensor 30.

  The environment recognition sensor 24 assumes a camera or a laser radar, and acquires an object existing around the vehicle as image data or obtains distance information.

  By processing the image data and distance information acquired by the environment recognition sensor 24, a moving object existing in the vicinity of the vehicle such as a traveling lane or road boundary or a leading vehicle or an oncoming vehicle is detected.

  The observation range depends on the specification of the environment recognition sensor 24. Here, it is assumed that the host vehicle monitors a forward area ahead.

  The vehicle motion sensor 26 is configured of a gyro, a wheel speed sensor, and the like, and measures the speed and the yaw rate of the vehicle at a predetermined cycle.

  The positioning sensor 30 mainly assumes GPS (Global Positioning System), and measures the absolute position, the moving direction, and the speed of the vehicle.

(Basic information generation unit 14)
Based on the detection (measurement) information received from each sensor of the on-vehicle sensor group 12, the base information generation unit 14 generates base information such as road information and trajectory information required to generate a road structure database.

  In the first embodiment, as shown in FIG. 2, the various data are obtained from the in-vehicle sensor group 12 (see FIG. 1) mounted on each of the plurality of vehicles 15, the center server CS (first In the embodiment, it is assumed that the information is transmitted to the system including the base information generation unit 14 and the roadside area information generation unit 16.

  As shown in FIG. 1, the base information generation unit 14 includes a road information extraction unit 32, a travel information extraction unit 34, a surrounding information extraction unit 36, and a vehicle position estimation unit 38.

  The road information extraction unit 32 detects the presence of a physical road boundary using the observation information of the environment recognition sensor 24. The detection target is a guardrail or a curb.

  The road information extraction unit 32 may detect all areas where there are no obstacles such as guardrails or curbs that prevent the passage of the vehicle, or even if only areas having intervals greater than the length that the vehicle can pass are extracted. Good. An example of a road boundary is shown in FIG.

  In the road structure image 50 shown in FIG. 3A, the guardrail 56 is laid along the curb 54 provided at one end of the straight road 52, and the position where the guardrail 56 is interrupted is the cross road 58. There is.

  Further, in the road structure image 60 shown in FIG. 3B, there is no guardrail along the straight road 62, and a T-shaped road 64 exists. As a feature used to estimate the vehicle position, road surface paint such as the lane mark 66 and the stop line 68 may be extracted as shown in FIG. 3 (B).

  As shown in FIG. 1, the traveling information extraction unit 34 detects the traveling locus of the vehicle including the right turn and the left turn. In addition, the traveling information extraction unit 34 extracts the stopped place and the decelerated place based on the speed information.

  The peripheral information extraction unit 36 detects moving objects existing in the vicinity of the vehicle such as a preceding vehicle or an oncoming vehicle using the environment recognition sensor 24 and extracts a locus by tracking in the time direction while considering the vehicle position Do.

  In addition, the peripheral information extraction unit 36 also detects vehicles that exit the lane from the roadside area and vehicles that are stopping to exit (the side of the vehicle can be observed when seen from the host vehicle).

  FIG. 4 shows an example of the peripheral information images 70A, 70B and 70C. In the peripheral information images 70A and 70B of FIGS. 4A and 4B, the vehicle 72 stops in an attempt to leave the alley onto the main road 74. The peripheral information image 70C of FIG. 4C is a peripheral information image showing a state where the vehicle 72 is about to enter the alley from the main road 74.

  Although not shown in FIG. 4, it is preferable to detect landmarks such as signs, signals, signs, and cylindrical objects as features used when the vehicle position estimation unit 38 described later estimates the vehicle position. .

  The vehicle position estimation unit 38 draws map data from positioning information from the GPS, and collates the landmarks registered in the map with the landmarks detected by the peripheral information extraction unit 36 and the road information extraction unit 32. Estimate the vehicle position in the map data. From this result, it can be known which lane the vehicle is traveling on and at which position it is.

  FIG. 5 shows an image of vehicle position estimation.

  As shown in FIG. 5, the position of the vehicle 76 can be estimated by the directions of the lane marks 78 and 80, the U-turn prohibition sign 82 and the speed limit sign 84 which are visible from the vehicle 76.

(Roadside area information generation unit 16)
As shown in FIG. 1, the roadside area information generation unit 16 aggregates base information generated in units of vehicles and generates connection information between lanes and the roadside area.

  The roadside area information generation unit 16 includes a data classification / statistical processing unit 40 and a road connection information generation unit 42.

  The data classification / statistics processing unit 40 receives the basic information from the basic information generation unit 14.

  The data classification / statistics processing unit 40 classifies the foundation information for each road block set in advance. The road block may be divided in link units from the intersection to the intersection, or the links may be further subdivided on a predetermined basis.

  Further, the data classification / statistics processing unit 40 distributes the base information into lane units according to the result of the vehicle position estimation. At this time, it is desirable to remove, as exceptional data, traveling data as if the vehicle traveled across the lane mark from the vehicle track and the result of the vehicle position estimation.

  FIG. 6 shows a case where the vehicle 92 (refer to the solid line) travels on the lane 88A on the left side in the traveling direction on the road 90 where there are a plurality of lanes (lanes 88A, 88B,...) Partitioned by the lane mark 86. It is a top view which shows the locus | trajectory (refer dotted line) when noticing the presence of the two-wheeled vehicle 94 ahead, and drive | working over the lane mark 86. FIG.

  The traveling data when such a lane change traveling is performed is removed as exceptional data.

  In addition, the data classification / statistics processing unit 40 integrates the trajectories of the own vehicle and the surrounding vehicles, and generates a spatial distribution of trajectories that exit to roadside areas other than roads where vehicles can enter or exit, or merge from roadside areas.

  For example, in the form of a two-dimensional occupancy grid map, the distribution of trajectories is expressed in the form of a probability or a histogram so that regions where many trajectories overlap can be extracted. A threshold may be provided for the probability value and the frequency of the histogram, and a region exceeding a predetermined threshold may be extracted as a region with high reliability. Furthermore, the detection results of the physical road boundary are superimposed, and it is verified whether or not the information is consistent with the information of the trajectory (reliability determination).

  FIG. 7 is an example of visualization of the distribution of paths and velocities.

  FIG. 7A is a plan view of an intersection X (within a dotted line frame) at which two roads R1 and R2 intersect and a roadside area 96 present in the vicinity thereof. It is FIG. 7 (B) which visualized and displayed distribution of a path | route and speed on the top view of the intersection X of this FIG. 7 (A). Although a facility 98 (a parking lot, an area to which a vehicle can enter) exists beyond the roadside area 96, in the first embodiment, the driving support is not performed up to the facility 98.

  In FIG. 7B, strip-like lines L having different densities represent paths on the image of the intersection X, and the density difference of the lines L represents velocity. The higher the concentration, the higher the speed.

  According to FIG. 7B, it is possible to visualize the spatial distribution of trajectories that exit to the roadside area 96 other than the roads R1 and R2 where the vehicle can enter and exit, or merge from the roadside area 96 to the roads R1 and R2. In addition, when visualizing, not only identification by density, but different colors may be used.

  Although the intersection X is shown in FIG. 7, locations where the vehicle may turn right, turn left, and stop outside the intersection X may be stored as a database.

  As shown in FIG. 1, the road connection information generation unit 42 generates connection information with the entrance / exit (position) to the roadside area and each lane. When determining the position of the entrance and exit point and the branch point from the lane, the speed distribution of the vehicle is used.

  Vehicles are often decelerated or stopped near the entrance and exit or lane junction for safety confirmation. These points can be determined from the result of statistical processing by superimposing the velocity distribution on the trajectory distribution in such a case.

  Also, with regard to connection information, for example, a single path without a central separation band and generally an entrance and exit requires connection to two opposite lanes. If there is a median, it will only enter from lanes near the roadside area. In addition, in the case of a road with multiple lanes, connection information is required such that the vehicle approaches from the nearest lane in both directions (see FIG. 8).

  FIG. 8A is a plan view showing the merging path 100 and the branch path 102 in a road with one lane on one side.

  FIG. 8 (B) is a plan view showing the merging path 100 and the branch path 102 in a road with two lanes on one side.

  FIG. 8C is a plan view showing the merging path 100 and the branching path 102 in a road with two lanes on one side where the central separation band 104 exists.

  Although the conditions of the merging route 100 and the branching route 102 depend on the number of lanes and the road structure as shown in FIG. 8A to FIG. 8C, the connection relationship is naturally made by collecting the trajectory information. It can be extracted.

  Finally, connection information is defined by the branch position for each lane, the point at the entrance of the roadside area, and the route connecting them. As shown in FIG. 8, the merging path 100 and the branching path 102 may be smoothly connected by curves such as splines, or may be expressed merely as indicating connection directions. The merging from the entrance of the roadside area to the lane is also set based on the collected trajectory distribution.

  As shown in FIG. 1, the road connection information generation unit 42 is connected to the road structure storage unit 18 via the roadside information addition / update unit 44.

  The roadside information addition / update unit 44 adds the generated connection information (branch position, entrance and exit, and route information) to the road structure storage unit 18. If there is already generated information, and it overlaps with it, it can be updated by updating to the newer one or by statistically combining with the old information.

  The road structure storage unit 18 adds the generated connection information (branch position, entrance and exit, route information) to the road structure storage unit 18. That is, the road structure storage unit 18 has already generated information, and the road structure storage unit 18 updates new information if the information overlaps, or statistically combines old information and new information. It can be updated by

  The operation of the first embodiment will be described below with reference to the flowchart of FIG.

  In step 150, the data detected by the in-vehicle sensor group 12 is acquired, and then, the process proceeds to step 152, and the base information is generated based on the data (sensor data) detected by the in-vehicle sensor group 12.

  In the next step 153, the generated infrastructure information is transmitted to the center server CS, aggregated, and the process proceeds to step 154.

  At step 154, the aggregated traveling data is classified. As classification items, each sensor of the on-vehicle sensor group 12, vehicle type, road type, lane unit, etc. may be mentioned.

  In the next step 156, the reliability of the collected information is determined by statistical processing, and the process proceeds to step 158.

  In step 158, as a result of the determination in step 156, it is determined whether the reliability is equal to or more than a threshold.

  If it is determined in step 158 that the reliability is less than the threshold (reliability <threshold), it is determined that the aggregated data amount is insufficient, and the process proceeds to step 153.

  If it is determined in step 158 that the reliability is equal to or higher than the threshold (reliability し き い 値 threshold), it is determined that the aggregated data amount is sufficient, and the process proceeds to step 160.

  In step 160, connection information between the road and the roadside area is generated, and then the process proceeds to step 162 to automatically generate space information in the roadside area. That is, the space arrangement of the landmarks present in the roadside area, the stop position of the vehicle, and the like are generated.

  In the next step 164, changes in road connection and in-area space information are determined, and the process proceeds to step 166.

  In step 166, as a result of the determination in step 164, it is determined whether there is a change in road connection or in-area space information.

  If it is determined in step 166 that there is no change, it is determined that the road structure already stored need not be updated, and the process proceeds to step 153.

  When it is determined in step 166 that there is a change, it is determined that the road structure already stored needs to be updated, and the process proceeds to step 168.

  In step 168, the information including the change point is stored in the road structure storage unit 18, and the routine ends.

  According to the first embodiment, it is possible to automatically generate information on the roadside area where vehicles may enter and leave, in addition to lane network information on the lane level.

  In addition, it is possible to store as a database the places where there is a possibility that the right turn, left turn and stop of the vehicle occur other than the intersection.

  In addition, driving behavior can be comprehensively extracted by aggregating and analyzing a large amount of vehicle data that has run on the road in lane units, connection information with an area accessible from the lane, and an appropriate entry / exit route Can be extracted.

  Furthermore, since the road structure database is generated by aggregating information detected by the on-vehicle sensor group 12 mounted on a vehicle traveling on a road and generated by automatic processing on a computer, data processing by the operator can be suppressed and cost can be reduced. The system can be realized.

"Second embodiment"
The second embodiment of the present invention will be described below. The same reference numerals as in the first embodiment denote the same parts in the second embodiment, and a description thereof will be omitted.

  The features of the second embodiment are as follows, with respect to the first embodiment described above.

(Feature a) Point of using information of infrastructure sensor installed near the road such as surveillance camera (Feature b) Point of combining available external map database information (Feature c) Not only connection information with roadside area The point at which space information in the roadside area is generated FIG. 10 is a schematic configuration diagram of a driving behavior data generation device 10A according to the second embodiment of the present invention.

  As shown in FIG. 10, the road structure data generation device 10A includes an onboard sensor group 12, a base information generation unit 14, and a roadside area information generation unit 16A. In the second embodiment, the infrastructure sensor 106 is used.

  In the second embodiment, as shown in FIG. 11, the various data are obtained from the in-vehicle sensor group 12 (see FIG. 10) mounted on each of the plurality of vehicles 15 and the infrastructure sensor 106 as a center server. It is assumed that it is transmitted to CS (in the second embodiment, a system including the base information generating unit 14 and the roadside area information generating unit 16A).

  The on-vehicle sensor group 12 used in the second embodiment includes an environment recognition sensor 24, a vehicle motion sensor 26, and a positioning sensor 30.

  The base information generation unit 14 includes a road information extraction unit 32, a travel information extraction unit 34, a surrounding information extraction unit 36, and a vehicle position estimation unit 38.

  In addition, the roadside area information generation unit 16A includes a data classification / statistical processing unit 40, a road connection information generation unit 42, a map information acquisition unit 108, and an in-area space information generation unit 110.

  Furthermore, as a storage medium for storing information, the road structure storage unit 18 is connected to the base information generation unit 14 and the roadside area information generation unit 16, and the external map storage unit 112 is connected to the roadside area information generation unit 16. There is.

(Use of infrastructure sensor 106)
Here, in the first embodiment described above, the information of the on-vehicle sensor group 12 mounted on the vehicle is collected. On the other hand, in the second embodiment, it is one feature to utilize the infrastructure sensor 106 as an information source different from the in-vehicle sensor group 12.

  As the infrastructure sensor 106, as shown in FIG. 12A to FIG. 12D, cameras 106A, 106B, 106C, and 106D corresponding to respective applications are applicable.

  That is, cameras 106A and 106B for monitoring (see FIG. 12A and FIG. 12B) are often installed around a road centering on an intersection. In addition, a camera 106C (see FIG. 12C) or the like for monitoring is often installed in a store or a building beside the road for the purpose of crime prevention and the like. Furthermore, in the parking lot beside the road, a camera 106D (see FIG. 12D) for monitoring a parked vehicle and the like are often installed.

  With the spread of the Internet of Things (IoT), there is a high possibility that detection (measurement) data by the infrastructure sensor 106 represented by the cameras 106A, 106B, 106C, 106D will be available in the future. The information of the fixed infrastructure sensor 106 is very useful in observing road conditions.

  As shown in FIG. 10, the infrastructure sensor 106 is connected to the moving object information extraction unit 111, and the image data captured by the infrastructure sensor 60 is the moving object information extraction unit together with the position / attitude information of the infrastructure sensor 106. It is sent to 111.

  The cameras 106A, 106B, 106C, and 106D (see FIG. 12) applied as the infrastructure sensor 106 are basically fixed, and the absolute position does not change, so the position information may be set only once at the time of installation. When the orientation (imaging optical axis) of the cameras 106A, 106B, 106C, and 106D can control the optical axis direction by pan, tilt, zoom, etc., parameters including optical axis direction information at the time of imaging are associated with the image. Is preferred.

  The moving object information extraction unit 111 extracts a moving vehicle from, for example, images captured by the cameras 106A, 106B, 106C, and 106D according to the following procedures (1) and (2), and tracks the position in the time direction. To detect the traveling track.

  (Procedure 1) First, candidates for moving object regions are detected in an image using background difference and time difference. A search area is set in the screen so as to include candidate areas, and a vehicle is detected using pattern recognition processing.

  (Procedure 2) Assuming that the installation height and the line of sight direction of the cameras 106A, 106B, 106C, 106D are known, vehicle position at the lane level is estimated using means such as the principle of triangulation (see FIG. 13) Do.

  As shown in FIG. 13, the vehicle V is located on the road surface, and any one of the cameras 106A, 106B, 106C, 106D (for example, the camera 106A) immediately above the position separated from the vehicle V by the distance d. ) Is installed. It is assumed that the camera 106A is installed at a height h from the road surface.

  The position of the vehicle V can be specified because the direction when the vehicle V is imaged by the camera 106A is the angle θ with respect to the perpendicular line of the camera 106A.

  In addition, if the camera 106A has sufficient resolution, lane marks on the road surface can also be detected, so the traveling lane can also be determined.

  As shown in FIG. 10, the moving object information extraction unit 111 is connected to the roadside area information generation unit 16A.

  The moving object information extraction unit 111 transmits the trajectory information of the moving object detected in the above-described procedure to the roadside area information generation unit 16A, and integrates the information from the in-vehicle sensor group 12 for use.

  It is also effective to merge (integrate) the information of the infrastructure sensor 106 by setting an appropriate weight with respect to the information of the in-vehicle sensor group 12. The weight may be changed according to the position of the imaging target. By integrating the information of the infrastructure sensor 106, the accuracy and reliability of the generated information can be improved, and the change point can be detected quickly.

(Use of external map information)
In the case of a large store or public facility, the entrance to the parking lot may be described in the general map data.

  In addition, information in the parking lot, such as the arrangement of the parking space, may be provided by communication.

  In the future, when passing through the gate at the entrance of the parking lot, a service may be considered in which information such as a map in the parking lot and an empty space is provided to the vehicle by wireless communication. If such public information (external map or aerial photograph) can be obtained, it is possible to process and integrate the information to compensate for the lack of in-area spatial information or to verify the generated structure.

  That is, open map data such as Open Street Map, and if the facility is an outdoor parking lot, aerial images can be image-processed to obtain information on parking positions, travel routes, and entrances and exits. That is, it is possible to create facility information from public information as well as travel guidance information distributed from the facility.

  Therefore, in the second embodiment, the map information acquisition unit 108 is added to the roadside area information generation unit 16A, and data conversion for integrating the information provided from the external map storage unit 112 is performed.

  For example, the position of the entrance of the roadside area, the travelable route in the area, the parking space, the arrangement of characteristic structures (concrete pillars, road surface paint, signs / signs such as guidance and regulation, artificial markers, etc.) The above information is read, converted into a desired format, and sent to the road connection information generation unit 42 and the in-area space information generation unit 110.

(Regional space information generation on the roadside)
In addition to the connection between the road network in the road connection information generation unit 42 and the entrance of the roadside area, the in-area space information generation unit 110 automatically generates space information required for route guidance in the roadside area.

  Technically, it uses the method called SLAM (Simultaneous Localization and Mapping) as a base.

  If GPS positioning is possible outdoors or the like, absolute position information is also added.

  For example, in patent document 1, although a road surface image is produced | generated in order to extract a lane mark, the runway of a road side area | region does not necessarily have a suitable mark.

  For example, when a camera is used as the in-vehicle sensor group 12, in a method called V-SLAM (Visual SLAM), salient feature points are extracted from the image, the same feature points are searched and collated before and after movement, and triangulation is performed. The three-dimensional position of the feature point can be estimated by the principle of

  If the three-dimensional positions of many feature points are stored, the position of the vehicle can be estimated even at different times by collating them with the current observation results.

  However, storing many feature points requires a large amount of memory. In addition, since extraction of feature points is affected by changes in lighting, robustness against time and weather is a problem.

  Therefore, the feature points should be minimized and only spatially and temporally stable. Instead, characteristic structures are extracted and their spatial arrangement is made into a database as described in the above-mentioned "Utilization of external map information".

  For example, boundary lines of runways in the area, road surface paint such as parking space and temporary stop, concrete pillars and poles, signboards and signs such as guidance and regulation, artificial markers, and the like.

  These objects can be automatically extracted from camera images using image processing techniques such as pattern recognition. Information collected from multiple vehicles is collated in the vicinity area, and statistical processing is used to calculate the reliability of observation. The three-dimensional position of the object satisfying the predetermined reliability is registered as spatial information.

  According to the second embodiment, the roadside area 48 and the road network 47 in the finally generated road structure database (see FIG. 7B) shown in FIG. 7 in the first embodiment described above. Can be generated in more detail.

  An example of the road structure database finally generated is shown in FIG.

  FIG. 14 (A) is a general road map 46, and FIG. 14 (B) is a detailed map generated specifically for a specific area from the road map 46 in the road structure data generation device 10. It is 46A.

  The roadside area 48 is set in addition to the lane level road network 47 (indicated by the dotted line in FIG. 14B), and the solid line of the connection relationship line 49 of the roadside area 48 and the road network 47 (FIG. 14B). ) Is added. Further, in the roadside area 48, a landmark 48A, a stop position 48B, and a travelable route 48C are automatically generated.

  The operation of the second embodiment will be described below with reference to the flowchart of FIG. Note that “A” is added to the end of the reference numerals for the same processing steps as the predetermined procedure (see FIG. 9) in the first embodiment described above.

  In step 150A, the data detected by the in-vehicle sensor group 12 is acquired, and then, the process proceeds to step 152A, and the base information is generated based on the data (sensor data) detected by the in-vehicle sensor group 12.

  In the next step 153A, the generated base information is transmitted to the center server CS, aggregated, and the process proceeds to step 154A.

  At step 154A, the aggregated traveling data is classified. As classification items, each sensor of the on-vehicle sensor group 12, vehicle type, road type, lane unit, etc. may be mentioned.

  Here, in the second embodiment, data detected by the infrastructure sensor 106 is added as the base information to be classified.

  That is, in step 170, data detected by the infrastructure sensor 106 is acquired, and then, the process proceeds to step 172 to generate base information including the trajectory of a moving object.

  In the next step 174, the generated infrastructure information is sent to the center server CS, aggregated, and the process moves to step 154A.

  As a result, in the center server CS, the base information based on the data detected by the in-vehicle sensor group 12 and the base information based on the data detected by the infrastructure sensor 106 are aggregated.

  In the next step 156A, the reliability of the information collected by statistical processing is determined, and the process proceeds to step 158A.

  In step 158A, as a result of the determination in step 156A, it is determined whether the reliability is equal to or greater than a threshold.

  If it is determined in step 158A that the reliability is less than the threshold (reliability <threshold), it is determined that the aggregated data amount is insufficient, and the process proceeds to step 153A.

  If it is determined in step 158A that the reliability is equal to or higher than the threshold (reliability し き い 値 threshold), it is determined that the aggregated data amount is sufficient, and the process proceeds to step 160A.

  Here, in the second embodiment, when connection information between a road and a roadside area is generated, related information obtained from existing map data is added. The related information obtained from the map data includes road connection information and parking position information.

  That is, in step 176, the related information is acquired from the existing map data, and the process proceeds to step 160A.

  In step 160A, connection information between the road and the roadside area is generated. At this time, in the second embodiment, since road connection information, parking position information, and the like are acquired from existing map data in addition to basic information on roads, connection information between a road and a roadside area is generated. Accuracy can be improved.

  In the next step 162A, spatial information in the roadside area is automatically generated. That is, the space arrangement of the landmarks present in the roadside area, the stop position of the vehicle, and the like are generated.

  In the next step 164A, a change in road connection and space information in the area is determined, and the process proceeds to step 166A.

  In step 166A, as a result of the determination in step 164A, it is determined whether there is a change in road connection or in-area space information.

  If it is determined in step 166A that there is no change, it is determined that the road structure already stored need not be updated, and the process proceeds to step 153A.

  If it is determined in step 166A that there is a change, it is determined that the road structure that has already been stored needs to be updated, and the process proceeds to step 168A.

  In step 168A, the information including the change point is stored in the road structure storage unit 18, and the routine ends.

  According to the second embodiment, the road structure database collects information of the in-vehicle sensor group 12 mounted on the vehicle traveling on the road, and the infrastructure sensors 106 such as the cameras 106A to 106D shown in FIG. Generated by automatic processing above. The data processing by the operator can be suppressed and the system can be realized at low cost.

  In the first embodiment and the second embodiment, when there is a change in road connection or space information in the area, it is stored in the road structure storage unit 18 (step 166 of FIG. 9 → 168 and the process of step 166A → 168A in FIG. 15), the data in the road structure storage unit 18 is old (when a certain amount of time has elapsed since the previous update), and the reliability is higher than a predetermined level regardless of time If it is determined that the road connection has not been made, the road connection and the space information in the area may be updated without any change. Furthermore, in combination with a case where there is a change, updating may be performed periodically, or may be performed by manual instruction operation.

DESCRIPTION OF SYMBOLS 10 Road structure data generation apparatus 12 Vehicle-mounted sensor group 14 Foundation information generation part 16 Roadside area information generation part 18 Road structure storage part 24 Environment recognition sensor 26 Vehicle motion sensor 30 Positioning sensor 32 Road information extraction part 34 Driving information extraction part 36 Surrounding information Extraction unit 38 Vehicle position estimation unit 40 Data classification / statistical processing unit 42 Road connection information generation unit 44 Roadside information addition / update unit 46 Road map 46A Detailed map 48A Landmark 48B Stop position 48C Travelable route 50 Road structure image 52 Straight line Road 54 curb 56 guardrail 60 road structure image 62 straight road 64 T-junction 66 lane mark 68 stop line 70A, 70B, 70C peripheral information image 72 vehicle 74 main road 76 own car 78, 80 lane mark 82 U turn prohibited sign 84 speed Limit sign 86 lanemer 88A, 88B Lane 90 Road 92 Vehicle 94 Two-wheeled vehicle 96 Two-way area 98 Facility 100 Merged route 102 Branched route 104 Central division R1, R2 Road X intersection (Second embodiment)
10A Road Structure Data Generation Device 16A Roadside Area Information Generation Unit 106 Infrastructure Sensors 106A, 106B, 106C, 106D Cameras 108 Map Information Acquisition Unit 110 Space Information Generation Unit in Region 111 Moving Object Information Extraction Unit 112 External Map Storage Unit

Claims (7)

Acquisition means for acquiring driving information related to driving of the vehicle from an on-board sensor mounted on the vehicle;
Basic information generation means for generating basic information to be a basis for recognizing a road structure based on driving information acquired by the acquisition means and road structure information known in advance;
An analysis unit that analyzes trajectory information when the vehicle is traveling based on the base information generated by the base information generation unit;
Classification means for classifying track information analyzed by the analysis means in lane units of a traveling path;
Extracting means for extracting an attention point including an end point of a road boundary, a road bifurcation point, and a starting point of an entrance to a facility by statistical processing of a traveling track of a vehicle;
Road connection information generation means for selecting a lane connectable to the attention point based on the trajectory information classified by the classification means, and generating a travel route connecting the selected lane and the attention point;
Road structure data generator having: The vehicle further includes exceptional running determination means for determining whether or not there is an exceptional running including at least one of lane change of the running path and obstacle avoidance behavior for each of the trajectory information analyzed by the analyzing means.
The travel locus determined to be exceptional travel is excluded from the classification targets by the classification unit,
The road structure data generation device according to claim 1, characterized in that: The basic information generated by the basic information generation means is
Based on the driving information and the road structure information, the information includes traveling information of the vehicle, position information of the vehicle, road information of the traveling path, and peripheral information of the traveling path.
The road structure data generation device according to claim 1 or 2, characterized in that:   The attention point is a facility having an area where the vehicle can travel, which is an entrance to the road, and a route network on which the vehicle can travel, a parking space, a boundary of a travelable range, and a landmark in the facility. The road structure data generation apparatus according to any one of claims 1 to 3, further comprising space information generation means for generating space information in the facility that includes the information. The space information generation means
The space information in the facility generated by the space information generation means is verified by acquiring traveling guide information of a vehicle related to the facility, and at least one of public information including a city map and an aerial photograph to which data are distributed 5. The road structure data generating device according to claim 4, further comprising an editing unit that performs editing including update, correction. And moving object information detection means for detecting moving object information from an infrastructure sensor including a surveillance camera installed along the traveling path,
6. The moving object information detected by the moving object information detection means is added to the base information passing through a traveling path of the area where the infrastructure sensor is installed. Road structure data generation device according to item 1. The driving support device is used when performing guidance of a connection relationship between trajectory information of a vehicle traveling on a road, and an attention point including an end point of a boundary of the road, a road branch point, and a base point of an entrance to a facility.
A road structure database storing road structure data including a travel route connecting the selected lane and the attention point by selecting a lane to which the attention point and the locus information can be connected.