JP2013061270A - Present position calculation device and present position calculation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a present position calculation device and present position calculation method capable of properly calculating a true present position even on a monotonous road that does not have any target to be by suppressing accumulations of errors.SOLUTION: The present position calculation device of this invention includes an observation image input part 1 for receiving an observation image being a two-dimensional image at the present position of one's own vehicle 17, a reference image management part 9 for managing a reference image being a two-dimensional image prepared in advance in association with a predetermined place, a pixel pattern collation part 5 as a collation part for collating the observation image with a reference image corresponding to a temporary present position being a stored present position by image matching, and a present position calculation part 6 for calculating a true present position of the one's own vehicle 17 on the basis of a collation result in the pixel pattern collation part 5.

Description

  The present invention relates to a current position calculation device and a current position calculation method, and more particularly to a current position calculation device and a current position calculation method that can calculate a true current position of a moving object.

  The current position calculation device has conventionally calculated the current position of the moving body by various methods. For example, car navigation devices that use GPS location information to calculate the current position of a vehicle are often used.

  On the other hand, under the situation where it is difficult to acquire position information by GPS or the like, the current position is calculated and corrected using a vehicle speed pulse and an angular velocity sensor (gyro sensor) provided in a vehicle that is a moving body.

  However, such correction using the vehicle speed pulse and angular velocity sensor has a problem that errors generated when calculating the current position are accumulated. In addition, there is a problem in that an operation below the accuracy of the angular velocity sensor cannot be detected, and angular errors also accumulate.

  Thus, Patent Document 1 discloses a method for specifying the current position of a vehicle by using a map data list having landmarks (electric poles, traffic lights, road signs, etc.).

  Specifically, the distance between the farthest landmark that can be confirmed from the vehicle and the current position of the vehicle is measured in a three-dimensional manner using a stereo image of a pair of TV cameras that serve as both left and right eyes, for example. Ask for. Then, the current position of the vehicle is specified from the landmark coordinates on the map corresponding to the landmark.

  When finding the correspondence between landmarks and landmark coordinates, create the observation map list with the type of landmark and the mutual distance between the landmarks using the stereo image of the vehicle, List matching is performed and the correspondence is obtained.

JP-A-8-247775

  In the method shown in Patent Document 1, a target road can be set only at a specific point where there is a landmark (electric pole, traffic light, road sign, etc.) on the road, and there is no road that is likely to be a target. There was a problem that the landmarks could not be determined, etc.

  The present invention has been made to solve the above-described problems, and suppresses the accumulation of errors, and can calculate a true current position appropriately even on a monotonous road that has no possibility of becoming a target. An object is to provide a position calculation device and a current position calculation method.

  The present position calculation apparatus according to the present invention includes an observation image input unit to which an observation image, which is a two-dimensional image at a current position of a moving body, is input, and a reference that is a two-dimensional image prepared in advance in association with a predetermined point. A reference image management unit for managing images, a collation unit that collates the observed image with the reference image corresponding to the temporary current position that is a stored current position by image matching, and a collation in the collation unit And a current position calculation unit that calculates a true current position of the mobile body based on the result.

  In the present position calculation method according to the present invention, (a) a step of inputting an observation image that is a two-dimensional image at the current position of a mobile object, and (b) a reference image that is a two-dimensional image related to a predetermined point is prepared. (C) collating the observed image with the reference image corresponding to the temporary current position that is the stored current position by image matching; and (d) collating in the step (c). And a step of calculating a true current position of the moving body based on the result.

  According to the present position calculation apparatus according to the present invention, an observation image input unit to which an observation image, which is a two-dimensional image at the current position of the moving body, is input, and a two-dimensional image prepared in advance in association with a predetermined point. A reference image management unit that manages a reference image, a collation unit that collates the observed image with the reference image corresponding to the temporary current position that is a stored current position by image matching, and the collation unit And a current position calculation unit that calculates the true current position of the moving body based on the collation result in the above, so that accumulation of errors is suppressed by using a two-dimensional image at the current position, and the moving body is subjected to image matching. The true current position of can be calculated.

  According to the present position calculation method according to the present invention, (a) a step of inputting an observation image that is a two-dimensional image at the current position of the moving body; and (b) a reference image that is a two-dimensional image related to a predetermined point. (C) collating the observed image with the reference image corresponding to the provisional current position, which is the stored current position, and (d) the step (c) And calculating the true current position of the moving body based on the collation result in the above, thereby suppressing the accumulation of errors by using a two-dimensional image at the current position, and performing the true matching of the moving body by image matching. The current position can be calculated.

It is a figure which shows the structure of the present position calculation apparatus concerning Embodiment 1. FIG. FIG. 3 is a diagram illustrating an example of an observation image according to the first embodiment. FIG. 3 is a diagram illustrating an example of a pixel pattern according to the first embodiment. It is a figure which shows the example of the map data concerning Embodiment 1. FIG. It is a figure which shows the example of communication of the map data concerning Embodiment 1. FIG. FIG. 3 is a diagram illustrating an example of a pixel pattern according to the first embodiment. FIG. 3 is a diagram illustrating configurations of a correction unit and a current position calculation unit according to the first embodiment. 3 is a flowchart illustrating an operation of the current position calculation apparatus according to the first embodiment. 3 is a flowchart illustrating an operation of the current position calculation apparatus according to the first embodiment. 3 is a flowchart illustrating an operation of the current position calculation apparatus according to the first embodiment. 3 is a flowchart illustrating an operation of the current position calculation apparatus according to the first embodiment. 3 is a flowchart illustrating an operation of the current position calculation apparatus according to the first embodiment. It is a figure which shows operation | movement of the present position calculation apparatus concerning Embodiment 1. FIG. FIG. 6 is a diagram for explaining the operation of the current position calculation apparatus according to the first embodiment. FIG. 6 is a diagram for explaining the operation of the current position calculation apparatus according to the first embodiment. FIG. 6 is a diagram for explaining the operation of the current position calculation apparatus according to the first embodiment. FIG. 6 is a diagram for explaining the operation of the current position calculation apparatus according to the first embodiment.

<A. Embodiment 1>
<A-1. Configuration>
FIG. 1 conceptually shows the overall configuration of a current position calculation apparatus according to the present invention. In the following description, a case where the current position of the “vehicle” is calculated will be described. However, the present invention is not limited to the case where the position of the vehicle is calculated. In general, the current position of the moving object is calculated. It is a current position calculation device that can.

  As shown in FIG. 1, the current position calculation device is an observation image to which a two-dimensional image (observation image) obtained by continuously capturing all directions around the vehicle from the vehicle side at the current position of the vehicle as a moving body is input. An input unit 1 and an observation pixel pattern creation unit 2 that creates an observation pixel pattern from an observation image in the observation image input unit 1 are provided. The observation image input unit 1 may further include an observation map data storage unit 3 for adding predetermined attribute data to the observation image and linking and storing it as observation map data on the planar map data. It can also be used via a network or the like.

  Further, for example, on the network, the reference map data storage unit 10 stores reference map data including a two-dimensional image (reference image) captured from the vehicle side at the position, which is linked in advance to a predetermined point on the planar map data. A predetermined reference map data from the reference map data storage unit 10, a reference image management unit 9 that manages the reference image, and a reference pixel pattern that creates a reference pixel pattern from the reference image in the reference image management unit 9 The true current position of the vehicle based on the creation unit 7, the pixel pattern matching unit 5 that matches the observed pixel pattern and the reference pixel pattern by image matching, the positional information by GPS and the matching result in the pixel pattern matching unit 5 Is calculated from the collation result in the current position calculation unit 6 for calculating And a correcting unit 4 that corrects the values.

  The observation image input unit 1 acquires, as an observation image, for example, a panoramic image 18 (around view) as shown in FIGS. 2A and 2B captured by an omnidirectional camera 16 attached to a vehicle or the like. be able to. When observed using the omnidirectional camera 16 from the position of the host vehicle 17 as shown in FIG. 2B, a panoramic image 18 as shown in FIG. 2A can be acquired (FIG. 2A). The panoramic image 18 shown in FIG. 2 is obtained by expanding the panoramic image 18 shown in FIG. A horizontal line 14 is near the center of the image, and road 11, road 12, and road 13 are shown in order.

  The input observation image does not have to be a panoramic image, and may be any observation image of a building or the like within a predetermined range from the vehicle side.

  As shown in FIG. 3, the observation pixel pattern creation unit 2 cuts out a part of the observation image, for example, a horizontal pixel pattern of a predetermined height from the observation image (panoramic image 18), and obtains the observation pixel pattern 21. create. The height of the part to be cut out may be centered on the height position that is most advantageous for position identification, for example, the horizontal line height position with little disturbance (height not related to street cars, people, street-facing etc.) desirable.

  The observation map data storage unit 3 and the reference map data storage unit 10 store map data 102 (observation map data or reference map data) linked to the planar map data 103, for example, as shown in FIG. ing.

  At the time of linking, the reference map data storage unit 10 adds attribute data (for example, latitude and longitude, image map name, data base point angle θ (data acquisition initial angle)) to the reference image (the panoramic image 111 illustrated in FIG. 4B). , Image tilt angle, horizontal line vertical ratio) can be stored. The correspondence relationship between the attribute data and the reference image is as shown in FIG. 4C, for example, where the data base angle θ106, the image tilt angle (pitch φ107, roll ψ108), and the horizontal line vertical ratio 109 of the horizontal line 14 are defined. .

  The observation map data storage unit 3 can also store observation map data obtained by adding the attribute data to the observation image at the time of linking, and the storage format of the observation map data and the reference map data increases versatility. Therefore, it is desirable to be unified. In this way, the observation map data stored in the observation map data storage unit 3 is the map data acquired by the reference image management unit 9 in the same manner as the reference map data stored in the reference map data storage unit 10. Can be used.

  As a specific example of the map data, for example, there is one as shown in JP-T-2010-531007. Corresponding map data can be acquired by accessing a map data server as shown in FIG. 5 storing map data and notifying the temporary current position of the vehicle. The temporary current position will be described later.

  In general, a digital map is made based on a topographic map of the Geospatial Information Authority of Japan, and the format is made in an industry standard Kiwi-W format or the like that conforms to the Japanese Industrial Standard (JIS-D0810). On the Internet, digital map search using Google, Yahoo, etc. is common. Therefore, versatility can also be improved by making observation map data or reference map data into the same storage format as existing Google Street View or the like.

  The reference image management unit 9 acquires predetermined reference map data in the reference map data storage unit 10 and manages the reference image. The coordinate position on the planar map data of the reference map data to be acquired is set based on the temporary current position of the vehicle.

  As shown in FIG. 6, the reference pixel pattern creation unit 7 creates a reference pixel pattern 71 by cutting out a horizontal pixel pattern of a predetermined height, for example, from the reference image. The reference pixel pattern 71 preferably has a horizontal height corresponding to the observation pixel pattern 21 (see FIG. 3), and a cutout width that is a predetermined height is made wider than the corresponding observation pixel pattern 21. It is desirable. This is because an observation error occurring in the observation pixel pattern 21 is taken into consideration when matching between the observation pixel pattern 21 and the reference pixel pattern 71 is performed. On the contrary, the cutout width of the observation pixel pattern 21 can be increased.

<A-2. Operation>
Next, the operation of the current position calculation apparatus according to the present invention will be described.

  FIG. 7 is a configuration diagram specifically showing the correction unit 4 and the current position calculation unit 6 conceptually shown in FIG. In particular, the configuration including the inertial navigation sensor 40 and the in-vehicle microcomputer 50 corresponds to the correction unit 4.

  The inertial navigation sensor 40 includes a speedometer 41 (vehicle speed pulse), an angular velocity sensor 42, and a geomagnetic sensor 43. The position specifying sensor 60 includes a GPS receiving unit 61. The inertial navigation sensor 40 and the position specifying sensor 60 are each connected to the in-vehicle microcomputer 50.

  The position specifying sensor 60 can receive position information by GPS at the GPS receiver 61.

  Further, an auxiliary storage device 51 is connected to the in-vehicle microcomputer 50.

  As shown in FIG. 8, the in-vehicle microcomputer 50 performs initialization for executing the program (step S1), and determines whether or not there is a key input from the user (step S2). If there is a key input, an input operation corresponding to the key input is performed (step S3). If there is no key input, for example, a map around the current position of the vehicle can be displayed on the display screen (step S4).

  The current position calculation apparatus according to the present invention performs an operation of calculating the current position of the vehicle while performing operations such as screen display. The overall flow of the current position calculation process will be described below.

  First, the GPS receiving unit 61 of the position specifying sensor 60 acquires GPS position information. Based on the acquired position information and the positioning data of the inertial navigation sensor, the current position of the vehicle is specified. At this time, in order to increase the accuracy of position identification, it is also possible to use an average GPS positioning position that is an average value of position information acquired a plurality of times by GPS.

  Specifically, as shown in FIG. 9, first, position information by GPS is acquired (step S9). Next, the positioning position (latitude and longitude) of the vehicle is specified from the acquired position information, and it is determined whether the GPS positioning is successful. Specifically, (1) whether the coordinate difference between the positioning position of the vehicle specified by GPS and the current position of the vehicle determined last time is within a predetermined range, or (2) the σ value of the average GPS positioning position Is determined based on whether the reception level is not less than the specified value, and (3) whether the reception level is not more than the specified value (step S10). Here, the current position of the vehicle determined last time is the current position of the vehicle recognized immediately before performing this operation, and is stored in the current position calculation unit 6. The closer the recognized time is, the more accurate the position information is. Further, the current position of the vehicle determined last time stored in the current position calculation unit 6 is used as a temporary current position in an image matching process described later.

  If it is determined that the GPS positioning is successful, the process proceeds to step S11. If it is determined that the GPS positioning is not successful, the process proceeds to step S12.

  In step S11, the positioning position of the vehicle specified by GPS is set and stored as the true current position of the vehicle.

  In step S12, the process proceeds to image collation processing. The image matching process will be described in detail later.

  The current position of the vehicle determined by the GPS receiver 61 is corrected using a speedometer 41 (vehicle speed pulse) as a distance sensor of the inertial navigation sensor 40, an angular velocity sensor 42 (gyro sensor) as a direction sensor, and a geomagnetic sensor 43, Set the current position of the vehicle.

  Specifically, as shown in FIG. 10, the travel distance is calculated using the speedometer 41 (step S30). Next, the azimuth is calculated using the angular velocity sensor 42 and the geomagnetic sensor 43 (step S31).

  Next, the current position of the vehicle determined by the GPS receiver 61 is corrected using the calculated travel distance and direction, set as the true current position of the vehicle, and stored (step S32).

  Next, an operation of image collation processing performed under a situation where it is difficult to acquire position information by GPS will be described.

  The following operations are performed in urban areas where foreign buildings and the like are lined with high-rise buildings where reception of radio waves from GPS satellites is difficult. Similarly, in tunnels, underground and multistory parking lots, large buildings, etc. where it is difficult to receive radio waves from GPS satellites, the following operations are possible if conditions are met.

  When the GPS positioning fails in the processing flow for calculating the current position shown in FIG. 9 and the process proceeds to the image collating process in step S12, first, as shown in FIG. Information on the temporary current position, which is the current position set in the calculation process, is acquired (step S60).

  Next, the observation image input unit 1 acquires an observation image (step S61), adds attribute data to the acquired observation image, and stores it as observation map data in the observation map data storage unit 3 (step S62). The observation pixel pattern creation unit 2 creates, for example, a horizontal observation pixel pattern from the acquired observation image (step S63). Note that the operation of adding the attribute data to the observation image and storing it is not an essential operation.

  On the other hand, the reference image management unit 9 acquires reference map data corresponding to the temporary current position in the reference map data storage unit 10 from the information on the temporary current position obtained from the current position calculation unit 6, and further, the reference map data To obtain a reference image (step S64). Then, the reference pixel pattern creation unit 7 creates, for example, a horizontal reference pixel pattern from the acquired reference image (step S65).

  Next, the pixel pattern matching unit 5 performs matching processing between the observed pixel pattern and the reference pixel pattern (step S66).

  Here, the operation of the pixel pattern matching process will be described with reference to FIG.

  First, the pixel pattern matching unit 5 performs pixel pattern matching (image matching) between the observed pixel pattern and the reference pixel pattern (step S79).

  The pixel pattern matching is performed by dividing the vehicle into four parts, for example, a right front part, a right rear part, a left front part, and a left rear part, as shown in FIG. Based on the azimuth information included in the reference pixel pattern 71 arranged to surround the vehicle and the traveling direction of the vehicle, the front matching start point 100 and the rear matching start point 101 are set before and after the vehicle traveling direction (FIG. 13 ( a)).

  Also in the observed pixel pattern 21 (see FIG. 13C), the front matching start point 100 and the backward matching start are made forward and backward in the vehicle traveling direction in correspondence with the reference pixel pattern 71 (see FIG. 13B). Point 101 is set.

  For example, when matching is performed using the right front matching transition 52, the direction (circumference) of the right front matching transition 52 is based on the front matching start point 100 of the observation pixel pattern 21 and the front matching start point 100 of the reference pixel pattern 71. (Clockwise in the direction) by pixel pattern matching to reach the final right matching position 112 of the observed pixel pattern 21 (a position 90 ° clockwise from the front matching start point 100). At this time, a position matched with the observed pixel pattern 21 in the reference pixel pattern 71 is set as a right front matching final position 57.

  Similarly, a right rear matching final position 59 is obtained in the direction of the right rear matching transition 54 (counterclockwise in the circumferential direction) from the rear matching start point 101 as a base point, and further, the direction of the left front matching transition 53 and the left rear matching transition Also in each of the directions of 55, the observation pixel pattern 21 is made to reach the final left matching position 113 (position 90 ° counterclockwise from the front matching start point 100), and the position matched with the observation pixel pattern 21 is left front. The matching final position 56 and the left rear matching final position 58 are obtained.

  In this embodiment, the matching final position of the observation pixel pattern 21 is fixed and matching is performed. However, the matching final position of the reference pixel pattern 71 may be fixed and matching may be performed.

  In this embodiment, the pixel pattern of both the observation image and the reference image is cut out and matched. However, both the observation image and the reference image or one of them is not cut out of the pixel pattern, and the entire image is matched. It is also possible to do this.

  In this method, since it is possible to obtain a correspondence relationship for each pixel while maintaining monotonous continuity between pixels, the positions of the reference pixel pattern 71 and the observation pixel pattern 21 to be compared are slightly shifted, and the image is Even if they are different, it is desirable because highly stable matching can be performed.

  On the plane map data, a left-front matching final position 56, a right-front matching final position 57, and a left-back matching final position 58 from the vehicle center with a line in the direction perpendicular to the traveling direction of the vehicle passing through the vehicle center as a reference line. The angles formed with respect to each of the right rear matching final positions 59 are a data base point angle θLF, a data base point angle θRF, a data base point angle θLB, and a data base point angle θRB, respectively (see FIG. 13A).

  The data base point angle θ (data base point angle θRF, data base point angle θRB, data base point angle θLF, data base point angle θLB) is specified by the method as described above (steps S80 to S83). The specific order is not limited to the case shown in the flowchart.

  Next, it is determined whether or not the data base point angle θ is valid (step S84). Whether or not the data is valid is determined, for example, based on whether or not each data base angle θ is within a range of ± 60 °. (1) When all four data base point angles are within a range of ± 60 ° (when all data base point angles θ are valid), or (2) Data base point angle θ on the right side of the vehicle (data base point angle θRF) And the data base point angle θRB) are within a range of ± 60 ° (when the data base point angle θR on the right side of the vehicle is valid), or (3) the data base point angle θ on the left side of the vehicle (the data base point angle θLF and the data If the base point angle θLB) is within a range of ± 60 ° (when the data base point angle θL on the left side of the vehicle is valid), the process proceeds to step S85. On the other hand, in other cases (when the data base point angle θR on the right side of the vehicle and the data base point angle θL on the left side of the vehicle are invalid), the process proceeds to step S91.

  Here, FIG. 14 shows a case where the data base angle θ is within a range of ± 60 °.

  First, it is assumed that the reference pixel pattern to be compared with the observed pixel pattern is linked on the plane map at an interval of about 10 m in order to suppress defects due to the blind spot of the camera.

  In addition, the observation pixel pattern 21 collects positions such as the height of a horizontal line with little disturbance, but a characteristic standing tree or building that can be matched in the image regardless of the height of the horizontal line is reflected in the image. Matching is possible. The assumed line of the position of the fixed object parallel to the road capable of image matching is set as an image collection prediction line.

  If the road width is 10 m and the distance between the image collection prediction lines of the observation pixel pattern 21 is 20 m, the distance from the vehicle located at the center of the road to the image collection prediction line of the observation pixel pattern 21 in the direction perpendicular to the traveling direction is 10m.

  As shown in FIG. 14, the point on the image collection prediction line of the observation pixel pattern 21 where the data base point angle (data base point angle θRF or data base point angle θRB) viewed from the vehicle is + 30 ° is approximately 6 m ahead of the vehicle. . Similarly, the point on the image collection prediction line of the observation pixel pattern 21 where the data base angle when viewed from the vehicle is + 60 ° is approximately 17 m ahead of the vehicle.

  In the range (correctable area) up to ± 6 m in the longitudinal direction of the vehicle, it is assumed that there is one reference pixel pattern (only C in FIG. 14), and the data base point angle θ is based on the one reference pixel pattern. Correct the position of the vehicle in the direction of travel.

For example, as shown in FIG. 15, when the distance between the image collection prediction lines of the observed pixel pattern 21 is 2 W and the data base point angle θ is as shown in the figure, as shown in FIG.
Correction value Δh = W × tan (θ)
Can be calculated as When the road width is unknown, the calculation is made with W = 10 (m), for example. By performing the correction, the positioning position of the vehicle can be obtained.

  Similarly, in the range up to ± 17 m in the vehicle front-rear direction (retractable area), there are four reference pixel patterns (A to D in FIG. 14), and data is based on the four reference pixel patterns. The position of the vehicle in the traveling direction is corrected based on the base angle θ.

  Specifically, position correction is performed as shown in FIG. When the reference pixel patterns 71 of A to D shown in FIG. 14 are coordinate positions as shown in FIG. 16A, the average angle of the data base angle θRF and the data base angle θRB, or the data base angle θLF and the data The average angle with the base point angle θLB changes as shown in FIG.

  Since the coordinate position where these average angles are 0 ° is the positioning position of the vehicle, the positioning position of the vehicle can be obtained from the transition of the average angles A to D.

  When the data base angle θ is outside the range of ± 60 °, the matching visual field end is in a range exceeding ± 17 m in the vehicle front-rear direction, and pixel pattern matching with more reference pixel patterns is performed before the positioning position is calculated. There is a need. Further, the transition width in each pixel pattern matching is short, and the matching accuracy is lowered. Therefore, it is effective when the data base angle θ is within a range of ± 60 °, for example. Note that the angle range can be changed according to the reference pixel pattern creation interval, matching calculation speed, and the like.

  In step S85, as described above, whether the average angle between the data base point angle θRF and the data base point angle θRB or the average angle between the data base point angle θLF and the data base point angle θLB is within a range of ± 30 °. Determine whether. When the average angle is within the range of ± 30 °, the process proceeds to step S90. And based on the said average angle, position correction of the advancing direction of a vehicle is performed and a positioning position is calculated | required (refer FIG. 15). If the average angle is outside the range of ± 30 °, the process proceeds to step S86.

  In step S86, it is determined whether the average angle is greater than 30 ° and not greater than 60 °. If so, the process proceeds to step S87. If not, that is, if the average angle is not less than −60 ° and less than −30 °, the process proceeds to step S88.

  In step S87, reference pixel patterns are acquired in order in the vehicle traveling direction (forward), and the matching operation is repeated until the average angle becomes a negative value or three reference pixel patterns are acquired. When this operation ends, the process proceeds to step S89.

  In step S88, the reference pixel pattern is acquired in the reverse direction (backward) with respect to the traveling direction of the vehicle, and the matching operation is repeated until the average angle becomes a positive value or three reference pixel patterns are acquired. When this operation ends, the process proceeds to step S89.

  In step S89, it is determined whether or not the repetitive matching operation has been properly completed. If the operation has been properly completed, the process proceeds to step S90, and if not, the process proceeds to step S91.

  In step S90, the positioning position of the vehicle is obtained from the transition of the average angle (see FIG. 16B). In step S91, the collation failure flag is set and the process ends.

  Next, the current position calculation unit 6 determines whether or not the collation between the observed pixel pattern and the reference pixel pattern is successful (step S67 in FIG. 11). If the collation is successful, the collation is performed to step S68. If it fails, the image collation process is terminated and the process returns to the current position calculation process.

  In step S68, the positioning position is set and stored as the true current position, and the process returns to the current position calculation process.

  FIG. 17 shows an example of the above pixel pattern matching.

  A reference pixel pattern 201 (including an observed pixel pattern 200 (see FIG. 17A) in which the pixel position can be specified by the m coordinate and the n coordinate and a pixel pattern similar to the observed pixel pattern 201 can be specified by the i coordinate and the j coordinate. FIG. 17B) is prepared.

  The correspondence of each pixel in the similar image pattern is examined in order by shifting the matching from the matching start point to the final matching position (see FIG. 17C).

  A reference pixel pattern corresponding to the final matching position 202 on the observed pixel pattern 200 is obtained by calculating the minimum cumulative distance that minimizes the cumulative value of the image position and the distance between the images by sequentially accumulating the minimum cumulative distance. The final matching position on 201 is obtained. In this way, it is possible to obtain the correspondence for each pixel while maintaining monotonic continuity.

  Here, it takes some time for the above pixel pattern matching operation, and when the position of the vehicle can be known, it is delayed from the time when the observation image was acquired, and the position of the vehicle has changed. Can be considered.

  In such a case, using a speedometer 41 (vehicle speed pulse) as a distance sensor of the inertial navigation sensor 40, an angular velocity sensor 42 as a direction sensor, and a geomagnetic sensor 43, a change in the position of the vehicle during the pixel pattern matching operation ( It is possible to correct the current position by calculating a movement distance and a movement direction based on the current position calculated by image matching. In other words, the position (current position) where the observation image is created, calculated by pixel pattern matching, is the position (true current position) where the vehicle is actually reached when the information is output from the current position calculation unit 6. ) Can be corrected.

  Here, when the vehicle is not moving, the position where the observation image was created (current position) and the position where the vehicle actually reached (true current position) are the same position. There is no need for such amendment.

  It is also possible to adjust the position by map data matching with respect to the current position of the vehicle obtained by pixel pattern matching.

  The pixel pattern matching operation described above may be performed in parallel even in a situation where position information can be acquired by GPS or the like. If there is a difference between the true current position of the vehicle calculated by the operation and the current position calculated from position information by GPS or the like, the current position calculated by GPS may be used as a reference. Good. However, the pixel pattern matching operation can be stopped if there is a problem such as the calculation speed being affected by the pixel pattern matching operation.

  Based on the true current position of the vehicle calculated in this way, route guidance to the destination can be further developed.

<A-3. Effect>
According to the embodiment of the present invention, in the current position calculation device, an observation image input unit 1 to which an observation image that is a two-dimensional image at the current position of the vehicle 17 as a moving body is input, and a predetermined point Image matching is performed between a reference image management unit 9 that manages a reference image that is a two-dimensional image prepared in advance, an observation image, and a reference image corresponding to a temporary current position that is a stored current position. The current position is provided with a pixel pattern matching unit 5 as a matching unit for matching by the above and a current position calculating unit 6 for calculating the true current position of the host vehicle 17 based on the matching result in the pixel pattern matching unit 5. The accumulation of errors can be suppressed using the two-dimensional image (observation image) at, and the true current position of the vehicle 17 can be calculated by image matching.

  Further, according to the embodiment of the present invention, in the current position calculation device, the temporary current position is determined based on the known position information in the own vehicle 17 as the moving body, so that it is recognized immediately before. The reference image can be acquired based on the current position.

  Further, according to the embodiment of the present invention, in the current position calculation device, the observation image and the reference image are panoramic images that display the surroundings of the vehicle 17 as a moving body, so that all surrounding directions can be obtained. The true current position of the vehicle 17 can be calculated by performing image matching with continuously captured images.

  Further, according to the embodiment of the present invention, in the current position calculation device, the pixel pattern matching unit 5 serving as the matching unit includes the front and the front in the traveling direction of the host vehicle 17 serving as the moving body in the observation image and the reference image. Image matching is performed using the back as the base point, and the verification result is generated based on the end position (matching final position) of the image matching, thereby valid / invalid of the data depending on the size of the data base point angle based on each final matching position And the presence / absence of further image matching, and the true current position of the vehicle 17 can be calculated.

  Further, according to the embodiment of the present invention, in the current position calculation device, the pixel pattern matching unit 5 as a matching unit includes an observed image and a plurality of reference images corresponding to a predetermined range from the temporary current position. Then, matching is performed by image matching, and the current position calculation unit 6 calculates the true current position of the vehicle 17 as a moving body based on the matching result using a plurality of reference images in the pixel pattern matching unit 5. Even if the current position of the vehicle 17 is more than a predetermined distance from the temporary current position, the true current position of the vehicle 17 is calculated by using a reference image associated with the temporary current position and the like. can do.

  Further, according to the embodiment of the present invention, in the current position calculation device, at least one of the observation image and the reference image extracts a pixel pattern that is a part of the image from the image and is used for collation. Image matching can be facilitated.

  Further, according to the embodiment of the present invention, in the current position calculation device, the pixel pattern has a predetermined height around the horizontal line height in the image, so that an image with less disturbance can be used for comparison. it can.

  Further, according to the embodiment of the present invention, in the current position calculation device, an observation map data storage unit that stores observation images by adding attribute data that is data relating to the current position of the vehicle 17 as a moving body. 3 is further provided, the observation map data is stored in the observation map data storage unit 3, and the image can be referred to for image matching together with the reference map data.

  Further, according to the embodiment of the present invention, in the current position calculation device, the current position calculation unit 6 moves the current position of the host vehicle 17 as the moving body calculated based on the collation result. By further including a correction unit 4 that corrects based on the distance and the moving direction and calculates the true current position, the current position calculated by image matching is different from the current position that the vehicle has actually reached. Even in such a case, the true current position can be appropriately calculated from the moving distance and moving direction of the host vehicle 17.

  According to the embodiment of the present invention, in the current position calculation method, (a) a step of inputting an observation image that is a two-dimensional image at the current position of the vehicle 17 as a moving body; A step of preparing a reference image which is a two-dimensional image related to the point of (c), a step of collating the observed image with a reference image corresponding to the temporary current position by image matching, and a step (d) (c ) To calculate the true current position of the vehicle 17 based on the result of the comparison in the above), thereby suppressing the accumulation of errors by using a two-dimensional image at the current position, The true current position can be calculated.

  It should be noted that the present invention can be modified or omitted in any constituent elements in the present embodiment within the scope of the invention.

  DESCRIPTION OF SYMBOLS 1 Observation image input part, 2 Observation pixel pattern creation part, 3 Observation map data storage part, 4 Correction part, 5 Pixel pattern collation part, 6 Current position calculation part, 7 Reference pixel pattern creation part, 9 Reference image management part, 10 Reference map data storage unit, 11-13 road, 14 horizon line, 16 omnidirectional camera, 17 own vehicle, 18,111 panoramic image, 21,200 observation pixel pattern, 40 inertial navigation sensor, 41 speedometer, 42 angular velocity sensor, 43 Geomagnetic sensor, 50 In-vehicle microcomputer, 51 Auxiliary storage device, 52 Right front matching transition, 53 Left front matching transition, 54 Right rear matching transition, 55 Left rear matching transition, 56 Left front matching final position, 57 Right front matching final Position, 58 Left rear Ching final position, 59 Right rear matching final position, 60 Position specifying sensor, 61 GPS receiving unit, 62 Cellular communication unit, 63 Cellular base station, 64 Base station positioning dedicated server, 71, 201 Reference pixel pattern, 100 Forward matching start point , 101 Back matching start point, 102 Map data, 103 Planar map data, 106 Data base angle θ, 107 Pitch φ, 108 Roll ψ, 109 Horizontal line vertical ratio, 112 Right matching final position, 113 Left matching final position, 202 Final matching position.

Claims (14)

An observation image input unit to which an observation image that is a two-dimensional image at the current position of the moving object is input;
A reference image management unit that manages a reference image that is a two-dimensional image prepared in advance in association with a predetermined point;
A collation unit that collates the observed image with the reference image corresponding to the temporary current position, which is the stored current position, by image matching;
A current position calculation unit that calculates a true current position of the mobile body based on a verification result in the verification unit;
Current position calculation device. The observation image and the reference image are panoramic images that display the periphery of the moving body,
The current position calculation apparatus according to claim 1. The matching unit performs the image matching in the observation image and the reference image with reference to the front and rear in the traveling direction of the moving body, and generates the matching result based on the end position of the image matching. Characterized by the
The current position calculation apparatus according to claim 2. The collation unit collates the observed image with a plurality of the reference images corresponding to a predetermined range from the temporary current position by the image matching,
The current position calculation unit calculates the true current position of the mobile body based on the verification result using the plurality of reference images in the verification unit.
The current position calculation device according to claim 1. At least one of the observation image and the reference image is used for the collation by extracting a pixel pattern that is a part of the image from the image,
The present position calculation apparatus in any one of Claims 1-4. The pixel pattern has a predetermined height around a horizontal line height in an image,
The current position calculation apparatus according to claim 5. A correction unit that corrects the current position of the moving body calculated based on the collation result based on a moving distance and a moving direction of the moving body, and calculates the true current position; Further comprising:
The current position calculation apparatus according to claim 1. (A) inputting an observation image that is a two-dimensional image at the current position of the moving object;
(B) preparing a reference image which is a two-dimensional image related to a predetermined point;
(C) collating the observed image with the reference image corresponding to the temporary current position that is the stored current position by image matching;
(D) comprising a step of calculating a true current position of the mobile body based on a collation result in the step (c).
Current position calculation method. The observation image and the reference image in the steps (a) and (b) are panoramic images that display the periphery of the moving body,
The current position calculation method according to claim 8. The step (c) performs the image matching on the observation image and the reference image with reference to the front and rear in the moving direction of the moving body, and generates the matching result based on the end position of the image matching It is a process to perform,
The current position calculation method according to claim 9. The step (c) is a step of collating the observed image with a plurality of the reference images corresponding to a predetermined range from the temporary current position by the image matching,
The step (d) is a step of calculating the true current position of the moving body based on the collation result using the plurality of reference images in the step (c).
The current position calculation method according to claim 8. At least one of the observed image and the reference image in the steps (a) and (b) is used for the collation by extracting a pixel pattern that is a part of the image from an image,
The current position calculation method according to any one of claims 8 to 11. The pixel pattern has a predetermined height around a horizontal line height in an image,
The current position calculation method according to claim 12. In the step (d), the current position of the moving body calculated based on the collation result in the step (c) is corrected based on a moving distance and a moving direction of the moving body, and the true current position is determined. It is a process of calculating,
The current position calculation method according to any one of claims 8 to 13.

Images