JP4577827B2 - Next road prediction device for traveling vehicle - Google Patents

Description

  The present invention relates to a next road prediction device for a traveling vehicle that predicts a next road in a traveling direction at an intersection where the vehicle travels.

  In a navigation device that provides guidance according to a route to a destination, the current location of the vehicle is detected, a map around the current location is displayed, and features along the intersection and the route are guided. In the detection of the current location in this case, map matching of roads based on dead reckoning navigation and map data using various sensor data such as vehicle speed, G (acceleration), gyroscope, and GPS is performed.

In the guidance of guidance intersections that turn left and right, especially when multiple intersections continue close to each other, if the accuracy of the current location detection is low, due to errors with the current location in route guidance, the passing intersection in front of the guidance intersection or beyond It is easy to cause troubles that cause you to turn off the route by mistakenly turning right and left at the intersection. As one of the countermeasures, when approaching an intersection in the route, guidance display of arrows such as “straight forward”, “right turn”, “left turn” and the like and information on each lane at a plurality of intersections including passing intersections are displayed. Thus, proposals have been made to eliminate driver anxiety (see, for example, Patent Documents 1 and 2).
JP 2000-251197 A JP 2003-240581 A

  However, for guidance intersections and passing intersections that turn left and right, the conventional system recognizes the actual driving lane and driver's habits even when guiding the driving direction and lane information using arrows, and adapts to the driving. The guidance is not provided, but it is unilateral, and guidance corresponding to the actual driving situation is not provided.

  The present invention solves the above-described problems, and recognizes actual driving lanes and driver behavior and predicts the next road in the traveling direction at an intersection so that driving guidance and control can be performed.

Therefore, the present invention provides a next road prediction device for a traveling vehicle that predicts a next road in a traveling direction at an intersection where the vehicle travels, recognizes an image acquired by a camera, and lane information on a road on which the current vehicle travels Image recognition means for acquiring a next road, a next road prediction processing means for predicting a next road in the traveling direction at the intersection based on the lane information, and a driver at the intersection.
Learning processing means for recognizing and learning the selection of the next road in the traveling direction , wherein the next road prediction processing means predicts the next road in the traveling direction at the intersection based on learning by the learning processing means. And

The learning processing means is characterized by accumulating a comparison between the predicted next road in the advancing direction at the intersection and the actual driving result to obtain and output a confidence level.

  According to the present invention, while recognizing road information while traveling, the next road in the traveling direction at the intersection is recognized, the driving behavior of the driver at that time is learned, and the next road in the traveling direction at the intersection is determined based on the learning result. Since the prediction is performed, it is possible to perform appropriate travel guidance and control based on the prediction.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an embodiment of a next road prediction device for a traveling vehicle according to the present invention. 4 is a current location information management unit, 7 is a database, 8 is an image recognition device, 9 is a driver input information management unit, 21 Indicates a learning processing unit, 22 indicates a next road prediction processing unit, and 23 indicates an output device.

  In FIG. 1, the current location management unit 4 manages guess information, macro information, and micro information, and acquires current location information from the macro information and micro information. The estimated information is an estimated trajectory obtained by calculating the azimuth and distance of the vehicle from various sensor data such as vehicle speed, G (acceleration), gyroscope, and GPS. Further, the macro information which is various sensor information is the estimated trajectory and the database 7. Based on the map matching process using the road map of the road, in addition to that, using the new device information, database information, etc., to manage which road is running more accurately, road on / off ( Whether or not it is on the road), road type, area information, confidence (information freshness, reliability, accuracy, degree of certainty from the update time), matching road, coordinates, route on / off (on route) Or not). Micro information performs feature determination based on image recognition, further performs lane determination based on image recognition, driver input information, optical beacon information, and estimation information, and uses position determination and lane determination results to perform position matching, macro The current position of information is corrected, and the total number of lanes, lane positions, and in-lane positions generated as a result of micro matching.

  The database 7 stores data related to various map data, road data, feature types belonging to each road, feature positions, and confidence levels. The image recognizing device 8 captures an image ahead of the traveling direction of the vehicle with a camera, recognizes paint information on the road, recognizes the number of lanes, its own lane position, the position in the lane, the number of lanes, the lane increase / decrease direction, and the road shoulder information. Further, the crossing state, the paint information, and the confidence level are acquired as events, and further, the feature recognition process is performed, and the recognition result, the feature type, the feature position, the confidence level, and the like are acquired. The driver input information management unit 9 detects a steering angle associated with the driver's steering wheel operation with a steering angle sensor, detects a right / left turn instruction with a direction indicator, and acquires steering information and turn signal information as events.

  The learning processing unit 21 tracks current location information and collates road information, lane information, travel information, steering information, turn signal information, and road data in the database 7 acquired by the image recognition device 8 and the driver input information management unit 9. The next road prediction processing unit 22 recognizes and learns the selection of the next road in the traveling direction at the intersection, and the road recognition information, the lane information, and the travel information are detected by the image recognition device 8 and the driver input information management unit 9 based on the current location information. Information, steering information, and turn signal information are acquired, and the next road in the traveling direction at the intersection is predicted based on the learning result by the learning processing unit 21. The output device 23 outputs the prediction information.

  According to the configuration of the above embodiment, by grasping the current position of the vehicle on the road with high accuracy, and grasping the detailed operation state of the vehicle, the driver's habits, which direction the driver wants to go thereafter, In other words, it is possible to predict the next road in the direction of travel at the intersection where the vehicle travels, and by predicting the next road, voice guidance, warning (vehicle control), driving assistance and control based on the prediction can be performed. .

  FIG. 2 is a diagram illustrating an example of next road prediction based on road information and learning, FIG. 3 is a diagram illustrating an example of next road prediction based on vehicle speed information, and FIG. 4 is a diagram illustrating an example of next road prediction based on driver input information. FIG. 5 is a diagram illustrating an example of the next road prediction process.

  In the next road prediction device for a traveling vehicle according to the present embodiment, in the next road prediction in the traveling direction at the intersection of the traveling vehicles, the next road / lane data that can be transferred to each lane of each road included in the map data is prepared, and the camera To acquire road information, road information including arrow paint on the road, information on the direction of travel including information on the direction of travel, information on the direction sign of the direction of travel, etc. The selection result is learned and the probability of the next road is obtained and used.

  For example, at the intersection of a road with 2 lanes, if a right-turn lane or a left-turn lane is set by arrow paint as shown in Fig. 2, when the arrow paint is acquired by image recognition, it usually travels on that lane The next road of the vehicle is the right turn destination road and the left turn destination road. Therefore, by predicting the next road in the direction of travel as you approach the intersection, you can give voice guidance to decelerate, turn right, turn left, and if you don't have a turn signal, turn the turn signal can do. However, by recognizing and learning the selection of the next road in the direction of travel at the intersection, for example, when driving on a right turn lane and the next road is different from the prediction, there are many straight ahead roads, the right turn ahead as the next road It is possible to reduce the speed by outputting a voice guidance that calls attention by reflecting the degree of confidence on the road and the straight ahead road, and controlling the brake as vehicle control.

  In addition, as shown in FIGS. 3 and 4, at the intersection where the right-turn lane or the left-turn lane is not set by arrow paint, the next road can be predicted by detecting the current position, the vehicle speed change, and the turn signal. it can. When detecting a change in vehicle speed, for example, if driving at a constant speed, the next road will be a straight road, and if decelerating, the next road will be a left turn road (for left lane driving) and a right turn road (for right lane driving). It becomes easier to predict. Further, when the driver's brake timing and curve speed are learned and the turn signal is detected in the same manner, the next road can be predicted with higher accuracy by learning the driver's turn signal timing.

  In the next road prediction process, for example, as shown in FIG. 5, first, arrow paint is recognized (step S11), the own lane position is specified (step S12), and the next road is predicted (step S13). Then, it is determined whether or not the predicted accuracy rate is larger than the criterion (step S14). If the predicted accuracy rate is larger than the criterion, an application such as guidance is started (step S15), and the determination process in step S14 is performed. If the predicted correct answer rate is not greater than the criterion, the next lane travel is performed as it is (step S16), and the predicted correct / incorrect learning is performed (step S17).

  Next, the vehicle current location information management device used in the present embodiment will be described. FIG. 6 is a diagram showing an embodiment of a vehicle current location information management device, where 1 is a micro matching processing unit, 2 is a macro matching processing unit, 3 is a dead reckoning processing unit, 4 is a current location management unit, and 5 is vehicle control. Device, 6 is a vehicle information processing device, 7 is a database, 8 is an image recognition device, 9 is a driver input information management unit, 11 is a position collation and correction unit, 12 is a feature determination unit, 13 is a micro matching result unit, 14 Indicates a lane determination unit.

  In FIG. 6, dead reckoning processing unit 3 is a module that calculates a vehicle's azimuth and distance from various sensor data such as vehicle speed, G (acceleration), gyroscope, GPS, etc. to obtain an estimated trajectory, and estimates the current vehicle position. The estimated trajectory and various sensor information are managed as estimated information and sent to the current location management unit 4. The vehicle position obtained in this way does not match the road on the map data because it does not match the map data by obtaining the estimated trajectory by directly using the sensor data such as the vehicle speed, G, gyroscope, and GPS. .

  The macro matching processing unit 2 is based on the map matching process using the conventional estimated trajectory obtained by the dead reckoning processing unit 3 and the road map of the database 7, and in addition, using new device information, database information, etc. It is a module that manages which road is running more accurately, road on / off (whether it is on the road), road type, area information, confidence (information freshness from the update time, Information such as reliability, accuracy, degree of probability), matching road, coordinates, route on / off (whether or not it is on the route) is managed as macro information and sent to the current location management unit 4.

  The micro-matching processing unit 1 is a module that manages the detailed position of the vehicle in a narrow area. The micro-matching processing unit 1 mainly performs feature determination based on image recognition, and further performs image recognition, driver input information, optical beacon information, and estimation information. Lane determination based on the feature, and using the result of the feature determination and lane determination to perform position matching and correction of the current position of the macro information, as well as the total number of lanes, the lane position, and the position in the lane of the micro matching result as micro information It is generated, managed, and sent to the current location management unit 4.

  The feature information includes information on various structures belonging to the road. For example, signals, footbridges, road signs, street lights, poles / electric poles, guardrails, shoulders / pedestrian steps, median strips, manholes on the road, paint (crossing) Paint on sidewalks, bicycle crossing roads, stop lines, turn left / right, lanes, center lines, etc.). The feature information has the feature type, feature position, update timing, reliability of the information itself, etc. as confidence level (information freshness, reliability, accuracy, accuracy from the update timing). Thus, when a feature is recognized as a result of image recognition, the current position can be corrected with high accuracy based on the position of the feature.

  The present location management unit 4 manages the micro information obtained from the micro matching processing unit 1, the macro information obtained from the macro matching processing unit 2, and the guess information obtained from the dead reckoning processing unit 3, and appropriately performs the micro matching processing on the information. The current location information is generated from the macro information and the micro information and is sent to the vehicle control device 5 and the vehicle information processing device 6 while being passed to the unit 1 and the macro matching processing unit 2.

  The vehicle control device 5 performs vehicle travel control such as cornering brake control and speed control based on the current location information acquired from the current location management unit 4, and the vehicle information processing device 6 performs the current location acquired from the current location management unit 4. It is a navigation device, VICS, or other application device that guides a route by guiding each intersection, feature, etc. to a destination based on information. The database 7 stores data on various road data, feature types belonging to each road, feature positions, and confidence levels.

  The image recognizing device 8 captures an image ahead of the traveling direction of the vehicle with a camera, recognizes paint information on the road, recognizes the number of lanes, its own lane position, the position in the lane, the number of lanes, the lane increase / decrease direction, and the shoulder information. , The straddle state, the paint information, and the confidence level are sent as events to the micro-matching processing unit 1, and further, the specified feature recognition process is performed in response to the request from the micro-matching processing unit 1, and the recognition result, The object type, feature position, confidence level, etc. are sent to the micro-matching processing unit 1.

  The driver input information management unit 9 detects the steering angle associated with the driver's steering wheel operation with a steering angle sensor, detects a right / left turn instruction with a direction indicator, and sends the steering information and turn signal information as events to the micro matching processing unit 1. To do.

  The micro matching processing unit 1, the macro matching processing unit 2, and the dead reckoning processing unit 3 will be described in further detail. FIG. 7 is a diagram illustrating a configuration example of the macro matching processing unit, and FIG. 8 is a diagram illustrating a configuration example of the dead reckoning processing unit.

  As shown in FIG. 6, the micro matching processing unit 1 includes a position matching & correction unit 11, a feature determination unit 12, a micro matching result unit 13, and a lane determination unit 14. The feature determination unit 12 searches the feature from the database 7 based on the current position of the macro information, requests the image recognition device 8 to perform image recognition of the feature based on the feature type, feature position, and confidence, The distance to the feature is specified based on the recognition result acquired from the image recognition device 8, the feature type, the feature position, and the confidence level. The lane determination unit 14 includes optical beacon information of the vehicle information processing device 6, inference information of the current location management unit 4, steering information and winker information events from the driver input information management unit 9, recognition lane number from the image recognition device 8, Among them, own lane position, lane position (right or left in lane), lane increase / decrease number, lane increase / decrease direction, shoulder information (existence etc.), straddle status (whether it straddles lane / white line, etc.), paint information (Straight or right / left turn, pedestrian crossing, bicycle crossing, etc.), identify the lane position of the vehicle and the position in the lane based on the event of confidence, and the result of the determination is the position matching & correction unit 11 and the micro matching result unit Pass to 13.

  The position collation & correction unit 11 includes the feature recognition information of the feature determination unit 12 obtained by the feature determination, the lane position of the lane determination unit 14 obtained by the lane determination, the in-lane position, and the current position of the macro information. In the case of mismatch, the current position of the macro information is corrected to the current position calculated based on the feature recognition information. The micro matching result unit 13 passes the micro information such as the total number of lanes, the lane position, the position in the lane, and the confidence level of the lane determination unit 14 obtained by the lane determination to the current location management unit 4.

  For example, when the recognition information of a manhole is obtained as a feature, the position of the manhole and the distance to the manhole are specified from the recognition information, so the current position of the vehicle and the current macro information in the direction of travel determined from the distance When there is a mismatch due to collation with the position, the current position of the macro information can be corrected. Also, if the position of the manhole in the road width (left and right) direction, not the direction of travel, does not match due to collation between the current position of the vehicle and the current position of the macro information due to either the left or right or near the center, etc. The current position of the macro information can be corrected.

  Similarly, if the lane determination indicates that the vehicle is traveling on a road with two lanes, for example, the lane position is closer to the shoulder and the position in the lane moves to the right from the center of the lane, further lanes on the center line side , The current position of the macro information can be corrected if there is a mismatch by comparing the current position of the host vehicle with the current position of the macro information. Also, if there is a change in the number of lanes, for example, when a new right turn lane increases on the right side or the number of lanes decreases from 3 to 2, or from 2 to 1, the macro is determined by determining the matching of the positions. The current position of information can be corrected.

  The macro matching processing unit 2 includes a macro matching result unit 21, a micro position correction reflection unit 22, a road determination unit 23, and a macro shape comparison unit 24 as shown in FIG. The macro shape comparison unit 24 compares the estimated trajectory of the estimated information managed by the current location management unit 4 with the road information in the database 7 and the map road shape based on the confidence level, and performs map matching, and the road determination unit 23 Determines on / off of the road at the current position, and determines the road at the current position. The micro position correction reflecting unit 22 reflects the correction information of the current position by the macro matching processing unit 1 of the macro information on the current position of the macro shape comparison unit 24 and the current position of the road determination unit 23. The macro matching result unit 21 sends coordinates, road type, area information, road on / off, matching road, route on / off, and confidence level as macro information to the current location management unit 4 according to the road determination by the road determination unit 23. To do.

  As shown in FIG. 8, the dead reckoning processing unit 3 includes a dead reckoning result unit 31, a guessed trajectory creation unit 32, a learning unit 33, and a correction unit 34, and includes a vehicle speed sensor 51, a G sensor 52, a gyro 53, and a GPS 54, respectively. The estimated locus is generated and is sent to the current location management unit 4 as estimated information together with various sensor information. The learning unit 33 is for learning the sensitivity and coefficient related to each sensor, and the correction unit 34 is for correcting a sensor error and the like. The estimated trajectory creation unit 32 creates a vehicle estimated trajectory from each sensor data, and the dead reckoning navigation result unit 31 sends the created estimated trajectory of the dead reckoning result and various sensor information to the current location management unit 4 as estimated information. .

  FIG. 9 is a diagram for explaining an example of the structure of a database, FIG. 10 is a diagram for explaining an example of micro-matching processing by feature determination, FIG. 11 is a diagram for explaining an example of micro-matching processing by lane judgment, and FIG. FIG. 13 is a diagram for explaining an example of an object or paint, and FIG. 13 is a diagram for explaining determination of a lane position, a position in a lane, and a straddle state.

  The guide road data file is stored in the database. As shown in FIG. 9A, the guide road data file includes a road number, a road number, a road number, It consists of length, road attribute data, shape data address, size and guidance data address, and size data, and is stored as data required for route guidance obtained by route search.

  As shown in FIG. 9B, the shape data has coordinate data including east longitude and north latitude for each of the number of nodes m when divided by a plurality of nodes (nodes) of each road. As shown in FIG. 9C, the guidance data includes intersection (or branch point) name, attention point data, road name data, road name voice data address, size, destination data address, size, and feature data. Consists of address and size data.

  Among these, for example, destination data includes a destination road number, a destination name, an address of destination name voice data, a size and destination direction data, and travel guidance data. Of the destination data, the destination direction data is invalid (no destination direction data is used), unnecessary (no guidance), straight ahead, right direction, diagonal right direction, right return direction, left direction, diagonal left direction, left return This is data indicating direction information.

  As shown in FIG. 9D, the feature data includes a feature number, a feature type, a feature position, a feature recognition data address, and a size for each feature number k of each road. The object recognition data is data necessary for recognition for each feature as shown in FIG. 9E, such as shape, size, height, color, distance from the link end (road end), and the like.

  The road number is set for each direction (outbound path, return path) for each road between the branch points. The road attribute data is road guidance auxiliary information data, and is data indicating information on the elevated / underpass and the information on the number of lanes, which are elevated, next to the elevated, underpass, or next to the underpass. Road name data includes information on road types such as expressways, city expressways, toll roads, and general roads (national roads, prefectural roads, etc.) and information indicating whether the main road or road is attached to expressways, city expressways, and toll roads. It is composed of road type data and an intra-type number which is individual number data for each road type.

  For example, as shown in FIG. 10, the micro-matching process by the feature determination first acquires the current position of the macro information (step S21), searches the database from the current position, and acquires the feature recognition data (step S22). ). It is determined whether there is a feature to be recognized (step S23). If there is no feature to be recognized, the process returns to step S21 and the same processing is repeated. If there is a feature to be recognized, the image recognition device 8 is requested to perform image recognition of the feature (step S24).

  After waiting for the recognition result to be acquired from the image recognition device 8 (step S25), the current position obtained from the feature recognition information and the current position of the macro information are collated (step S26). If the current position obtained from the feature recognition information matches the current position of the macro information, the process returns to step S21 and repeats the same process. If the current position of the macro information does not match, the current macro information The position is corrected based on the current position obtained from the feature recognition information.

  For example, as shown in FIG. 11, the micro-matching process based on the lane determination includes an event input from the driver input information management unit 9 and an event input from the image recognition device 8 (step S31), from the image recognition result and the driver input information. The lane position and the position within the lane are specified (step S32), and the total number of lanes, the lane position, the position within the lane, and the confidence level of the micro matching result are transmitted as micro information (step S33). Next, the lane position and the in-lane position are collated with the current position of the macro information (step S34), and it is determined whether or not the lane position and the in-lane position match the current position of the macro information (step S35). If the lane position and the in-lane position match the current position of the macro information, the process returns to step S31, and the same processing is repeated. If not, the current position of the macro information is corrected based on the lane position and the in-lane position. (Step S36).

  The various features and paints are, for example, manholes (b), lanes (b, c), median or center lines (d), stop lines (e), sidewalk steps (f), road signs as shown in FIG. (G) and traffic lights (h). These features can be recognized from the shape of the image, and the current position can be obtained from the recognized position. Recognized positions such as features and paint are identified by the position of the mesh when the screen is cut with a mesh as indicated by the dotted line, or by the angle of view of the target feature or paint. can do. Further, the lane position, the in-lane position, and the straddling state can be determined from the positions of the lower points on the screen of the lane (white line) a, the center line b, and the road shoulder c as shown in FIG.

  FIG. 14 is a diagram for explaining a determination example of a lane position, a position in a lane, and a straddle state using an estimated trajectory, and FIG. 15 is a diagram for explaining a determination example of a lane position, a position in a lane, and a straddle state using an optical beacon. is there.

  Even when the image recognition device 8 cannot be used, an estimated trajectory or an optical beacon can be used to determine the lane position, the in-lane position, and the straddle state. When using the estimated trajectory, for example, as shown in FIG. 14, the current location management unit 4 monitors the estimated information (the trajectory or the left / right movement amount), for example, to integrate the movement amount in the width (left / right) direction of the lane. By comparing with the lane width, the lane movement can be determined when the movement amount becomes the lane width, and the straddle state can be determined by 1/2. Further, a correction may be made to determine whether the position in the lane is right or left.

  Since information on lanes is included in the optical beacon, the optical beacon shown in FIG. 15 can be used regardless of the presence or absence of a camera and an image recognition device. Since there are cases where it is not possible, priority is given to optical beacon information. In addition, the final lane determination result is determined by combining both the current determination lane position and the optical beacon information. If the information does not match, for example, the degree of confidence can be reduced. Good.

  In addition, this invention is not limited to the said embodiment, A various deformation | transformation is possible. For example, in the above-described embodiment, the movement of the lane is detected by integrating the amount of movement of the lane in the left-right direction.

It is a figure which shows embodiment of the next road prediction apparatus of the traveling vehicle which concerns on this invention. It is a figure explaining the example of the next road prediction by road information and learning. It is a figure explaining the example of the next road prediction by vehicle speed information. It is a figure explaining the example of the next road prediction by driver input information. It is a figure explaining the example of the next road prediction process. It is a figure which shows embodiment of the present location information management apparatus of a vehicle. It is a figure which shows the structural example of a macro matching process part. It is a figure which shows the structural example of a dead reckoning process part. It is a figure explaining the structural example of a database. It is a figure explaining the example of the micro matching process by feature determination. It is a figure explaining the example of the micro matching process by lane determination. It is a figure explaining the example of various features and a paint. It is a figure explaining determination of a lane position, a position in a lane, and a straddle state. It is a figure explaining the example of determination of the lane position using a presumed locus | trajectory, the position in a lane, and a straddle state. It is a figure explaining the example of determination of the lane position using an optical beacon, the position in a lane, and a straddle state.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Micro matching process part, 2 ... Macro matching process part, 3 ... Dead reckoning process part, 4 ... Present location management part, 5 ... Vehicle control apparatus, 6 ... Vehicle information processing apparatus, 7 ... Database, 8 ... Image recognition apparatus, DESCRIPTION OF SYMBOLS 9 ... Driver input information management part, 11 ... Position collation & correction part, 12 ... Feature determination part, 13 ... Micro matching result part, 14 ... Lane determination part, 21 ... Learning processing part, 22 ... Next road prediction processing part, 23 ... Output device

Claims (2)

A next road prediction device for a traveling vehicle that predicts a next road in a traveling direction at an intersection where the vehicle travels,
Image recognition means for recognizing an image acquired by a camera and acquiring lane information on a road on which the vehicle is currently traveling ;
A next road prediction processing means for predicting the next road in the traveling direction at the intersection based on the lane information ;
Learning processing means for recognizing and learning the selection of the next road in the direction of travel at the intersection by the driver ,
The next road prediction processing unit for a traveling vehicle, wherein the next road prediction processing unit predicts a next road in a traveling direction at an intersection based on learning by the learning processing unit . 2. The next road of the traveling vehicle according to claim 1 , wherein the learning processing means accumulates a comparison between the next road in the traveling direction at the predicted intersection and the traveling performance and obtains and outputs a confidence level. Prediction device.

Images