JP6616967B2 - Map creation apparatus and map creation method - Google Patents

Description

  The present invention relates to a map creation device and a map creation method.

  Conventionally, Patent Document 1 discloses a map creation device that creates a map along a road using a photographed image obtained by continuously photographing the road surface and scenery on both sides of the road along the road length, and photographing position / direction information. Is known.

  In this conventional example, the map is created by first extracting feature points from the all-around video acquired from the vehicle carrying the all-around camera, tracking it to adjacent frames, and then moving the camera to the base. And the tracking point are calibrated, and the tracking data is analyzed to obtain an all-round image having the three-dimensional coordinates of the feature points, the camera position, and the posture. Next, from this all-round image, the coordinates are narrowed down based on the camera position and orientation, and a three-dimensional analysis is performed to extract a geometric simple figure consisting of points, lines, and end points, and from the tracking data, Based on the same principle, the three-dimensional position and orientation are detected and the three-dimensional coordinates are given. Since these coordinates are relative coordinates, absolute coordinates are added by post-processing from known data on the map or high-precision GPS data.

  After that, by comparing with the patterns prepared in advance, the geometric simple figures described above are classified and registered attributes such as road markings, road signs, traffic lights, and the like. Reconstruct the 3D space corresponding to the 3D space. Depending on the purpose, other necessary information is integrated and can be expressed from a bird's-eye view or from the viewpoint of the vehicle, that is, with data that can be converted into all directions and all viewpoints. The 3D machine map is completed.

Japanese Patent No. 5227005

  However, the map in the conventional example described above has a lot in common with the actual view when the viewpoint is adjusted to a three-dimensional composition such as a bird's-eye view, but it is easy to grasp the geography intuitively, The geographical range that can be grasped and displayed at a time will be narrowed, and the listability will be reduced, and when it is adjusted to a flat composition like a flat map, the listability will improve, but on the other hand it will be a real view. There is a disadvantage that the number of common points is extremely small and it is difficult to intuitively grasp the geography.

  In addition, after extracting a geometric simple figure from the photographed image and reconstructing a three-dimensional space from the geometric simple figure, or adding absolute coordinates to the relative coordinates by post-processing, there is an error. It is easy to accumulate, and this makes it difficult to intuitively grasp geography even if accuracy is lowered.

  The present invention has been made to solve the above drawbacks, making it easy to intuitively grasp the geography, and creating a map along the road with improved listability, And to provide a mapping method.

According to the present invention, the object is
A map creation device for creating a map along a road using a photographed image 3 obtained by continuously photographing the road surface 1 and both sides of the road along the road length, and photographing position / azimuth information,
A photographed image 3 is generated by joining the original images 12 simultaneously photographed by the camera 11 that shoots the road surface 1 and both sides of the road separately from the same photographing position using the photographing position / azimuth information, and the road surface 1 of the road An individual image generation unit 8 that generates a mosaic image 6 photographed in the vertical direction and a mosaic image 7 photographed in the horizontal direction on both sides of the road with respect to the road length of the road ;
A developed image in which a mosaic image 7 in the horizontal shooting direction with the position at the lower end is connected on the same plane at positions corresponding to both sides of the road in the mosaic image 6 in the vertical shooting direction to create a photograph map approximating a developed view. Generating unit 10;
The road intersection 15 is extracted from the photographed image 3 or the original image 12, an interpolation image 16 is generated from a branch photographing region of the intersection 15 in the photographed image 3 or the original image 12, and the interpolation image 16 is converted into the vertical image 16. This is achieved by providing a map creation device having an intersection adjustment unit that joins to the branch line corresponding position of the mosaic image 6 in the photographing direction .

  According to the present invention, the map creation device performs absolute orientation and mosaic processing by block adjustment including the shooting position / orientation information at the pass point 4 set in the shot image 3 and uses the three-dimensional model 5. By performing parallel projection conversion such as orthogonal transformation, the mosaic image 6 in the vertical shooting direction and the mosaic images 7 and 7 in the horizontal shooting direction on both sides of the road are generated with extremely high accuracy. Further, by connecting the mosaic images 6, 7, and 7 in the vertical and horizontal photographing directions thus positioned on the same plane with the road as a reference, a photographic map 9 approximating the developed view is generated.

  The obtained photographic map 9 has a lot in common with the actual view, so it is easy to grasp the geography intuitively, and because the geography is virtually reproduced under a planar composition. , Excellent listability. Further, since the mosaic images 6 and 7 in the vertical and horizontal shooting directions are generated from the shot image 3 by a strict method, the accuracy is excellent. For example, the bundle method can be used for the block adjustment described above, and the Shift method or the Harris method can be used for setting the pass point 4. In order to increase the position accuracy, it is desirable to use a ground reference point for block adjustment.

  Therefore, according to the present invention, for example, a user-friendly map as described above is obtained by inputting data acquired by mobile mapping using a measurement vehicle equipped with a camera, a laser range finder, GPS, IMU, and the like. be able to.

Also, the wide-angle image that generates the captured image 3 by joining the original images 12 and 12 that are simultaneously captured by the camera 11 that divides and captures the entire zenith direction from the same capturing position using the capturing position and orientation information A generator 13;
When the individual image generating unit 8 generates the mosaic images 6 and 7 based on the captured image 3 or the original image 12, the above-described map is generated using a so-called all-around image that is frequently used for mobile mapping. Compared to the case, since there is no missing area in the shooting range in the elevation angle direction, it is possible to reliably reflect high-rise buildings along the road on the map without any omissions, and the realism of the map is greatly improved. For example, when generating the above-described mosaic image 7 in the horizontal shooting direction, the reproducibility can be further improved by using the original image 12 in the shooting direction close to the horizontal shooting direction. For example, the camera 11 that divides and shoots the original image 12 may be a camera provided with a wide-angle lens or a fish-eye lens in addition to a standard lens.

  Further, when generating the mosaic images 6 and 7 in the vertical and horizontal shooting directions by the individual image generation unit 8, when the projection-type shot image 3 shot using the fisheye lens 14 is subjected to parallel projection conversion processing in the vertical and horizontal directions. Since the image is not distorted compared to the central projection image, an easy-to-view map can be obtained. Further, by using the fisheye lens 14, the photographing range can be made extremely wide, thereby reducing the number of original images 12 for simultaneously photographing all zenith directions. Can reduce the load. The above-described joining of the photographed images 3 and 3 using the fisheye lens 14 can associate each tie point with an epipolar line, for example.

  In addition, the intersection 15 is extracted based on the photographed image 3 or the three-dimensional model 5, the branch line photographing region of the intersection 15 in the photographed image 3 or the original image 12 is processed to generate an interpolation image 16, and When the intersection adjustment unit 17 for pasting the interpolated image 16 so as to be connected to the branch line corresponding position of the mosaic image 6 in the vertical photographing direction is present, the existence of the intersection 15 can be grasped better by the map. be able to. In addition, by pasting the interpolation image 16, an unnatural display dropout appears at the branch line corresponding position of the mosaic image 7 in the horizontal shooting direction, and the presence of the entire map including the presence of the branch line is improved. In addition, the above-described extraction of the intersection 15 is performed by searching for discontinuous portions of the surface virtually constituted by the surfaces of the buildings on both sides of the road in the photographed image 3 and the three-dimensional model 5, pattern matching, or contour tracking. 15 or a region estimated to be a branch line, and the interpolation image 16 includes, for example, an orthographic projection image of a region in the vicinity of the branch line intersection 15 like the road surface 1, An image obtained by processing a shooting area in the horizontal shooting direction of the branch line in the original image 12 can be used.

Further, according to the present invention, the above-described object is
A map creation method for creating a map along a road using a photographed image 3 obtained by continuously photographing the road surface 1 and scenery on both sides of the road along the road length, and photographing position / direction information,
A photographed image 3 is generated by joining the original images 12 simultaneously photographed by the camera 11 that shoots the road surface 1 and both sides of the road separately from the same photographing position using the photographing position / azimuth information, and the road surface 1 of the road Generating a mosaic image 6 photographed in the vertical direction and a mosaic image 7 photographed in the horizontal direction on both sides of the road with respect to the road length of the road;
Creating a photographic map approximating a developed view by connecting mosaic images 7 in the horizontal shooting direction with the lower end at the position corresponding to both sides of the road in the mosaic image 6 in the vertical shooting direction on the same plane; ,
The road intersection 15 is extracted from the photographed image 3 or the original image 12, an interpolation image 16 is generated from a branch photographing region of the intersection 15 in the photographed image 3 or the original image 12, and the interpolation image 16 is converted into the vertical image 16. This is achieved by providing a map generation method including the step of connecting to the branch line corresponding position of the mosaic image 6 in the photographing direction .

Furthermore, the captured image 3 that is input to the above-described map creation device or used in the above map creation method is:
A low-altitude imaging camera 11A that captures a road surface 1 of a road and a landscape 2 including a relatively low-altitude area of a high-rise building 18 beside the road within a predetermined elevation angle range;
A mobile having a shooting range of the low-altitude shooting camera 11 and a high-altitude shooting camera 11B that takes a comparatively high-rise area of a high-rise building 18 on the roadside with a shooting range as a shooting range that overlaps at the outer peripheral edge. It can be obtained using a mapping measuring device.

This mobile mapping measuring apparatus divides and shoots the entire zenith direction, in other words, the elevation angle direction including the zenith, by the low altitude imaging camera 11A and the high altitude imaging camera 11B. As a result, even a relatively high-rise area of the high-rise building 18 along the road is surely photographed without omission. Note that the low altitude shooting camera 11A and the high altitude shooting camera 11B may be separate cameras 11 or may be a single housing camera 11 that divides and shoots a relatively low altitude region and a relatively high layer region. It is also possible to configure.

Therefore, it is possible to acquire a captured image 3 obtained by continuously shooting the road surface 1 and the roadside scenery 2 along the road length that is input to the above-described map creation device or used in the map creation method. Since there is no missing area in the shooting range in the direction, the entire view of the high-rise building 18 along the road can be reflected in the map obtained by the map creation device or the like, and the presence can be enhanced.

  The relatively high-rise area of the high-rise building 18 along the road measures, for example, the lane closest to the roadside when the road on which the measurement vehicle is traveling is one lane, one lane on one side, or two or more lanes on one side. When the vehicle travels, the vehicle can easily go out of the shooting range of the low altitude shooting camera 11A. For this reason, for example, a camera control unit is provided in the above-described mobile mapping measurement device, and images taken by the low altitude shooting camera 11A are sequentially analyzed. If the shooting by the high altitude shooting camera 11B is stopped when the conditions such as not being in a near lane are satisfied, unnecessary shooting by the high altitude shooting camera 11B, unnecessary joining processing for the shot image 3, etc. Can be prevented. Moreover, it replaces with the above driving lanes, and the distance to the roadside is sufficient as a control criterion for the high altitude photographing camera 11B.

  As is apparent from the above description, according to the present invention, a map creation device capable of creating a map along a road that makes it easy to intuitively grasp the geography and improves the listability, and a map A creation method can be provided, and a mobile mapping imaging device that can be used to collect data for creating the map can be provided, whereby a pedestrian or vehicle using the road can be provided. You can easily reach your destination without making a mistake.

It is a hardware block diagram of the map creation apparatus which concerns on this invention. It is a figure which shows a measurement vehicle, (a) is a side view explaining a measurement range and a measurement state, (b) is a top view. It is a figure explaining a camera, (a) is a perspective view, (b) is a figure which shows the imaging | photography range of a horizontal direction, (c) is a figure explaining the synthetic | combination state of the original image image | photographed separately. FIG. 6 is a diagram for explaining geometric correction of an original image, in which (a) is a diagram for explaining an error that occurs in a single photograph due to a change in the principal point position of the lens, (b) is an image diagram of geometric correction, and (c) is an image of the lens. It is a figure explaining the error which arises in a three-dimensional photograph with the change of a principal point position. It is a figure explaining the production | generation work of the photographic map from the picked-up image of a zenith direction, (a) is a figure which shows the picked-up image of a zenith direction, (b) is a figure explaining the mechanism of a block adjustment. It is a figure explaining the production | generation work of the photograph map from the picked-up image of all zenith directions, (a) is a figure explaining the extraction state of the pass point in the picked-up image of all zenith directions, (b) is extracted from the picked-up image, It is a figure which shows the image which caught the scenery by the side of a road from the front almost. It is a figure explaining the production | generation work of the photographic map from the picked-up image of all the zenith directions, (a) is a figure which shows the parallel projection mosaic image of a horizontal direction, (b) is a figure which shows the parallel projection mosaic image of a vertical direction, c) is an image diagram showing the three-dimensional model from the vertical direction. It is a figure explaining the production | generation work of the photographic map from the picked-up image of all the zenith directions, and is a figure which shows the state which affixed the parallel projection mosaic image of the horizontal direction on the parallel projection mosaic image of the vertical direction. It is a figure explaining the production | generation work of the photographic map from the picked-up image of all the zenith directions, (a) is a figure which shows the image which caught the branch line of the intersection from the substantially front, (b) is the production | generation image figure of an interpolation image. It is a figure explaining the production | generation work of the photographic map from the picked-up image of all the zenith directions, and is a figure which shows the completed photographic map.

  1 to 10 show an embodiment of the present invention. FIG. 2A shows an outline of the measurement work on the map creation target road. The measurement work is performed by running the measurement vehicle 20 on the road surface 1, that is, by mobile mapping. The measurement vehicle 20 is equipped with a mobile mapping measurement device A.

  As shown in FIGS. 2A and 2B, the mobile mapping measuring device A includes a GNSS receiver 21, an IMU 22, a laser distance measuring device 23, a camera 11, and a gantry 24 to which these are attached. And an odometer 25. As shown in FIG. 2B, the gantry 24 is installed on the roof of the measurement vehicle 20, and the odometer 25 is mounted in the measurement vehicle 20 so as to be able to measure the rotation angle of the wheels 20a. Further, on the gantry 24, the three antennas 21a of the GNSS receiver 21 described above are spaced apart, the IMU 22 and the camera 11 are disposed in the center portion in plan view, and the laser distance measuring device 23 is disposed in front of the vehicle and A total of two units are arranged on the rear side. As a result, the laser distance measuring device 23 can measure a distance in a height range from the road surface 1 to an appropriate ground height over almost the entire circumference of the measurement vehicle 20. In addition, it is also possible to use a laser distance measuring device 23 that can measure the entire circumference alone instead of the above one.

  In addition, as shown in FIGS. 2A, 3A, and 3B, the camera 11 is an omnidirectional camera 11 (low-level camera) that can capture the zenith direction, that is, the zenith direction in addition to the entire circumference. The altitude range shooting camera 11A and the high altitude range shooting camera 11B are integrated) and are supported by the stand 24a at a position where the line of sight is good above the measuring vehicle 20. The omnidirectional camera 11 divides and shoots the surrounding direction (low altitude region) into five parts, and further shoots all zenith directions at the same time by a total of six parts including photographing in the zenith direction (high altitude region). The adjacent divided photographing ranges in the peripheral direction, and the respective photographing ranges in the peripheral direction and the photographing range in the zenith direction overlap with each other at an appropriate overlap rate. The above-described camera 11 has a very wide angle of view by using the fish-eye lens 14, thereby enabling photographing of the height range from the zenith to the road surface 1 described above. In FIG. 3, a two-dot chain line is a line indicating individual imaging ranges to be divided and 36 A is a principal point of the fisheye lens 12.

  The measurement work around the road by the measurement vehicle 20 described above is performed by repeatedly driving various measurement devices such as the camera 11 at predetermined distance intervals and the like as the measurement vehicle 20 travels on the road surface 1 and the obtained captured image 3 and the like. Is recorded in a storage device (not shown) mounted on the measurement vehicle 20. As a result, all directions including the zenith direction along the road length are continuously photographed from the road side. Similarly, almost all circumferences are continuously laser-ranged, and further, measurement is performed at the time of photographing and laser ranging. The position / posture information of the vehicle 20 is recorded. In order to increase the accuracy of the map, it is desirable to appropriately set a ground reference point on the side of the road before the measurement, and to photograph this with the road surface 1 and the like and perform laser distance measurement.

  The map creation process performed using the measurement result obtained by the above measurement operation will be described below. In order to facilitate understanding, the following description will be given while sequentially illustrating images and the like that are in the process of processing. However, in actuality, such an image is not generated each time, and processing is performed only by calculation on the measurement result. Done. Further, the following map creation process can be automatically performed by a computer that operates a program in which the following procedure is described. A block diagram of a computer configured as such a map creation apparatus B is shown in FIG. .

  The map creation apparatus B includes an input unit 26 to which the above-described measurement results are input, a calculation unit 27 that generates a map by calculation on the input measurement results, and an output that outputs the generated map to a display outside the figure. Part 28. The input unit 26 receives position / orientation information based on the captured image input unit 29 to which the above-described divided captured image (original image 12) of the camera 11 is input, the GNSS receiver 21, the IMU 22, and the odometer 25. An imaging position / orientation information input unit 30 and laser distance measurement data, that is, a laser data input unit 31 to which data capable of specifying a three-dimensional position / shape of a measurement target are input. The shooting position / orientation information input unit 30 also receives the mounting position and mounting posture of the camera 11 and the laser distance measuring device 23 in the measuring vehicle 20, and thereby matches the position / posture of the measuring vehicle 20 described above. The shooting position / orientation by the camera 11 or the laser distance measuring device 23 can be specified in detail.

  When the data input to the input unit 26 is completed as described above, the map data generation process by the arithmetic unit 27 described above is then started based on the input data. The computing unit 27 generates the omnidirectional captured image (captured image 3) from the divided captured image 12 obtained by dividing the omnidirectional image into 6 parts as described above, and the omnidirectional captured image 3. At the intersection 15 in the image obtained by the developed image generating unit 10, the individual image generating unit 8 that generates individual mosaic images that are constituent parts of the map, the developed image generating unit 10 that joins the individual mosaic images together. And an intersection adjustment unit 17 that applies adjustment processing to the corresponding portion.

  The wide-angle image generation unit 13 generates a single omnidirectional photographic image 3 by combining the divided photographic images 12, 12,... Taken at the same time. The divided captured images 12 are joined after block adjustment. More specifically, geometric correction and lens distortion correction are applied to each divided photographed image 12 by the correcting unit 32, and the divided photographed images 12, 12 are connected to each other by using, for example, epipolar lines by the tie point extracting unit 33. The corresponding position is detected, and the vehicle position / orientation analysis unit 34 obtains the shooting position and orientation information of each of the divided shot images 12, and the wide-angle coupling unit 35 uses the aerial triangulation or the forward intersection method to perform the division. The relative orientation, absolute orientation, and mosaic processing of the captured images 12, 12,.

  FIGS. 4A and 4B illustrate strict processing by geometric correction to the circumferentially divided captured image 12, and the principal point 36A of the lens of the divided captured image 12 and the virtual lens in the omnidirectional image. An error caused by the deviation α from the principal point 36B is also geometrically corrected. In the figure, 37 is a line that virtually represents the image plane of the divided photographed image 12, and 38 is a line that virtually represents the projection plane. Note that the image processing accompanying the change of the lens principal point can also be performed using the three-dimensional model 5 as in the individual image generation unit 8 described later.

  FIG. 3C shows a state in which each divided photographed image 12 after geometric correction is block-adjusted, in other words, shows a combined arrangement of the respective divided photographed images 12 in the omnidirectional photographed image 3, and this omnidirectional photographed image. 3, the horizontal direction of the paper surface corresponding to the peripheral direction of the measurement vehicle 20 is long. In the figure, reference numeral 38A denotes an area occupied by the divided photographed image 12 in the zenith direction, and reference numeral 38B denotes an area occupied by each divided photographed image 12 in the peripheral direction. FIG. 5A illustrates the omnidirectional image 3 obtained by the wide-angle image generation unit 13 described above. In the figure, 39 is a white line, 40 is a building, and 41 is a building. The road surface 1 extending in the back of the right side of the image is the front of the measuring vehicle 20, the central part of the image is the left side of the measuring vehicle 20, and the left side of the image. The road surface 1 extending in the back represents the rear of the measurement vehicle 20.

  When the omnidirectional captured image 3 is completed as described above, generation of individual mosaic images by the individual image generation unit 8 is started. As will be described later, in order to generate a map that approximates a development view, the individual mosaic images are obtained with respect to the mosaic image in the vertical shooting direction of the road (the mosaic image 6 in the vertical shooting direction) and the scenery 2 and 2 on both sides of the road. It consists of a set of mosaic images in the horizontal shooting direction (mosaic image 7 in the horizontal shooting direction), and the individual image generation unit 8 obtains a plurality of the above-described omnidirectional shooting images 3 obtained continuously in the traveling direction of the measurement vehicle 20. These images are generated by virtually changing the shooting direction by parallel projection conversion processing.

  As shown in FIG. 5B, the omnidirectional photographed images 3 are joined by block adjustment of the omnidirectional photographed images 3 by aerial triangulation based on the photographing position / orientation information. Specifically, the bundle method is used for block adjustment, and prior to this, a pass point 4 is set for each omnidirectional image 3. FIG. 6A shows the setting state of the pass point 4 by the pass point extracting unit 42. In this embodiment, the pass point 4 is extracted by the Harris method that is resistant to image scale change and rotation. It should be noted that the extraction of the pass point 4 can be performed by another method such as a shift method. Further, similarly to the wide-angle image generation unit 13 described above, correction processing such as geometric correction is performed by the correction unit 43.

  FIG. 6B shows an example of the block adjustment process, in which an image capturing the landscape 2 on the side of the measurement vehicle 20, that is, the roadside, is extracted from the omnidirectional image 3. By joining, a long mosaic image can be generated along the road length. The alternate long and two short dashes line shown in FIG. 3 indicates that the divided shooting image area 38A in the zenith direction (the shooting range of the high altitude shooting camera 11B) and the divided shooting image area 38B in the peripheral direction (the shooting range of the low altitude shooting camera 11A). By capturing the zenith area as described above, it is possible to reflect the relatively high altitude area of the high-rise building 18 on the roadside in the mosaic image.

  The block adjustment of the omnidirectional image 3 is performed by using the shooting position and azimuth information obtained by the vehicle position and orientation analysis unit 34 of the wide-angle image generation unit 13 described above. Block adjustment by simultaneous adjustment of the bundle method is performed by using the information and the pass point 4 of the photographed image 3 and the above-described ground reference point as necessary, and joint and mosaic processing are performed.

  The parallel projection conversion process by the individual image generation unit 8 is performed using a DSM (numerical surface layer model) which is a three-dimensional model 5. Therefore, prior to the parallel projection conversion process, the numerical surface layer model generation unit 45 of the individual image generation unit 8 generates a DSM 5 composed of TIN (irregular triangular network model) or the like using laser distance measurement data and shooting position / orientation information. Is done. The parallel projection conversion unit 46 that performs parallel projection conversion processing projects the above-described mosaic image onto the DSM 5 and virtually changes the imaging direction to the vertical and horizontal directions as described above, that is, captures the pixels of the projection image. Each of the mosaic images 6 and 7 in the vertical and horizontal shooting directions is generated by moving to a position corresponding to the change in direction. 7A shows an image of the mosaic image 7 in the horizontal shooting direction, and FIG. 7B shows an image of the mosaic image 6 in the vertical shooting direction, that is, an ortho image. The numerical surface layer model generation unit 45 specifies the three-dimensional position and shape by stereo matching of the omnidirectional images 3 and 3 with respect to the zenith direction periphery that is outside the distance measurement range by the laser distance measuring device 23, This is used to generate DSM5.

  When the mosaic images 6 and 7 in the vertical and horizontal shooting directions are completed in this way, the processing for joining the mosaic images 6 and 7 by the developed image generation unit 10 is started. The developed image generator 10 pastes the horizontal mosaic images 7 and 7 on both sides of the road on both sides of the road area in the vertical mosaic image 6 so as to be connected on the same plane. In other words, in reality, the horizontal mosaic image 7 that has a three-dimensional positional relationship that rises from the side edge of the road surface 1 region in the vertical mosaic image 6 is a photograph that approximates the development view by virtually falling down to the outside of the road. A map 9 is generated. At the time of joining, the mosaic images 6 and 7 are precisely aligned in the road length direction by using the position coordinates of the mosaic images 6 and 7, and the scenery 2 and 2 on both sides of the road are The displayed mosaic images 7 and 7 and the mosaic image 6 of the road area are matched in the positional relationship in the left-right direction.

  FIG. 8 shows a photographic image 9 approximated to a developed view obtained as described above. As shown in this figure, in this photographic image 9, the roadside scenery 2 and 2 projected in the horizontal direction are arranged on both sides of the orthographic projection road surface 1 while leaving the positional relationship in the plane direction. It is easy to grasp the geography sensuously, and the list is excellent because the whole is compact. However, regarding the intersection 15, if the branch line 47 is not uphill, it will only appear as a line in the horizontal mosaic image 7, and therefore it will be difficult to distinguish.

  The intersection adjusting unit 17 improves the difficulty of distinguishing the image, and pastes an interpolated image 16 in which the existence of the branch line 47 is clearly depicted in an image region that is missing like a blank by the branch line 47. The intersection extraction unit 48 of the intersection adjustment unit 17 is constituted by the surface of the building 40 located on both sides of the road surface 1 from the three-dimensional model 5 as shown in FIG. 7C showing the orthographic projection of the three-dimensional model 5 described above. The virtual surface 49 is extracted, and the presence of the intersection 15, in other words, the presence of the branch line 47 and its position are specified by the gap existing between the peripheral edges of the surface 49. When extracting the surface 49 from the surface of the building 40, a relatively narrow gap between the buildings 40 and 40 is erroneously extracted as the branch line 47 on condition that there is no gap having a width dimension that is estimated to be the branch line 47. Can be prevented.

  In addition, the interpolation image generation unit 50 of the intersection adjustment unit 17 generates the above-described interpolation image 16. In this embodiment, the measurement vehicle 20 side that can be extracted from the omnidirectional image 3 as described above. An interpolated image 16 is generated based on an image capturing the direction, that is, a side image. The interpolated image 16 is generated by using the position of the branch line 47 specified as described above, specifying the corresponding image area of the side image, and then cutting it out. This cutout is limited to a length that the captured image 3 is considered to be clear to some extent. FIG. 9A depicts this cut-out operation, and a two-dot chain line is a cut-out region.

  In addition, since the image cut out as described above generally has to be shortened in the dimension corresponding to the longitudinal direction of the branch line 47, the interpolated image generation unit 50 is configured as shown in FIG. 9B. In this embodiment, image transformation processing into a size and shape that is conspicuous can be performed by using affine transformation.

  Finally, the image pasting unit 51 pastes the obtained interpolation image 16 on the horizontal mosaic image 7 according to the position of the branch line 47 specified as described above. FIG. 10 shows a map completed by the above processing. A two-dot chain line represents the periphery of the interpolation image 16.

  Therefore, for example, when walking along this road or driving a vehicle, the surrounding scenery 2, 2 and this photographic map 9, more preferably horizontal mosaic images 7, 7 are displayed. If compared, it is possible to easily identify the own position, and it is also easy to associate with the movement on the map when the person is actually moving along the road. In particular, since the presence of the branch line 47 is easy to understand, it is easy to compare with the current situation, for example, whether or not the branch line 47 has been crossed while actually walking. Furthermore, by looking forward at the map along the vertical mosaic image 6, ie, along the road, facilities such as pedestrian crossings, parking frames for parking meters, etc. The presence of road markings drawn on the road surface 1 can be easily grasped in advance.

  In the above, the case where a simple photographic map 9 is created as a map has been shown. However, in order to improve usability, the building name, 15 intersections, road names, road speed limits, etc. are displayed on the photographic map 9. It is also possible to make a map displayed in an overlapping manner. In the above description, the case where the divided photographed image 12 is individually geometrically corrected and the plural images are joined to generate the omnidirectional photographed image 3 has been described. However, as shown in FIG. It is also sufficient to generate the omnidirectional image 3 by simultaneously calculating geometric correction including elimination of the shift β when the images 12 are joined.

DESCRIPTION OF SYMBOLS 1 Road surface 2 Scenery 3 Photographed image 4 Pass point 5 Three-dimensional model 6 Mosaic image in the vertical photographing direction 7 Mosaic image in the horizontal photographing direction 8 Individual image generating unit 9 Photo map 10 Expanded image generating unit 11 Camera 11A Low altitude region photographing camera 11B High-altitude photography camera 12 Original image 13 Wide-angle image generation unit 14 Fisheye lens 15 Intersection 16 Interpolated image 17 Intersection adjustment unit 18 High-rise building

Claims (2)

A map creation device that creates a map along a road using captured images obtained by continuously shooting the road surface and both sides of the road along the road length, and shooting position / azimuth information,
A photographed image is generated by joining the original images simultaneously photographed by a camera that shoots the road surface and both sides of the road from the same photographing position using the photographing position / orientation information, and is perpendicular to the road surface of the road. An individual image generation unit that generates a captured mosaic image and a mosaic image captured in the horizontal direction on both sides of the road with respect to the road length of the road ;
A developed image generation unit that creates a photographic map that approximates a developed view by connecting mosaic images in the horizontal shooting direction with the position at the lower end on the same plane at positions corresponding to both sides of the road of the mosaic image in the vertical shooting direction. When,
The intersection of the road is extracted from the photographed image or the original image, an interpolation image is generated from the branch photographing region of the intersection in the photographed image or the original image, and the interpolation image corresponds to the branch line of the mosaic image in the vertical photographing direction. A map creation device having an intersection adjustment unit that connects to a position . A map creation method for creating a map along a road using a photographed image obtained by continuously photographing the road surface and scenery on both sides of the road along the road length, and photographing position / direction information,
A photographed image is generated by joining the original images simultaneously photographed by a camera that shoots the road surface and both sides of the road from the same photographing position using the photographing position / orientation information, and is perpendicular to the road surface of the road. Generating a captured mosaic image and a mosaic image captured in the horizontal direction on both sides of the road with respect to the road length of the road;
Creating a photographic map approximating a developed view by connecting the mosaic images in the horizontal shooting direction on the same plane at the positions corresponding to both sides of the road of the mosaic image in the vertical shooting direction on the same plane;
The intersection of the road is extracted from the photographed image or the original image, an interpolation image is generated from a branch photographing region of the intersection in the photographed image or the original image, and the interpolation image corresponds to the branch line of the mosaic image in the vertical photographing direction. A mapping method comprising the steps of stitching to a location .

Images