JP2012173843A - Travel route generation device and travel route generation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve route generation by estimating a self-location more reliably even if there is a significant change in a travelling direction of vehicle MM between a case of constituting a travel history and a case of generating a travel route.SOLUTION: A travel route generation method in the invention includes: a step for individually taking each image of at least both vehicle front and vehicle rear with two or more cameras 2 that is capable of individually taking each image of at least both vehicle front and vehicle rear, and for storing each taken image as a stored image upon associating it with a camera location and an imaging direction angle; a step for verifying the stored images based on the imaging direction angle and the camera location, and for selecting stored images including images for which a degree of overlap with images taken by the cameras 2 is estimated to be relatively high; and a step for generating a route by estimating a location of own vehicle based on the stored images selected and the images taken by the cameras 2.

Description

  The present invention relates to a technique for generating a travel route based on a captured image captured by a camera.

  As a technique for generating a travel route based on an image captured by a camera, for example, there is a technique described in Patent Document 1. This prior art is configured so that an image behind the vehicle can be captured by a rear camera attached to the rear of the vehicle. Then, when the vehicle moves forward, a traveling locus at the time of forward movement is detected and stored based on an image captured by the rear camera, and when the vehicle moves backward, driving assistance is performed based on the detected traveling locus. That is, a travel locus stored in advance is used as a travel route.

JP 2007-237930 A

The prior art uses a travel locus calculated from an image acquired during forward travel as a travel route during reverse travel. However, in the above-described conventional technique, when the vehicle is turned in the opposite direction to the direction in which the travel track is stored and the same route is traveled in the reverse direction, the travel track may not be used as the travel route. is there. In other words, when generating a return travel route based on information on the travel path of a path that has gone to the destination, in the above-described conventional technology, the imaging direction of the camera is reversed, so the self-position estimation from the camera image Is difficult, and path generation becomes very difficult.
The present invention focuses on the above points, and estimates the self-position more reliably even when there is a large change in the traveling direction of the vehicle between the case where a travel history is configured and the case where a travel route is generated. The purpose is to enable path generation.

  In order to solve the above-described problem, one embodiment of the present invention provides at least a front of a vehicle and two or more cameras that can individually capture at least the vehicle front and the vehicle rear mounted on a vehicle capable of grasping the vehicle position. In addition, the rear of the vehicle is individually imaged. Then, the camera position and the imaging direction angle based on the vehicle position when the feature points in each captured image are imaged are accumulated and accumulated as an accumulated image, and the accumulated image is collated based on the imaging direction angle and the camera position. A stored image that is estimated to have a relatively large overlap between the captured image captured by the camera and the image is selected. A travel route of the vehicle is generated based on the feature points in the selected accumulated image.

  According to the present invention, images in a plurality of directions with completely different imaging direction angles can be stored as stored images. For this reason, it is possible to select the stored image in consideration of the imaging direction angle (camera direction) of the camera. Even if there is a large change in the direction of travel, such as when the direction of travel changes in the reverse direction, as a stored image for estimating the self-position, a stored image other than a stored image before a minute time can also be applied. In addition, since the self-position can be estimated with such a more appropriate accumulated image, the path can be generated more reliably.

Further, by selecting the stored image based on the imaging direction angle and the camera position, it is possible to select a more appropriate stored image even if stored images from different cameras are mixed.
As described above, according to the present invention, even when there is a large change in the traveling direction of the vehicle between the case where the travel history is configured and the case where the travel route is generated, the self-position is more reliably estimated to generate the route. It becomes possible.

It is a driving assistance device system block diagram carried in vehicles concerning an embodiment based on the present invention. It is a processing block diagram of the driving assistance device concerning an embodiment based on the present invention. It is a conceptual diagram which shows that suitable image selection can be performed from several camera images (accumulated image). It is a figure which shows the flowchart of the process at the time of driving | running | working history structure. It is a figure which shows the flowchart of the process at the time of driving | running | working assistance. It is a figure for demonstrating operation | movement etc. of this invention, Comprising: (a) is a figure explaining the time of driving | running | working history composition, (b) is a figure explaining the time of driving | running route generation.

(Please make corrections as appropriate based on the amendments to the above claims.)
“First Embodiment”
Next, embodiments of the present invention will be described with reference to the drawings.
The travel support device of this embodiment includes the travel route generation device of this embodiment as a component.
(Constitution)
First, the configuration of the driving support device 100 according to the present embodiment will be described with reference to FIGS. 1 and 2. FIG. 1 is a system configuration diagram of the travel support apparatus 100 of the present embodiment, and FIG. 2 is a block diagram schematically showing the processing contents performed by the travel support apparatus 100 of the present embodiment.

  As shown in FIG. 1, the driving support device 100 of the present embodiment includes a plurality of cameras 2 capable of individually capturing images in front or rear of the vehicle front-rear direction, a driving support ECU 3, a vehicle information DB 4, a driving history DB 5, The display 6, the touch panel 7, the vehicle ECU 8, the vehicle sensor group 9, and the control unit group 10 are configured. These components are mounted on the vehicle MM as shown in FIG.

  The plurality of cameras 2 includes a front camera 2 (2a) that can image the front in the vehicle front-rear direction and a rear camera 2 (2b) that can image the rear in the vehicle front-rear direction. For these cameras 2, for example, a camera using a solid-state image sensor such as a CCD may be used. And the front camera 2 (2a) is installed in the direction which can image the vehicle front, for example in the back side of the in-vehicle mirror part of the vehicle MM. The rear camera 2 (2b) is installed, for example, on the upper part of a license plate attached to the front part of the vehicle. Images individually captured by each camera 2 are sent to the driving support ECU 3 respectively.

  The travel assistance ECU 3 is an electronic control unit that performs processing such as self-position estimation using a captured image and travel route calculation. The travel assist ECU 3 may also be used as an ECU used for other controls. The driving support ECU 3 is configured by a CPU, a ROM, a RAM, and the like in the same manner as a general ECU, and a program for realizing various processing units described later is stored in the ROM. Functionally, the driving support ECU 3 includes processing blocks of a feature point detection unit 31, a data storage unit, a vehicle position estimation unit 33, an accumulated image selection unit 34, a travel route calculation unit 35, and a driving support unit 36. Prepare. The processing contents of these processing blocks will be described later.

The vehicle information DB 4 is a storage unit in which various parameters related to the vehicle MM and the camera 2 are stored. In the vehicle information DB 4, for example, information related to the vehicle model such as the minimum turning radius, vehicle width, tread length, tire radius of the vehicle MM, and information related to the camera 2 design values such as the internal parameters of the camera 2, the optical axis direction, and the angle of view. Is remembered.
The travel history DB 5 is a storage unit that stores a captured image captured by the camera 2 as an accumulated image in association with additional information such as a camera position at the time of imaging, an imaging direction angle, a vehicle position, and a vehicle posture. As the unit, for example, a storage medium such as a hard disk or a memory card may be used.

The display 6 is installed at a center console in the room of the vehicle MM, for example. For example, it is used as a display device that displays a predicted route and a recommended route of a host vehicle superimposed on a camera image during driving support. The display 6 may also be used as a display generally used for a navigation device.
The touch panel 7 is configured by, for example, a general touch panel that is arranged on the display unit of the display 6 and can specify the coordinate position of the part touched by the user and the movement locus of the contact part. In the present embodiment, it is used when the user selects whether to start route saving or instructs whether to start driving support. As an instruction input interface from the user, an input interface such as a button, a mouse, a keyboard, and a voice input may be used instead of the touch panel 7.

  The vehicle ECU 8 is an electronic control unit that controls operations of an engine, a brake, and the like. The vehicle ECU is connected to a control unit group 10 including control units such as an engine control ECU, a brake control ECU, and a steering control EPS. The vehicle ECU 8 calculates command values such as the accelerator opening, the brake control amount, the steering angle, etc. based on the output from the driving support ECU 3, and transmits the calculated command values to the control unit group 10. As a result, the vehicle MM can be controlled along the travel locus calculated by the travel support ECU 3.

  The vehicle sensor group 9 includes various vehicle sensors such as an accelerator sensor, a brake sensor, a steering sensor, a vehicle speed sensor, and an acceleration sensor, and is connected to the travel support ECU 3. The driving assistance ECU 3 can calculate the vehicle speed, acceleration, steering angle, accelerator operation amount, brake operation amount, and the like of the vehicle MM2 from the output values of the vehicle sensor group 9.

Next, each processing block of the driving support ECU 3 will be described.
The feature point detection unit 31 detects a feature point existing in the screen of the target image with respect to a captured image of the camera 2 or a past captured image (accumulated image) accumulated in the travel history DB 5. Part. A feature point is a point that can be followed when the screen moves. Generally, a point having a luminance difference in both the vertical direction and the horizontal direction in an image is extracted as a feature point. That is, feature points can be detected by image processing such as edge processing, and various methods (SIFT, SURF, fast, etc.) have been proposed as well-known processing methods as robust feature point detection methods. What is necessary is just to detect by extracting a feature point with these processing methods.

The data storage unit 32 captures each captured image captured by each camera 2, the feature point obtained by the feature point detection unit 31, the camera position based on the host vehicle position estimated by the host vehicle position estimation unit 33, and the imaging The information is stored in the travel history DB 5 as an accumulated image in association with additional information such as the imaging direction angle at the time.
Here, since the imaging direction angle can be specified by the imaging direction (installation direction) of the camera 2 with respect to the direction of the vehicle MM, there is no problem even if the direction of the vehicle MM is stored as information on the imaging direction angle of each camera 2. . Similarly, since the installation position of the camera 2 with respect to the vehicle MM is also specified, the vehicle position and the direction of the vehicle MM may be used as the camera position information. In addition, as a vehicle position, a vehicle gravity center point is used, for example. Note that the imaging direction angle is the visual axis direction in which the camera 2 is facing, and is set to the center position of the imaging angle of view, for example.

  Further, the feature points of the captured image of the camera 2 are compared by using the feature points of the captured image of the camera 2 and the feature points of the stored image selected from the travel history DB 5 by the stored image selection unit 34. Then, the three-dimensional position is measured and further converted into the world coordinate system, that is, specified by the coordinates in the world coordinate system and stored. At this time, the position measurement result from the camera 2 to the feature point is represented in the camera 2 coordinate system, and in order to convert this to the world coordinate system, the position of the camera 2 in the world coordinate system and the tertiary of the feature point It is necessary to obtain the original configuration at the same time. As described above, SLAM (Simultaneous Localization And Mapping) may be applied as a method for simultaneously generating a map and estimating a self-location. Various algorithms have been published for this SLAM. Examples of SLAM using the camera 2 include Mono-SLAM and PTAM.

The accumulated image selection unit 34 selects an accumulated image suitable for collation with the captured image captured by the camera 2 from the travel history DB 5.
The accumulated image selection unit 34 first acquires the imaging direction angle of the camera 2 (imaging direction angle of the captured image). The imaging direction angle of the camera 2 may be calculated from information on the vehicle position at the time of imaging and information such as the mounting position of the camera 2 accumulated in the vehicle information DB 4. The imaging direction angle is a direction angle in a top view. Next, the obtained imaging direction angle and the imaging direction angle corresponding to each accumulated image in the travel history DB 5 are compared, and the deviation of the direction from the imaging direction angle of the camera 2 (difference angle) is set in advance. The accumulated image group within the set deviation angle is extracted. The set deviation angle is set to an angle having a value equal to or less than 50% of the angle of view of the camera 2, for example. Next, among the extracted accumulated image groups, the accumulated images are within a preset set approach distance, and are captured at a position closest to the target camera 2 (the vehicle position may be substituted). Select the stored image.

  A specific selection example will be described with reference to FIG. A candidate (3) having a large image overlap with respect to the current image is set as a suitable stored image capturing position. The setting deviation angle of the imaging direction angle is set to be half or less of the camera 2 field angle. Since the camera 2 is in the opposite direction, that is, the deviation angle θ1 with respect to the candidate (1) is larger than the set deviation angle, the candidate (1) can be excluded. For this reason, candidate (2) and candidate (3) remain based on the deviation angle. Next, since the closest captured image is selected from the remaining candidates, the candidate (3) can be selected. By limiting the imaging direction angle, it can be expected that about half of the images overlap. Moreover, the possibility that there are many common feature points increases by selecting a closer stored image. As a result, the self-position estimation unit 33 contributes to improving the self-position estimation accuracy.

As described above, the stored image selection unit 34 can select a captured image suitable for matching with the current captured image from the travel history DB 5 in which captured images from the plurality of cameras 2 are mixed.
The own vehicle position estimation unit 33 uses the captured image captured by the camera 2, the stored image selected by the stored image selection unit 34, the feature points in the stored image, and the vehicle information stored in the vehicle information DB 4. To estimate the vehicle position.

  First, the self-position estimation unit first obtains prior knowledge (a vehicle position estimation result (for example, a vehicle position estimation result obtained in the immediately preceding processing step) in the self-position estimation performed previously) and a vehicle speed calculated from the vehicle sensor group 9. , Vehicle information such as steering angle) is used to perform pre-estimation processing of the vehicle position. Thereafter, when the position and orientation of the vehicle MM are obtained by the pre-estimation, the self-position estimating unit uses the feature MM that can be reflected in the captured image of the camera 2 among the feature points accumulated in the travel history DB 5. Select points. Specifically, the feature points are selected from the feature points in the stored image selected by the stored image selection unit 34 and the captured image captured by the camera 2. The position of the selected feature point in the captured image is estimated, and the feature point is searched for in the image around the estimated position value in the image. By the above search, a deviation between the position in the image assumed from the pre-estimated value and the position in the image actually detected is detected. Thereafter, the self-position estimating unit calculates the own vehicle position by correcting the prior estimated value of the own vehicle position from the difference between the feature point position in the image and the feature point position calculated from the previous estimated value of the own vehicle position. This is called posterior estimation. As described above, the self-position estimation unit according to the present embodiment performs a posteriori estimation by adjusting the result of the prior position estimation so that the position of the feature point coincides with the actual measurement position. Obtain the vehicle position estimation result. The final vehicle position estimation result (vehicle position information) estimated by the self-position estimation unit is accumulated in the travel history DB 5.

  The travel route calculation unit 35 reads the past vehicle position trajectory stored in the travel history DB 5 based on the current vehicle position obtained using the stored image selected by the stored image selection unit 34, and the travel route. And In other words, the travel route calculation unit 35 generates a travel route based on the accumulated image selected by the accumulated image selection unit 34 and the past vehicle position trajectory accumulated in the travel history DB 5. At this time, a three-dimensional feature point group obtained by collating a plurality of captured images stored in the travel history DB 5 is also calculated at the same time. The trajectory of the past vehicle position that is the travel route is calculated from the camera position at the time of imaging of the captured image used for calculating the feature points, and is defined by the relative relationship with the three-dimensional feature point group.

In addition, the travel route calculation unit 35 can use the direction of the read travel route in the reverse direction. As a method of selecting a direction, for example, the vehicle may travel in the forward direction if the current position on the route is close to the start point, or in the reverse direction if the current position is close to the end point, or the driver may indicate the direction or destination through the touch panel 7. You may decide by doing.
The travel support unit 36 performs travel support using the travel route determined by the travel route calculation unit 35. Specifically, after the generated travel route information is converted into the camera 2 coordinate system of the camera 2 (for example, the front camera 2 (2a)), the travel route is superimposed on the captured image and displayed on the display 6. Furthermore, the driving support unit 36 can control the accelerator, the brake, and the steering, and perform automatic driving by sending the amount of deviation between the driving locus and the position / posture of the vehicle MM to the vehicle ECU 8.

Next, among the processes performed by the travel support apparatus, a process for configuring a travel history and a process at the time of travel support using the travel history will be described with reference to the flowcharts of FIGS. 4 and 5. The travel history configuration process and the travel support process are continuously performed in the travel support ECU 4 at a preset control cycle, for example, at an interval of 100 msec.
First, the travel history configuration process will be described with reference to the flowchart of FIG.
In the following description, the creation is started when there is a disclosure instruction from the driver. However, the creation is automatically started when the driving support ECU 4 satisfies a preset condition. May be.

First, the processing flow when the travel history is configured will be described with reference to FIG.
In step S <b> 110, a button is displayed on the display 6 as to whether or not the travel history storage is to be started, and the driver's input can be accepted through the touch panel 7. And when it determines with there being a start instruction based on a driver | operator's input, it transfers to step S111. When there is no start instruction, the determination in step S110 is repeated at a preset control cycle, thereby waiting for an input from the driver.
In step S111, it is determined whether or not a past travel history exists in the travel history DB 5. If it is determined that there is no travel history, the process proceeds to step S112. If it is determined that a travel history exists, the process proceeds to step S113.

  In step S112, a travel history initialization process such as identification of the self-position that is the starting point of the travel locus is performed. First, the feature point detection unit 31 detects a feature point from the captured image of the camera 2. Next, after a preset time has elapsed (for example, the imaging interval of the camera 2), the captured image is captured again from the camera 2, and the feature point detection unit 31 calculates a feature point from the captured captured image. It is assumed that the acquisition positions of the two images are different such as when the vehicle MM is moving. Then, the feature points are tracked by comparing the feature points in the new image with the respective feature points acquired previously. Here, as a feature point tracking method, a generally known image processing method such as KLT Tracker may be applied. Feature point tracking is continuously performed by continuously repeating the captured image capturing and feature point detection processes. Further, when the vehicle MM moves a certain distance, for example, 0.5 m, the feature point of the first captured image is compared with the feature point of the current captured image, and a known eight-point method or the like is used. Then, each feature point for map creation is made three-dimensional, and the initial vehicle position and initial vehicle attitude of the vehicle MM are defined based on the relative relationship between the feature point and the vehicle MM. Above, the process of step S112 is complete | finished and it transfers to step S113.

The following processing from step S113 to step S117 is a processing group that is repeatedly performed as the vehicle MM progresses.
In step S113, the stored image selection unit 34 collates with the current captured image from the stored image group in the travel history DB 5 using the imaging direction angle captured by the camera 2 and the camera position based on the vehicle position. An accumulated image (past captured image) suitable for the above is selected. Thereafter, the process proceeds to step S114.

In step S114, the vehicle position estimation unit 33 uses the relative relationship between the feature point of the captured image captured by the camera 2 and the acquired three-dimensional feature point group (information on the constructed map that configures the traveling environment). Estimate the vehicle position. Thereafter, the process proceeds to step S115.
In the present embodiment, the host vehicle position estimation unit 33 estimates the host vehicle position from the correlation between the feature points of the current captured image and the feature points of the accumulated image selected in step S113.

  Here, when the traveling direction of the vehicle MM is not greatly changed, it is considered that the accumulated image acquired before one control step is almost selected, but when the traveling direction of the vehicle MM is largely changed, An accumulated image different from the accumulated image acquired before one control step may be selected. This is because the accumulated images captured by the plurality of cameras 2 are mixed and accumulated in the travel history DB 5.

  In view of this, in step S113, the accumulated image is based on the imaging direction angle and the camera position only when the amount of change in the traveling direction of the vehicle MM within the preset time is changed by a preset angle, for example, 90 degrees or more. In other cases, processing may be performed so that the image captured immediately before is selected as the accumulated image. In addition, the stored image may be selected based on the imaging direction angle and the camera position only when switching from the travel history configuration time to the route generation time.

  In step S115, based on the own vehicle position information obtained in step S114, a three-dimensional feature point, which is information on a map constituting the traveling environment, is updated. Specifically, the feature points in the captured image acquired from the camera 2 are compared with the feature points of the stored image selected in step S113, and the feature points are obtained by using a generally known technique such as the 8-point method. Is added to a known feature point group. This process may be performed together with the self-position estimation in the process of step S114. That is, map update processing for generating a travel route may be performed together with self-position estimation. Thereafter, the process proceeds to step S116.

  In step S116, the data storage unit 32 stores the captured image of the camera 2, the three-dimensional feature point group, the vehicle position, and the imaging direction angle information group acquired in steps S112 to S115 in the travel history DB 5, and then The process proceeds to step S117. In the travel history DB 5, various vehicle information such as the vehicle speed and the steering angle calculated from the vehicle sensor group 9 may be stored in association with the vehicle position. By this processing, the data storage unit 32 stores the captured image captured by the camera 2 as a stored image in the travel history DB 5 every time a camera image is acquired for processing. Specifically, the captured image captured by the camera 2 is stored in the travel history DB 5 as an accumulated image in a state associated with the information on the camera position and the imaging direction angle by the above-described processing.

  In step S117, a button indicating whether or not to save the travel history is displayed on the display 6, and the driver's input is accepted through the touch panel 7. When it is determined that there is no termination instruction from the driver, the process returns to step S113 again. When it is determined that there is an termination instruction, the process is terminated. Even if there is no instruction to end the travel history configuration, the process may be configured to shift to the process at the next travel support when the next travel support start instruction is issued.

Next, processing at the time of driving support will be described with reference to the flowchart of FIG. It is assumed that the processing at the time of configuring the travel history is performed prior to travel support.
First, in step S120, a button as to whether or not to start driving support is displayed on the display 6, and the driver's input is accepted through the touch panel 7. If it is determined that there is a start instruction from the driver, the process proceeds to step S121, and if it is determined that there is no start instruction, the inquiry of step S120 or the input confirmation process of the start instruction is repeated, Wait for input.
In step S121, the vehicle position estimation unit 33 uses the current captured image captured by the camera 2 and, in particular, the storage selected by the storage image selection unit 34 in the relative relationship with the read three-dimensional feature point group. From the relative relationship with the feature points in the image, an initial vehicle position (before route generation) is estimated, and the process proceeds to step S122.

  In step S122, the travel route calculation unit 35 reads the history of the past vehicle position and the three-dimensional feature point group (the map information) from the travel history DB 5, and calculates the current vehicle position and the three-dimensional feature point group. Based on the relative relationship, particularly the relative relationship with the feature point in the stored image selected by the stored image selection unit 34, the trajectory of the past vehicle position corresponding to the traveling direction of the current vehicle MM is calculated as a candidate for the travel route. To do. Thereafter, the process proceeds to step S123. If it is determined that the trajectory of the past vehicle position cannot be formed in the vehicle traveling direction, a travel route on which the host vehicle MM can travel is calculated based on the three-dimensional feature point group (the map information). May be.

In step S123, based on the vehicle position information on the travel route, the direction of the route is determined so that the vehicle is used in the forward direction if it is close to the route start point, and in the reverse direction if it is close to the end point. The direction of the travel route is set so as to become, and the process proceeds to step S124.
Note that, based on the relationship between the obstacle and the obstacle specified by the three-dimensional feature point group centered on the accumulated image selected by the accumulated image selection unit 34, the traveling route becomes a more appropriate route. You may correct | amend a path | route.
In step S <b> 124, a button indicating whether or not to perform automatic driving is displayed on the display 6, and the driver's input can be received through the touch panel 7. When there is an automatic driving instruction from the driver, the process proceeds to step S125, and when there is no automatic driving instruction, the process proceeds to step S126.

  In step S125, automatic driving is performed based on the travel route set in step S123. For example, the vehicle speed is set to a constant speed, and the steering angle is calculated to correct the travel route to be supported and the vehicle position / posture. The vehicle ECU 8 calculates command values for the accelerator opening, the brake control amount, and the steering angle for realizing the calculation result, and sends the command values to the control unit group 10 so that the vehicle MM along the travel support route. The automatic operation process is carried out. Thereafter, the process proceeds to step S127.

In step S126, driving assistance is performed based on the driving route set in step S123, and the process proceeds to step S127. Specifically, the set travel route obtained by displaying the captured image captured by the front camera 2 (2a) facing the vehicle traveling direction on the display 6 and then converting it into the coordinate system of the front camera 2 (2a). And an expected trajectory based on the current steering angle are displayed in a superimposed manner. This assists the driver in driving along the driving support route. As another driving support method, for example, EPS control may be performed based on the steering angle calculated in step S125, and steer steering guidance may be performed using a reaction force.
In step S127, it is determined whether or not the vehicle position has reached a certain range at the end of the driving support route. If not, the process of step S125 or step 126 is repeatedly executed based on the determination result of step S124. On the other hand, when the end is reached, the process is terminated.
This is the end of the driving support process.

(Operation other)
Driving route generation technology is important in order to realize driving support or automatic driving in general driving environments such as narrow roads and parking lots, not simple situations such as straight driving and following driving on highways. .
In general, when a travel route is calculated, an environment map including information such as a travelable region or an obstacle is constructed based on sensor information, and the route is calculated together with vehicle information such as a vehicle position with respect to the environment. However, in such a technique, since the route is calculated using the environment information and the vehicle information for generating the ideal route, the calculation takes time. In addition, an expensive sensor may be required for environment recognition, or the path may not necessarily be calculated depending on the recognition status of the sensor.

  In contrast, the present embodiment is a memory-based travel route generation technology that uses past travel trajectories, and only reads the accumulated travel history, so that the calculation cost is low. Furthermore, in the case of a route that has been taken once in the past, a traveling locus can always exist, and it is sufficient to provide the minimum necessary sensors for obstacle avoidance and the like. For this reason, the route generation technique of this embodiment is generally low in cost.

  Moreover, in this embodiment, the two cameras 2 of the camera 2 which images at least the vehicle front and the camera 2 which images the vehicle front are provided. By mounting the camera 2 on the front and rear in this way, the feature point that was visible with the rear camera 2 (2b) is the front camera 2 (2a), and the feature point that was visible with the front camera 2 (2a) is the rear camera 2. In (2b), it becomes observable. For this reason, even if there is a large change in the direction of the vehicle MM, it is possible to generate a route based on the travel locus based on the past self-position history.

For example, as shown in FIG. 6A, consider a case where the vehicle enters the parking lot from the front of the vehicle and is parked at the parking position. It is assumed that the driver inputs a travel history start instruction during the parking.
In this case, it is possible to obtain a travel locus indicated by a thick arrow as shown in FIG. 6A based on the travel history of the self-position estimation results sequentially performed as the vehicle MM progresses. In addition, the feature points ◯ and Δ can be acquired from images captured by the front camera 2 (2a) and the rear camera 2 (2b) when traveling toward the parking position PA, respectively. In FIG. 6A, points indicated by ◯ are feature points captured by the front camera 2 (2a), and points indicated by Δ are points captured by the rear camera 2 (2b). Thus, there is a difference in the feature points captured by each camera 2.

  Assuming that the vehicle travels from the same parking position PA as shown in FIG. 6B in the state where the travel history corresponding to the parking position PA has already been configured as described above. In this case, the direction of the vehicle MM is greatly different between when the travel locus history is taken and when the route is generated.

  At this time, usually, the traveling position is generated by estimating the vehicle position based on the captured image captured by the front camera 2 (2a). In the present embodiment, as shown in FIG. 6, the accumulated image that has a small deviation from the imaging direction angle of the front camera 2 (2 a) and is close to the current front camera 2 (2 a) position (or vehicle position) 2 (2b) is an accumulated image (image obtained by capturing the feature point Δ). In this embodiment, the accumulated image selection unit 34 selects the accumulated image captured by the rear camera 2 (2b), the feature point in the accumulated image by the selected rear camera 2 (2b), and the front camera. 2 (2a) is used to estimate the self-position in the map information (constructed three-dimensional feature point group) by comparing with the feature points in the imaging screen imaged based on the past self-position history. It is possible to generate a travel route when leaving the parking lot.

  At this time, since the direction of the travel locus when entering the parking lot is opposite to the direction of the travel route when leaving the parking lot, the direction of the travel locus when entering the parking lot is reversed. In addition, a travel route is generated. That is, by mounting the camera 2 on the front and rear, it is possible to generate a travel route in the opposite direction to that when the trajectory is stored.

  Here, when two or more cameras 2 are mounted, it is difficult to select an image for constructing a three-dimensional feature point. As an example, consider the situation shown in FIG. 3 in which an optimal stored image is selected from a plurality of past images (stored images) for the current camera image. In general, in the construction of three-dimensional feature points, for any camera image, an accumulated image that contains many of the same feature points is selected, and the feature points are calculated by calculating the correlation of the feature points between the two camera images. 3D. Here, when only one camera 2 is used, the camera image captured at the position closest to the captured image (candidate (1) in FIG. 3) is usually selected. However, when there are a plurality of cameras 2, the closest camera image is not necessarily optimal. For example, in FIG. 3, the candidate (3) should be selected intuitively, but cannot be simply solved. In FIG. 3, three cameras 2 are illustrated for easy understanding of the problem. However, only the two front and rear cameras 2 have a problem of which one of the cameras 2 should acquire the stored image. There is no change.

  On the other hand, in the present embodiment, a stored image suitable for the current captured image can be selected by performing a search based on the vehicle position and the imaging direction angle. That is, by selecting an accumulated image with a small shift in the imaging direction angle, an image with a large image overlap can be selected. In addition, by selecting an image within a preset approach distance, the possibility of selecting a stored image having more common feature points, in particular, many feature points near the current camera position is increased. As a result, even when a plurality of cameras 2 attached to the front and rear are used, it is possible to stably generate a travel route.

Further, by attaching the cameras 2 to the front and rear, as described above, the vehicle position can be stably estimated even when the vehicle MM is turned in the opposite direction, and the route can be reproduced in the reverse direction. This makes it possible to generate a route twice as long as the recorded route, and has the effect of doubling the route storage efficiency.
Here, the data storage unit 32 constitutes an image storage unit. The own vehicle position estimation unit 33 constitutes vehicle position detection means. The stored image selection unit 34 constitutes an image selection unit. The feature point detection unit 31 constitutes feature point detection means. The travel route calculation unit 35 constitutes a travel route calculation unit.

(Effect of this embodiment)
(1) This apparatus includes two or more cameras 2 that can individually capture at least the vehicle front and the vehicle front, respectively, as the camera 2. The data storage unit 32 stores the captured images captured by the respective cameras 2 as stored images in association with information on the imaging direction angle that is the imaging direction and information on the camera position at the time of imaging. The accumulated image selection unit 34 compares the imaging direction angle and the camera position at the time of imaging captured by the camera 2 with the imaging direction angle and the camera position associated with each of the accumulated images. An accumulated image that is estimated to have a relatively large image overlap with a captured image captured by the is selected.
The travel route calculation unit 35 calculates the travel route of the vehicle from the feature points in the selected accumulated image.

  Images in a plurality of directions with completely different imaging direction angles can be stored as stored images. For this reason, it is possible to select the stored image in consideration of the imaging direction angle of the camera 2 (the direction of the camera 2). As a result, even if there is a large change in the direction of the host vehicle MM during route generation, such as when the direction of travel changes in the reverse direction, as an accumulated image for estimating the self position, an accumulated image other than a minute-time previous accumulated image Accumulated images can also be applied. In addition, since the self-position can be estimated from the stored image, a route can be generated.

Further, by selecting the stored image based on the imaging direction angle and the camera position, it is possible to select a more appropriate stored image even if stored images from different cameras 2 are mixed.
In particular, according to the present invention, since at least the front of the vehicle and the front of the vehicle can be individually imaged, for example, even if the vehicle travels in the reverse direction, the self-position is estimated more reliably. A route can be generated at the same time.
As described above, according to the present invention, even if there is a large change in the traveling direction of the vehicle MM, it is possible to more reliably estimate the self-position and generate a route.

(2) The stored image selection unit 34 is associated with the history information of the vehicle position detected by the vehicle position detection means, and also selects the history information of the vehicle position.
A travel route is generated based on the accumulated image selected by the accumulated image selection unit 34 and the history of the self-position estimated by the self-position estimation unit.
As described above, even if there is a large change in the traveling direction of the vehicle MM between the travel history configuration and the route generation, the self-position can be estimated more reliably, so that the route generation can be performed more reliably.

(3) The accumulated image selection unit 34 has a deviation angle between the imaging direction angle captured by the camera 2 and the imaging direction angle associated with the accumulated image in a top view within a preset deviation angle. In addition, an accumulated image in which the distance between the camera position when the camera 2 captures an image and the camera position associated with the accumulated image is within a preset approach distance is selected.
By doing in this way, it becomes possible to select the accumulation image which overlaps relatively much with the picked-up image which the camera 2 images by making the deviation angle into less than the preset deviation angle set beforehand. Further, by selecting an accumulated image within the set approach distance, it is possible to select an accumulated image having more common feature points and feature points in the captured image captured by the camera 2.

“Second Embodiment”
Next, a second embodiment will be described with reference to the drawings. In addition, the same code | symbol is attached | subjected and demonstrated to the structure similar to 1st Embodiment.
The basic configuration of this embodiment is the same as that of the first embodiment. However, the processing of the stored image selection unit 34 is different. Other configurations and processes are the same as those in the first embodiment. For this reason, details of other configurations and processes are omitted.

Next, processing of the accumulated image selection unit 34 that is a processing block of the driving support ECU 3 and other processing contents related to the processing will be described.
The stored image selection unit 34 selects a captured image suitable for matching with the captured image of the camera 2 from the travel history DB 5.
First, the feature point detection unit 31 calculates a feature point of the current captured image. The feature point detection unit 31 also detects feature points for each accumulated image in the travel history DB 5. Then, the accumulated image selection unit 34 calculates how many feature points in common with the current captured image exist for each accumulated image. Then, the accumulated image selection unit 34 selects an accumulated image having the most common feature points.

Here, the target accumulated image group has a deviation from the imaging direction angle of the current camera 2 equal to or smaller than a preset deviation angle, and a distance from the current camera 2 is equal to or smaller than a preset approach distance. A set of accumulated images taken at a position.
As a result, it is possible to select an accumulated image in which the same feature point is reflected more reliably, so that a three-dimensional feature point can be more reliably configured and self-position estimation can be performed more reliably.

  Here, when calculating the number of common feature points, instead of adding all feature points, calculate the distance to each feature point, and then only feature points within a certain distance range from the camera position. May be added. As a result, a captured image having many common feature points in the vicinity can be preferentially selected, so that errors are reduced when three-dimensional feature points are configured. This effect is due to the fact that there is less error when comparing images with nearby feature points than when comparing images with distant feature points.

As a method of determining the above-mentioned fixed distance range, an average distance to the feature point may be calculated, and a range that is a fixed multiple of the average distance, for example, 1.5 times may be used. Thereby, the distance can be determined optimally according to the distribution state of the surrounding feature points.
Since the selection target is narrowed down by adding restrictions based on the imaging direction angle and the camera 2 angle of view, the calculation load can be reduced even if the accumulated image is selected based on the number of feature points.
As described above, the accumulated image selection unit 34 can select a captured image suitable for the current captured image from the travel history DB 5 in which captured images from the plurality of cameras 2 are mixed.
In addition, since the process procedure of the driving assistance device 100 is the same as that of the first embodiment, a description thereof will be omitted.

(Operation other)
According to the second embodiment, the number of feature points commonly included in the captured image and the stored image is calculated, and the stored image having the largest number of common feature points is selected. A large number of stored images can be selected.
In addition, when selecting an accumulated image taking into account the distance to the feature point, it is possible to preferentially select an accumulated image including a portion where many feature points are gathered. As a result, an optimal captured image can be selected in a robust manner in accordance with the feature point distribution situation of the surrounding environment.
Compared to the case where one accumulated image is selected based on the imaging direction angle and the camera 2 angle of view as in the first embodiment, it is possible to select a more suitable accumulated image by considering common feature points. It becomes. In other words, in the second embodiment, a suitable captured image can be selected more robustly according to the surrounding environment depending on the distribution state of the feature points.

(Effect of this embodiment)
In the present embodiment, in addition to the effects described in the first embodiment, the following effects are also achieved.
(1) The stored image selection unit 34 compares the feature points included in the captured image captured by the camera 2 with the feature points included in the stored image, and includes the feature points included in the captured image captured by the camera 2. The stored image having the largest number of common feature points that are common feature points is selected.
Since the accumulated image having the largest number of common feature points is selected, it is possible to more reliably select the accumulated image having many common imaging parts.

(2) The image selection unit calculates a feature point distance that is a distance from the feature point detected from the captured image to the camera position when the captured image is captured, and the common feature is based on the calculated feature point distance. Limit points.
Since the stored image is selected in consideration of the distance to the feature point, the possibility of preferentially selecting a captured image including a portion where many feature points are gathered is increased. As a result, it is possible to select a more optimal captured image in accordance with the feature point distribution situation of the surrounding environment.

2 camera 2a front camera 2b rear camera 31 feature point detection unit 32 data accumulation unit 33 own vehicle position estimation unit 34 accumulated image selection unit 35 travel route calculation unit 36 travel support unit 100 travel support device MM vehicle θ1, θ2, θ3 deviation angle

Claims (6)

A travel route generation device that generates a travel route based on feature points of a vehicle surrounding environment in an image captured by two or more cameras mounted on the vehicle so that at least the front and the rear of the vehicle can be individually captured. Because
The vehicle position detecting means for detecting the vehicle position and the feature points in the picked-up images picked up by the respective cameras are set to the information on the image pickup direction angle as the pick-up direction and the vehicle position detected by the vehicle position detecting means when picked up Image storage means for storing as a stored image in association with information on the camera position based thereon;
The imaging direction angle captured by the camera and the camera position at the time of imaging are compared with the imaging direction angle and camera position associated with each of the accumulated images. Image selecting means for selecting an accumulated image that is estimated to have a relatively large overlap;
A travel route generation device comprising: travel route calculation means for calculating a travel route of a vehicle from the feature points in the selected image.   2. The travel route according to claim 1, wherein the image selection means is associated with the history information of the vehicle position detected by the vehicle position detection means, and is selected in consideration of the history information of the vehicle position. Generator.   In the top view, the image selection means has a deviation angle between the imaging direction angle captured by the camera and the imaging direction angle associated with the accumulated image within a preset deviation angle, and the camera captures an image. The travel route generation device according to claim 1 or 2, wherein the stored image is selected so that a distance between the camera position and the camera position associated with the stored image is within a preset approach distance. . Comprising a feature point detection means for detecting a feature point from the captured image;
The image selection means is a feature point common to the feature points included in the captured image captured by the camera by comparing the feature points included in the captured image captured by the camera with the feature points included in the accumulated image. The travel route generation device according to any one of claims 1 to 3, wherein an accumulated image having the largest number of common feature points is selected.   The image selection unit calculates a feature point distance that is a distance from the feature point detected from the captured image to the camera position when the captured image is captured, and limits the common feature point based on the calculated feature point distance. The travel route generation device according to claim 4, wherein With at least two cameras mounted on a vehicle capable of grasping the vehicle position and capable of individually imaging the environment around the vehicle at least in front of the vehicle and behind the vehicle, at least the vehicle front and the rear of the vehicle are individually imaged. The camera position and the imaging direction angle based on the vehicle position when imaging the feature point of the
The above-mentioned accumulated image is collated based on the imaging direction angle and the camera position, and an accumulated image that is estimated to have a relatively large overlap between the captured image captured by the camera and the image is selected.
A travel route generation method comprising generating a travel route of a vehicle based on the feature points in the selected accumulated image.

Images