JP5157067B2 - Automatic travel map creation device and automatic travel device. - Google Patents

Description

  The present invention relates to an apparatus for creating a map used for automatic traveling, and an automatic traveling apparatus.

Patent Documents 1 and 2 describe a method of creating road map data by causing a vehicle equipped with a measuring device to actually travel on a road. In this method, white lines and edges of a road are detected, and a measured value of the distance to the road is recorded to create road map data. In addition, scenes taken with a video camera are recorded in a superimposed manner.
JP 2000-338865 A JP 2003-329449 A JP 2002-98758 A

  However, when creating map data according to the prior art, the created map data includes a GPS error, a sensor error, and the like. In addition, in the creation of map data according to the prior art, when an obstacle such as a parked vehicle temporarily exists along the road, the measurement vehicle detours to avoid the obstacle. The map data collected in step 1 contains a large error. As described above, there is a problem that a large error is included in the map data, and accurate map data may not be created.

  Accordingly, an object of the present invention is to provide a map creation device that can create map data with less error.

  Another object of the present invention is to provide an automatic traveling apparatus capable of performing automatic traveling with high accuracy.

  In order to achieve the above-described object, the automatic travel map creation device according to the present invention divides an area along a road into a number of small areas, and associates a weight indicating the presence of a measurement target with each small area. The storage means for storing the map and the position data of the measurement target obtained each time the vehicle equipped with the measurement device travels on the same road, the measurement target is stored in a small area corresponding to the position data. And a same road information combining means for adding a weight indicating presence.

  According to this configuration, position data to be measured is obtained every time a vehicle equipped with a measuring device travels on a road. A weight indicating the presence of the measurement target is added to the small area corresponding to the position data of the measurement target in the map. Therefore, the map reflects the weight of a vehicle equipped with a measurement device that travels multiple times, so using this map will cause errors due to the measurement device and errors caused by the vehicle avoiding obstacles. It is possible to create map data that suppresses the above.

  In the automatic travel map creation device described above, it is preferable that the same road information combining unit specifies the position of the measurement target based on the weights associated with a large number of small areas of the map. According to this configuration, since the position of the measurement target is specified based on the weights corresponding to many small areas of the map, the position error of the measurement target can be reduced.

  Further, in the above-described automatic travel map creation device, it is preferable that the weight added to the map has a distribution that becomes maximum at the position of the measurement target and decreases as the distance from the measurement symmetry position increases. According to this configuration, the weight added to the map has a distribution that is maximum at the position of the measurement target and decreases with distance from the measurement symmetrical position, and therefore, an appropriate weight according to the existence probability of the measurement target. You can create map data.

  In the automatic travel map creation device described above, the distribution of the weight added to the map is preferably a distribution according to the position error of the measurement target. According to this configuration, the distribution of the weight added to the map is a distribution according to the position error of the measurement target, and therefore map data can be created with an appropriate weight according to the existence probability of the measurement target.

  In the above-described automatic travel map creation device, the measurement target is preferably a travel locus of the vehicle. According to this configuration, it is possible to create map data with a small error with respect to the traveling locus of the vehicle.

  In the automatic travel map creation device described above, the measurement target is preferably a road component. According to this structure, map data with few errors can be created for road components. Here, the road component is an object that generally exists along the road, such as a white line, a curb, a gutter, and a guardrail.

  In order to achieve the above-described object, an automatic traveling apparatus according to the present invention includes a position detection unit that measures the position of a vehicle, an imaging unit that captures the periphery of the vehicle and generates image data, and a position of a road component. Based on the error, a search range of the road component in the image data is set, a position calculation unit that calculates the position of the road component by processing the search range, and the road component calculated by the position calculation unit A vehicle position correction unit that corrects the vehicle position measured by the position detection unit based on the position; and a control command output unit that outputs a control command for automatic travel based on the corrected vehicle position. It is characterized by.

  According to this configuration, since the search range of the road component in the image data is set based on the position error of the road component, the search range is small and surely includes the road component. Therefore, by calculating the position of the road component by processing such a search range, the position of the road component can be quickly calculated, and erroneous detection of the road component can be prevented. it can. Since the vehicle position measured using the position detecting means is corrected based on the calculated position of the road component, automatic traveling can be performed with high accuracy.

  In the automatic travel device described above, the position error of the road component is a position error due to stereo vision, a position error due to a GPS radio wave reception state, a position error due to an azimuth angle error, a position error due to a pitch angle error, It is preferable to include at least one of a position error due to communication delay and a position error at the time of creating the map database.

  In the automatic travel device described above, the vehicle position correcting means corrects the vehicle position measured by the position detecting means when the error of the vehicle position measured by the position detecting means is larger than a predetermined threshold, and the position detecting means It is preferable to prohibit the correction of the vehicle position measured by the position detection means when the measured vehicle position error is smaller than a predetermined threshold. According to this configuration, the vehicle position is corrected only when necessary for control by correcting the vehicle position measured by the position detection means only when the error of the vehicle position measured by the position detection means is larger than the predetermined threshold. Can be corrected.

  Further, in the above-described automatic traveling apparatus, the imaging unit generates stereo upper image data and lower image data, and the position calculating unit calculates the road component or obstacle from each of the upper image data and the lower image data. It is preferable to extract a lateral edge corresponding to an object and calculate a position of a road component or an obstacle based on the extracted lateral edge. According to this configuration, the position of the road component or the obstacle can be suitably calculated based on the lateral edge extracted from the upper image data and the lower image data.

  In the above-described automatic traveling device, it is preferable that the position calculating means determines the type of road component or obstacle based on the position of the lateral edge. According to this configuration, it is possible to determine the type of road component or obstacle based on the position of the horizontal edge by utilizing the fact that the position of the horizontal edge varies depending on the type of road component or obstacle. .

  In the above-described automatic travel device, it is preferable that the position calculating means calculates the position of the area between the lateral edges and determines the type of the road component based on the position of the area between the lateral edges. According to this configuration, it is possible to more accurately determine the type of the road component by using the position of the region between the lateral edges.

  In the above-described automatic traveling device, the position calculating means determines that the road component is a gutter when the position of the region between the lateral edges is lower than the plane on which the vehicle is placed by a predetermined threshold or more. It is preferable to do. According to this configuration, it can be appropriately determined that the type of road component is a gutter.

  Further, in the above-described automatic traveling device, the position calculating means is configured such that when the position of the region between the lateral edges is lower than the plane on which the vehicle is placed and is aligned vertically with respect to the plane, It is preferable to determine that the component is a side groove. According to this configuration, it can be appropriately determined that the type of road component is a gutter.

  Further, in the above-described automatic traveling apparatus, the imaging means captures a front area, a left area, a right area, and a rear area of the vehicle, and generates image data of each area. Using the two image data in which the captured areas of the image data partially overlap as stereo-viewed image data, the position of the road component in the overlap area is calculated, and the calculated road component It is preferable to estimate the position of the road structure in the non-overlapping area based on the position of According to this configuration, the position of the road component in the region that is not viewed in stereo can be estimated appropriately.

  According to the present invention, map data with few errors can be created. Further, according to the present invention, automatic traveling can be performed with high accuracy.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

[Automatic map creation device]
First, a preferred embodiment of an automatic travel map creating apparatus for creating a map used by a vehicle during automatic travel will be described. As shown in FIG. 1, the automatic travel map creation device 1 includes an imaging device (imaging means) 10, a data processing device 12, and a detection unit 14 that detects the position / orientation / attitude of the vehicle. . The data processing device 12 is physically configured to include a CPU, RAM, ROM, and the like. Functionally, the data processing device 12 includes an image data capturing unit 16, a road component recognition unit 18, and a position calculation unit 20. A maximum error calculation unit 22, a measurement information writing unit 24, a measurement information storage unit 26, an identical road information extraction unit 28, an identical road information combination unit 30, a map data writing unit 32, and a map database 34. It is configured to include.

  The image capturing apparatus 10 captures a road around the vehicle and generates image data. In the present embodiment, the image capturing apparatus 10 includes two CCD cameras. The two CCD cameras are pointed in the same direction and are a little apart from each other. When two CCD cameras capture images simultaneously, two image data in which the periphery of the vehicle is viewed in stereo is generated. Here, the image data generated by each CCD camera is data having a plurality of pixel data, and is data having luminance data and color data as each pixel data.

  Note that the imaging apparatus 10 can also be configured to generate two pieces of image data in which the periphery of the vehicle is viewed in stereo using one CCD camera. This configuration will be described in detail later. Further, the imaging apparatus 10 may generate two pieces of image data in which the periphery of the vehicle is viewed in stereo by a motion stereo method using a single CCD camera. Further, the imaging device 10 may be configured by two millimeter wave radars, laser radars, and the like. When a radar is used for the imaging device, the image data generated by the radar has distance data from the own vehicle to the detected object as each pixel data.

  The vehicle position / orientation / attitude detection unit 14 includes a plurality of types of sensors attached to the vehicle. (1) A GPS (Global Positioning System) device is used as means for detecting the position of the vehicle. The GPS device detects the position of the vehicle in a world plane rectangular coordinate system, which will be described later, by processing signals from the satellite. (2) A lateral acceleration sensor and a steering angle sensor are used as means for detecting the azimuth (traveling direction) of the vehicle. The lateral acceleration sensor detects acceleration acting in the lateral direction of the vehicle. The steering angle sensor detects the angle of the steering wheel steered by the driver. (3) A roll sensor, a pitch sensor, a yaw sensor, and a vehicle height sensor are used as means for detecting the attitude of the vehicle. The roll sensor detects the roll angle of the vehicle, the pitch sensor detects the pitch angle of the vehicle, and the yaw sensor detects the yaw angle of the vehicle. The vehicle height sensor detects the height of the vehicle body with respect to the road surface.

  In the data processing device 12, the image data capturing unit 16 performs processing for capturing the two image data generated by the imaging device 10 into the data processing device 12. The road component recognition unit 18 performs processing for recognizing road components such as white lines, curbs, gutters, and guardrails for each of the two image data. Specifically, the road component recognition unit 18 performs (1) edge extraction processing using a Sobel filter, Canny filter, (2) line segment using Hough transform, least square method, or the like on image data. By performing detection processing, (3) difference summation, normalized correlation, pattern matching processing using a neural network, etc., a road component in the image is recognized.

The position calculation unit 20 performs a process of calculating a position where the road structure exists in the three-dimensional coordinate system based on the pixel position of the road structure in the two image data. An outline of a method for calculating the position of the road component in the three-dimensional coordinate system will be described with reference to FIG. FIG. 2 shows a situation where a road component is photographed by the left and right CCD cameras. The base line length, which is the distance between the optical centers of the left and right CCD cameras, is b, and both the left and right CCD cameras generate image data with a focal length f. In the left image data _{I L,} 2 axes _X L on the image, based on the _{Y L,} the pixel position Pl (xl, yl) of the road construction material is specified. On the other hand, the right side of the image data _{I R,} 2 axes _X R of the image, the _{Y R} as a reference, the pixel position Pr (xr, yr) of the road construction material is specified. The position calculation unit 20 calculates a position P (X _ps , Y _ps , Z _ps ) where the road component exists in the imaging coordinate system based on these geometrical relationships. The imaging coordinate system is a coordinate system with the imaging device 10 as a reference, and is a coordinate system in which coordinates are specified by three axes Xs, Ys, and Zs with the center of the left and right optical centers as the origin.

Specifically, the position calculation unit 20 calculates a position P (X _ps , Y _ps , Z _ps ) where the road component exists according to the following mathematical formula (1).

Next, the position calculation unit 20 uses the position P (X _ps , Y _ps , Z _ps ) of the road component in the imaging coordinate system and the position P (X _pc , Y _pc , Z _pc ) of the road component in the vehicle coordinate system. ), And further, a process of converting to a position P (X _pf , Y _pf , Z _pf ) of the road component in the vehicle modified coordinate system is performed. As shown in FIG. 3, the vehicle coordinate system is a coordinate system based on the vehicle, with the center position of the vehicle as the origin, and the three axes Xc parallel to the three axes Xs, Ys, and Zs of the imaging coordinate system. , Yc, Zc is a coordinate system for specifying coordinates. The origin of the vehicle coordinate system is separated from the origin of the imaging coordinate system by ax, ay, and az along the axes Xs, Ys, and Zs. The vehicle reformed coordinate system is a coordinate system based on the vehicle, and has the center position of the vehicle as the origin, the Xf axis extending in the horizontal plane and extending laterally of the vehicle, and extending in the horizontal plane and extending forward of the vehicle. This is a coordinate system in which coordinates are specified by the Yf axis and the Zf axis extending vertically. In addition, when each of the axes Xc and Yc of the vehicle coordinate system is projected onto the horizontal plane, it coincides with the axes Xf and Yf of the vehicle modified coordinate system. The axis Xc of the vehicle coordinate system is inclined by an angle φ in the roll direction with respect to the axis Xf of the vehicle modified coordinate system, and the axis Yc of the vehicle coordinate system is in the pitch direction with respect to the axis Yf of the vehicle modified coordinate system. It is inclined by an angle ρ.

Specifically, the position calculation unit 20 determines the position P (X _ps , Y _ps , Z _ps ) of the road component in the imaging coordinate system and the position of the road component in the vehicle coordinate system according to the following equation (2). Convert to P (X _pc , Y _pc , Z _pc ). Further, the position calculation unit 20 calculates the position P (X _pc , Y _pc , Z _pc ) of the road component in the vehicle coordinate system and the position P ( X _pf , Y _pf , Z _pf )

Next, the position calculation unit 20 changes the position P (X _pf , Y _pf , Z _pf ) of the road component in the vehicle modified coordinate system to the position P (φom, λom) of the road component in the world plane rectangular coordinate system. Along with the conversion, a process of converting to a position P (X _cg , Y _cg , Z _cg ) of the road component in the world plane coordinate system is performed. Here, as shown in FIG. 4, the world plane rectangular coordinate system is a coordinate system in which two axes are set for latitude φ and longitude λ with the center of the earth as a reference. The position P (φom, λom) of the road component in the world plane rectangular coordinate system is equal to the position P (X _pf , Y _pf , Z _pf ) of the road component in the vehicle modified coordinate system and the vehicle C obtained by the GPS device. Calculated from the position data. Further, as shown in FIG. 5, the world plane coordinate system is a coordinate system in which a reference point Ow is set at a position on the ground surface, and three axes are set for north-south Xg, east-west Yg, and altitude. The position P (X _cg , Y _cg , Z _cg ) of the road component in the world plane rectangular coordinate system is determined by the position P (X _pf , Y _pf , Z _pf ) of the road component in the vehicle modified coordinate system and the GPS device. It is calculated from the position data of the vehicle C obtained.

  The maximum error calculation unit 22 calculates the maximum error estimated to be included in the measurement position of the vehicle from the reception state of the GPS radio wave. Here, as for the error Eg caused by the reception state of GPS radio waves, there is a commercially available GPS device that outputs measurement accuracy data, so that data may be used. Further, the maximum error calculation unit 22 calculates the maximum error estimated to be included in the position data of the road component calculated by the position calculation unit from the camera distance resolution, the GPS radio wave reception state, and the like. FIG. 6 shows an example of the maximum error calculated by the maximum error calculation unit 22. The curve Err indicating the maximum error is calculated by adding the error Ec (D) caused by the camera distance resolution and the error Eg caused by the reception state of the GPS radio wave according to the following equation (4).

  The error Ec (D) resulting from the camera distance resolution can be calculated based on the angle of view, the number of pixels, the baseline length, and the distance D from the camera to the road component, and the distance D from the camera to the road component is The smaller it is, the smaller it is. It can be seen that the actual measurement value of the position error indicated by the black circle in FIG. 6 is smaller than the maximum error calculated by the maximum error calculation unit 22. In this embodiment, only the error Ec (D) caused by the camera distance resolution and the error Eg caused by the reception state of the GPS radio wave are considered, but the lateral acceleration detection sensor, the steering angle detection sensor, the roll sensor, and the pitch sensor are considered. If an error such as a yaw sensor or a vehicle height sensor is taken into account, a more accurate estimated maximum error can be calculated.

  The measurement information writing unit 24 writes information regarding the vehicle and road components obtained by the above-described calculation into the measurement information storage unit 26. Here, information related to vehicles is shown in Table 1 below, and information related to road components is shown in Table 2 below.

  In Table 2 above, the type ID of the road component is identification information for distinguishing road components of different types. For example, “1” is associated with the white line, “2” is associated with the side groove, and “3” is associated with the guardrail. Further, the label ID is identification information for distinguishing road components for each individual when there are a plurality of road components of the same type. For example, when there are two white lines, “1-1” is associated with the first white line, and “1-2” is associated with the second white line.

  In Table 2 above, the information on the road component is a point sequence in which a large number of points are arranged in a row, as conceptually shown in FIG. However, as conceptually shown in FIG. 7B, the measurement information writing unit 24 may write the road component line information in the measurement information storage unit 26 in place of the point sequence information of the road component. Good. Table 3 below shows the line information of road components.

  The same road information extraction unit 28 searches and extracts vehicle information and road component information created for the same road from the vehicle information and road component information stored in the measurement information storage unit 26. Here, the extracted information is information obtained by the same vehicle traveling on the same road a plurality of times, or information obtained by different vehicles traveling on the same road.

  The same road information combining unit 30 calculates the target travel locus of the vehicle and calculates the position of the road component based on the vehicle information and the road component information extracted by the same road information extraction unit 28. . In the following, the process of calculating the target travel locus by the same road information combining unit 30 will be described, and then the process of calculating the position of the road component will be described.

  The same road information combination unit 30 performs a process of generating a travel locus map based on each of the vehicle information extracted by the same road information extraction unit 28. Here, the traveling locus map is a map in which a large number of small square regions are two-dimensionally arranged by finely dividing a region along the road in the north-south direction and the east-west direction at predetermined intervals. In the travel locus map, a weight indicating the existence probability of the travel locus is associated with each small region.

  In order to generate a travel trajectory map, the same road information combining unit 30 first interpolates between a series of measurement points of the vehicle position by a straight line, a Pezier curve, or the like as shown in FIG. Find the running track. Next, the same road information combining unit 30 performs weighting on the travel locus. Here, the weighting on the travel locus is performed using a normal distribution as shown in FIG. In FIG. 9, the travel locus extends perpendicular to the paper surface, the horizontal position is represented by the axis u, and the weight is represented by the axis G.

The weighting formula using the normal distribution is expressed by the following formula (5).

Here, σ is a variable for determining the width of the normal distribution, and is determined in consideration of the maximum error Err estimated for the vehicle position. For example, σ = Err / 3 is set.

  By weighting the travel locus shown in FIG. 8 using the above equation (5), the weighted travel locus shown in FIG. 10 is calculated. As shown in FIG. 11, the weighting on the travel locus may be set such that the weight is maximized on the travel locus and linearly decreases in proportion to the distance from the travel locus.

  FIGS. 12A to 12D show image data of a travel locus map obtained by a vehicle equipped with a map creation device traveling four times on the same route. Here, in the travel locus map, the higher the luminance, the larger the weight, and the lower the luminance, the smaller the weight. In the above-described processing, a map that weights the travel trajectory is generated as the travel trajectory map, but a map that does not weight the travel trajectory may be generated. That is, the travel locus map may be generated by giving a constant value (for example, 1.0) to a small region through which the travel locus passes.

  Next, the same road information combining unit 30 adds the travel locus map to the travel locus weight addition map and rewrites the travel locus weight addition map every time the travel locus map is generated as described above. Here, the travel trajectory weight addition map is a map composed of a large number of small areas arranged two-dimensionally, like the travel trajectory map, and an added value of weight is associated with each small area.

  A specific example of the rewriting process of the travel locus weight addition map will be described. In the travel locus maps of FIGS. 12A to 12D, the travel locus maps of FIGS. 12A, 12C, and 12D have substantially the same shape, but FIG. Only the travel locus map of) has a different shape. Regarding the travel locus map of FIG. 12B, it is presumed that the vehicle equipped with the map creation device avoids other vehicles and obstacles while traveling on the road. FIG. 13 shows the result obtained by adding these travel locus maps to the travel locus weight addition map. In the travel locus map of FIG. 13, the weight addition value of the small region that has traveled many times is large and the luminance is large. On the other hand, the weight addition value of the small area where the number of times of traveling is small is small and the luminance is small.

  Next, the same road information combining unit 30 binarizes the area where the weight addition value is larger than the preset weight threshold Th1 and the area where the weight addition value is smaller than the weight threshold Th1 in the travel locus weight addition map. Turn into. For example, when the travel locus weight addition map shown in FIG. 14A is binarized, the map is converted as shown in FIG. In FIG. 14B, the weight addition value in the white area is larger than the preset weight threshold Th1, and the weight addition value in the black area is smaller than the preset weight threshold Th1.

  Next, as shown in FIG. 14C, the same road information combining unit 30 obtains the center position of the white area at the end of the area R as the first position of the target travel locus in the area R. Moreover, the same road information coupling | bond part 30 calculates | requires a 1st advancing direction by calculating the average value of the vehicle azimuth | direction angle (refer Table 1) in the 1st position vicinity. Then, as shown in FIG. 14 (d), the same road information combining unit 30 obtains a position that has traveled a certain distance in the vehicle traveling direction from the first position of the target travel locus, passes through the position, and The center position of the white area on the line segment orthogonal to the traveling direction of 1 is obtained as the second position of the target travel locus. Moreover, the same road information coupling | bond part 30 calculates | requires a 2nd advancing direction by calculating the average value of the vehicle azimuth | direction angle in the 2nd position vicinity. The same road information combining unit 30 performs the same process thereafter, and as shown in FIG. 14 (e), a series of positions (third position, fourth position, fifth position, ...) And a series of traveling directions (third traveling direction, fourth traveling direction, fifth traveling direction,...) Are calculated.

  Next, a process for calculating the position of the road component by the same road information combining unit 30 will be described. The same road information combining unit 30 performs processing for generating a road component map based on the road component information extracted by the same road information extracting unit 28. Here, the road component map is a map composed of a large number of small regions arranged two-dimensionally, like the travel locus map. In the road component map, a weight indicating the existence probability of the road component is associated with each small region.

  In order to generate a road component map, the same road information combining unit 30 first interpolates between a series of measurement points of the road component with a straight line, a Pezier curve, etc., and then a series of measurement points of the road component. Weighing straight lines or curves connecting Here, as with the above-described weighting to the travel locus, the line connecting the measurement points of the road components may be weighted.

  FIGS. 15A to 15E show maps of road components (side grooves) measured at each position when the vehicle is traveling on the road. When the vehicle is traveling on the road, the range of road components photographed by the imaging device 10 is limited. Therefore, in each of FIGS. 15A to 15E, the road components (side grooves) Only a part of is detected. By adding the partial maps of these road constituents to each other, the road constituent map is completed as shown in FIG.

  FIGS. 17A to 17D show image data of a road composition map obtained by a vehicle equipped with a map creation device traveling four times on the same route. Here, in the road composition map, the higher the luminance, the larger the weight, and the lower the luminance, the smaller the weight. In the above-described processing, a road weight map is generated as the road structure map, but a map that does not weight the road structure may be generated. That is, a road composition map may be generated by giving a constant value (for example, 1.0) to a small area through which the road composition passes.

  Next, the same road information combining unit 30 adds the road component map to the road component weight addition map and rewrites the road component weight addition map every time the road component map is generated as described above. I do. Here, the road component weight addition map is a map composed of a large number of small regions arranged two-dimensionally, like the road component map, and each small region is associated with a weight addition value. .

  A specific example of the road component weight addition map rewriting process will be described. The result of adding these road component maps to the road component weight addition map is shown in FIG. In the road component weight addition map of FIG. 18, the weight addition value of the small area traveled many times by the vehicle is large and the luminance is large. On the other hand, the weight addition value of the small area where the number of times of traveling is small is small and the luminance is small.

  Next, the same road information combining unit 30 divides the road component weight addition map into two areas where the weight addition value is larger than the preset weight threshold Th2 and the area where the weight addition value is smaller than the weight threshold Th2. Convert to value. Next, the same road information combination unit 30 uses the same method as that for obtaining a series of positions of the target travel locus to determine the existence positions of the series (first, second, third,...) Of road components. The extension direction of a series (first, second, third,...) Of road components is calculated by the same method as that for calculating and obtaining a series of traveling directions of the target travel locus. FIG. 19 shows the result of obtaining the location and extension direction of the road component. As shown in FIG. 20, the same road information combining unit 30 creates the above-described road component weight addition map for each type of white line, gutter, curb, guardrail, and the like.

  The map data writing unit 32 converts the information on the target travel locus and road components into the automatic travel map format, and then stores it in the map database 34. The format of the map for automatic driving is shown in Table 4 below. Note that the information shown in the following Table 4 corresponds to one point of the target travel locus.

  Table 4 above shows a data group stored corresponding to one point of the target travel locus, and such a data group is stored in the map database 34 for each point of the target travel locus. ing. Each point of the target travel locus has a 50 cm period, and the data group in Table 4 above is stored corresponding to each point of the 50 cm period.

  As shown in FIG. 21, when there are road components such as white lines, gutters, curbs, guardrails, etc., a line segment that has the shortest distance from the target travel locus to the road component is assumed. Here, the length of the line segment is the distance L, the angle between the target travel locus and the line segment is the existence direction β1, and the angle between the extending direction of the road component and the line segment is the posture angle β2. . The estimated maximum position error is an error caused by the camera distance resolution, the GPS radio wave reception state at the time of image acquisition, and the like. In the road component weight addition map shown in FIG. When the area is white and the area having a weight smaller than the predetermined threshold is black, the width W of the band-shaped white area is calculated by halving.

  Next, with reference to the flowchart of FIG. 23, the process flow of the above-described automatic travel map creation device will be described. First, the data processing device 12 takes in the image data generated by the imaging device 10 (S101). Next, the data processing device 12 processes the image data and performs a road component recognition process (S102). Next, the data processing device 12 takes in the measurement data of the position, orientation, and attitude of the vehicle (S103). Next, the data processing device 12 calculates the position of the road component (S104). Next, the data processing device 12 calculates the maximum error estimated for the position of the vehicle and the road component from the camera distance resolution, the GPS radio wave reception state, and the like (S105).

  Next, the data processing device 12 writes information on the vehicle and road components in the measurement point information storage unit (S106). Next, the data processing device creates a travel locus map and further creates a travel locus weight addition map based on the information about the vehicle. Similarly, the data processing apparatus creates a road composition map based on the information on the road composition, and further creates a road composition weight addition map (S107). Then, the data processing device 12 determines whether or not to update the information on the target travel locus and the road component position stored in the map database (S108). If the map database is updated, the process proceeds to S109. On the other hand, if the map database is not updated, the process returns to S101.

  The data processing device 12 binarizes the travel locus weight addition map with the predetermined threshold Th1, and calculates the target travel locus (S109). In addition, the data processing device 12 binarizes the road component weight addition map with the predetermined threshold Th2, and calculates the location of the road component (S110). Then, the data processing device 12 converts the information on the target travel locus and the location of the road component into the automatic travel map format, and then stores it in the map database (S111, S112). Finally, the data processing device 12 determines whether or not to end the process. If the process is not terminated, the process returns to S101.

[Automatic traveling device]
Next, a vehicle equipped with the automatic travel device 3 that uses the automatic travel map described above will be described. As shown in FIG. 24, the automatic travel device 3 includes an imaging device (imaging means) 40, a data processing device 42, and a detection unit 44 that detects the position / orientation / attitude of the vehicle. The data processing device 42 is physically configured to include a CPU, a RAM, a ROM, and the like, and functionally includes a map database 46, an image data fetching unit 48, and a road component position calculating unit (first item). 1 position calculation unit) 50, search distance setting unit 52, in-image position estimation unit 56, error influence calculation unit 58, minimum search frame setting unit 60, recognition object type setting unit 62, and object A recognition unit (second position calculation unit) 64, a host vehicle position correction unit 66, and a control command output unit 68 are included.

  In the automatic travel device 3, the imaging device 40 and the vehicle position / orientation / attitude detection unit 44 are the same as the imaging device 10 and the vehicle position / orientation / attitude detection unit 14 of the map creation device 1, and thus the description thereof is omitted. Similarly, in the automatic traveling device 3, the image data capturing unit 48 is the same as the image data capturing unit 16 of the map creation device 1, and thus description thereof is omitted. The map database 46 stores map data for automatic travel created by the map creation device 1.

  The road component position calculation unit (first position calculation unit) 50 takes in the vehicle position data obtained by the GPS device. In addition, the road component position calculation unit 50 extracts and imports data (see Table 4) relating to the point closest to the vehicle position data from the target travel locus data stored in the map database 46. Then, the road component position calculation unit 50 obtains a relative position from the vehicle to the point of the target travel locus using the captured data, and further, from this point of the target travel locus to the road component. To calculate the position of the road component in the vehicle coordinate system. The relative position of the road component calculated here includes a measurement error by the GPS device.

  Further, the road component position calculation unit 50 takes in the search distance data from the search distance setting unit 52, compares the relative position of the road component with the search distance, and compares the relative values of the road components within the search distance. Only the position is retained, and the relative position of the road component outside the search distance is discarded. The search distance setting unit 52 performs a process of determining a search distance for road components. Here, the search distance may always be set to a fixed distance, or may be set to a distance that changes according to the vehicle condition.

The in-image position estimation unit 56 performs processing for estimating the position of the road component in the image. In order to perform this processing, the in-image position estimation unit 56 extracts and captures data relating to the point closest to the vehicle position data, as with the road component position calculation unit 50. Then, the in-image position estimation unit 56 calculates the position data (X _pg , Y _pg , Z _pg ) of the road component in the world plane rectangular coordinate system from the data regarding the points of the target travel locus. The position data of the road component is calculated according to the following formula (6).

Next, the in-image position estimation unit 56 calculates the position data (X _pg , Y _pg , Z _pg ) in the world plane rectangular coordinate system and the position data (X _ps , Y in the imaging coordinate system) according to the following equation (7). _ps , Z _ps ).

Next, the in-image position estimating unit 56 calculates the position data (X _ps , Y _ps , Z _ps ) in the imaging coordinate system according to the following mathematical formula (8), as the pixel position ( X _L , Y _L ). In addition, the in-image position estimation unit 56 uses the position data (X _ps , Y _ps , Z _ps ) in the image capturing apparatus 40 coordinate system according to the following mathematical formula (9) to _obtain the pixel position in the image captured by the right camera. Convert to (X _R , Y _R ).

Here, P ₁ , P ₂ , H ₁ , and H ₂ are camera parameters derived by calibration of the stereo camera.

  The error influence calculation unit 58 performs a process of calculating the sum of various position errors such as a position error due to stereo vision and a position error of the GPS device for the extracted road component. Specifically, the error influence calculation unit 58 performs the calculation of the following formula (10). The types of position errors are shown in Table 5 below.

The position error due to stereo vision is calculated by the following formula (11). The position error due to the stereo view is a function of the distance Ds from the own vehicle to the road component.

  The position errors Px, Py, and Pz resulting from the GPS radio wave reception state may be calculated based on the measurement accuracy data because some commercially available GPS devices output measurement accuracy data.

The position error Hd caused by the azimuth error (the component in the horizontal plane of the vehicle direction error) Ae is calculated according to the following equation (12). Since the azimuth error Ae affects only the X-axis direction, the position error Hd occurs only in the X-axis direction. Since the maximum value Ae of the azimuth error is known (for example, 0.05 °), the position error Hd is a function of the distance Ds to the road component.

The position error Pt caused by the pitch angle error (vertical component of the vehicle direction error) Pe is calculated according to the following equation (13). Since the pitch angle error Pt affects only the Z-axis direction, the position error Pt occurs only in the Z-axis direction. Since the maximum value Pt of the pitch angle error is known (for example, 0.05 °), the position error Pt is a function of the distance Ds to the road component.

  Strictly speaking, the position error Cm caused by the delay in data communication in the data processing device 42 is a value that is a function of the vehicle speed, the angular velocity, etc., but is handled as a constant value (for example, 100 mm) in this processing.

  As the position error at the time of creating the map database 46, estimated maximum position errors Err_W, Err_L and Err_h (see Table 4) stored in the map database 46 are used. Here, when the line of sight of the camera is directed in the direction in which the road component exists, Err_W may be associated with the X axis, Err_L with the Y axis, and Err_h with the Z axis. However, if the camera's line of sight does not point in the direction in which the road component exists, the position error can be calculated based on the estimated maximum position errors Err_W, Err_L and Err_h. That's fine.

The minimum search frame setting unit 60 performs a process of setting a frame for searching for each road component in images taken by the left and right cameras. As shown in FIG. 25, the pixel position (X _L , Y) of the road component (curbstone) calculated by the in-image position estimation unit 56 in consideration of the position error integrated values ΣerrX, ΣerrY, ΣerrZ in the respective axial directions. _L ), (X _R , Y _R ) are given a suitable margin. Specifically, assuming a road component as a center, a cubic space of ΣerrX × 2 in the X-axis direction, ΣerrY × 2 in the Y-axis direction, and ΣerrZ × 2 in the Z-axis direction is assumed, and this cubic space is projected onto the image. Set the search range as a search frame. Note that the minimum search frame setting unit 60 sets a search frame for each road component when there are a plurality of road components.

  The recognition target object type setting unit 62 performs a process of setting the type of the road object to be recognized. Here, the recognition target object type setting unit 62 is determined as being within the search distance by the above-described road component position calculation unit 50 from the plurality of search frames set by the minimum search frame setting unit 60. Only the search frame of the road component is set.

The object recognition unit (second position calculation unit) 64 detects the road object to be recognized within the search frame set in the left and right images using edge extraction or pattern recognition technology, and The pixel position Pl (xl, yl) of the road component in the image data and the pixel position Pr (xr, yr) of the road component in the right image data are obtained. The object recognition unit 64 calculates a position P (X _ps , Y _ps , Z _ps ) where the road component exists in the imaging coordinate system based on these pixel positions Pl and Pr, and finally these positions. P (X _ps , Y _ps , Z _ps ) is converted into an existing position P (X _pf , Y _pf , Z _pf ) in the vehicle modified coordinate system. The relative position of the road component calculated here does not include a measurement error by the GPS device.

  The own vehicle position correction unit 66 performs GPS based on the position data of the road structure calculated by the road structure position calculation unit 50 described above and the position data of the road structure recognized by the object recognition unit 64. A process of correcting the position data of the own vehicle detected by the device is performed. This process will be described with reference to FIG.

  The own vehicle position correction unit 66 has performed the process up to this point so that the position data Pg of the own vehicle, the position data Pt of the points of the target travel locus, the position data Pm of the road components stored in the map database together with the target travel locus, and the The data β2 in the current direction is grasped. The own vehicle position correction unit 66 assumes a straight line L1 from the point Pt of the target travel locus to the road component Pm, and an intersection position Pn between the straight line L1 and the road component Q detected by the object recognition unit 64. Is calculated. Then, the vehicle position correction unit 66 calculates a difference ΔL between the position Pm of the road component by GPS and the position Pn of the road structure by image recognition. In addition, the host vehicle position correction unit 66 calculates an angle β3 formed by the straight line L1 and the road component Q.

  The own vehicle position correcting unit 66 calculates the absolute value of the difference between the extending direction β2 of the road component Pm stored in the position database and the extending direction β3 of the road component calculated by the object recognizing unit 64 | β2−. β3 | is calculated, and it is determined whether or not the absolute value is larger than a preset threshold value. Here, when the absolute value is smaller than the threshold value, the detection status is normal, and subsequently, correction processing of the own vehicle position is performed. On the other hand, if the absolute value is larger than the threshold value, the detection status is abnormal, and the process is terminated without performing the correction process of the vehicle position.

  When the absolute value is smaller than the threshold value, the own vehicle position correcting unit 66, when the perpendicular is lowered to the straight line L1 extending in the existing direction from the measured own vehicle position Pg, as shown in FIG. The position where the perpendicular intersects is calculated, and further, the position of the point Pr moved by the difference ΔL from the intersection position on the straight line L1 extending in the existence direction is calculated. The position of this point Pr is the corrected vehicle position. Thereby, the measurement error by the GPS device is corrected.

  When the measurement error of the GPS device is within the range of the circle E indicated by the broken line, if the vehicle position is corrected to the point Pr as described above, the position data of the vehicle is excessively corrected. It becomes. In order to avoid such a situation, when the point Pr is outside the range of the measurement error P of the GPS device, the position of the intersection Pu between the straight line L1 and the circle E is calculated as the corrected vehicle position.

  The own vehicle position correcting unit 66 corrects the vehicle position measured by the GPS device when the measurement error of the GPS device is larger than a predetermined threshold, and when the measurement error of the GPS device is smaller than the predetermined threshold, Correction of the vehicle position measured by may be prohibited. According to this configuration, the vehicle position can be corrected only when necessary for control by correcting the vehicle position measured by the GPS device only when the measurement error of the GPS device is larger than the predetermined threshold.

The control command output unit 68 adjusts the vehicle speed Pr, the steering angle, etc. so that the corrected vehicle position Pr is adjusted to points (P _{t + 1} , P _{t + 2} , P _{t + 3} ,...) On the target travel locus. Outputs a command to control. Note that a NOx concentration distribution map may be stored, and the control command output unit 68 may automatically switch from engine driving to motor driving in a region where there is a large amount of NOx.

  Next, the processing flow of the automatic travel device 3 will be described with reference to the flowchart of FIG. The data processing device 42 takes in the image data generated by the imaging device 40 (S201). Next, the data processing device 42 takes in the measurement data of the position, orientation, and attitude of the vehicle (S202). Next, the data processing device 42 takes in the information of the target travel locus from the map database 46, and extracts the data of the point closest to the measured vehicle position from the many points that define the target travel locus (S203). ). Next, the data processing device 42 calculates the relative position of the road component in the vehicle coordinate system based on the measurement data of the vehicle position by the GPS device and the data of the closest point of the target travel locus (S204).

  Next, the data processing device 42 calculates, as a search range, a pixel range in which a road component can exist in the image data captured by the left and right cameras (S205). The data processing device 42 detects the road component from the search range of the left and right image data, and calculates the relative position of the road component in the vehicle coordinate system (S206). The data processing device 42 compares the relative position of the road component calculated in step 204 with the relative position of the road component calculated in step 206, and corrects the own vehicle position (S207). Based on the corrected vehicle position, the data processing device 42 determines a control value so that the vehicle travels on the target travel locus, and executes control (S208). The data processing device 42 determines whether or not to end the above-described processing. If the process is not terminated, the process returns to S201 (S209).

[Improved method for recognizing road components]
Next, in the automatic travel map creation device 1 and the automatic travel device 3 according to the above-described embodiment, an improved structure of the imaging devices 10 and 40 attached to the vehicle, and an image photographed by the imaging devices 10 and 40. A method for detecting road components from data will be described.

As shown in FIG. 29, the imaging devices 10 and 40 are configured by attaching a stereo lens unit 72 to one CCD camera 70. In the stereo lens unit 72, two lenses 74H and 74L are arranged so as to be separated from each other by about 50 mm. The light beam collected by the upper lens 74H is reflected by the _two mirrors 76H ₁ and 76H ₂ and reaches the CCD camera 70. Similarly, the light beam condensed by the lower lens 74L is reflected by the _two mirrors 76L ₁ and 76L ₂ and reaches the CCD camera 70. The imaging devices 10 and 40 generate one image data I including two image portions obtained by viewing the imaging target O in stereo, and output the image data I to the data processing devices 12 and 42. By configuring the imaging devices 10 and 40 in this manner, the image data I can be made suitable for image processing, and the imaging devices 10 and 40 can be reduced in price or downsized.

  The imaging devices 10 and 40 described above are attached to the position of the side mirror of the vehicle as shown in FIG. By directing the camera line-of-sight direction of the imaging devices 10 and 40 obliquely downward, it is possible to detect white lines, side grooves, guide rails, parallel running vehicles, and the like on the side of the vehicle. Note that, by selecting a lens with a wide field of view, it is preferable that the side of the vehicle can be imaged over the entire range (180 °), as shown in FIG.

  Next, the flow of processing by the data processing devices 12 and 42 will be described with reference to the flowchart of FIG. The process shown in the flowchart of FIG. 31 is a process performed by the road component recognition unit 18 of the map creation device 1 and the object recognition unit 64 of the automatic travel device 3.

  First, the data processing devices 12 and 42 acquire image data captured by the imaging devices 10 and 40 (S301). An example of the acquired image data is shown in FIG. The data processing devices 12 and 42 extract the horizontal edge from the image data, and obtain the result shown in FIG. 32B (S302). Next, the data processing devices 12 and 42 calculate a search range in which road component recognition processing is to be performed (S303). The derivation of the search range performed here is the same processing from step 202 to step 205 described with reference to the flowchart of FIG.

  Next, as shown in FIG. 32C, the data processing devices 12 and 42 assume a straight line extending in the vertical direction in the horizontal edge extraction result, and pixels at a plurality of intersections of the horizontal edge and the straight line. Find the position. Then, the data processing devices 12 and 42 obtain an intersection corresponding to the same road component from the upper image portion and the lower image portion using the epipolar constraint (S304). As a result, as shown in FIG. 32 (c), the intersection corresponding to the guide rail in the upper image portion and the intersection corresponding to the guide rail in the lower image portion are obtained as a pair of intersections having a corresponding relationship with each other. It is done. Similarly, an intersection point corresponding to the side groove of the upper image portion and an intersection point corresponding to the side groove of the lower image portion are obtained as a pair of intersection points corresponding to each other.

  Next, the data processing devices 12 and 42 calculate the three-dimensional position of the intersection in the vehicle coordinate system from the pair of intersections that are in a corresponding relationship with each other (S305). As shown in FIG. 33, the lateral edge extends along the front-rear direction of the vehicle, and when viewed from the front of the vehicle according to the arrow A, the intersection is located on the side of the vehicle as shown in FIG. .

  Next, the data processing devices 12 and 42 perform a process of determining the type of road component (white line, curbstone, guardrail, gutter, road boundary, etc.). Here, regarding the intersection having a height equal to or higher than the threshold Th1 set in advance from the plane on which the own vehicle is placed, the data processing devices 12 and 42 are such that the type of road component is an obstacle such as a guide rail. Determine. However, the type of the road component is unknown at the intersection having a height within the predetermined threshold Th1 from the plane on which the host vehicle is placed. Therefore, the data processing devices 12 and 42 perform the following processing.

As shown in FIG. 35A, the data processing devices 12 and 42 store rectangular ranges R _H and R _{L including} two points corresponding to each other in each of the upper image portion and the lower image portion. Then, as shown in FIG. 35 (b), a plurality of small regions SR ₁ , SR ₂ , SR ₃ are further set between the two points D ₁ , D _{2 in} the rectangular ranges R _H , R _L. . Here, by performing pattern matching processing such as normalized correlation and difference summation when setting a plurality of small regions, the small region of the upper image portion has a higher correlation value than the small region of the lower image portion. Is set to a position. Note that the plurality of small regions may be set to overlap each other.

  Next, the data processing devices 12 and 42 determine the three-dimensional position of the small area in the vehicle coordinate system based on the pixel position of the small area of the upper image portion and the corresponding pixel position of the small area of the lower image portion. calculate. As shown in FIG. 36 (a), when the three-dimensional position of the small area is lower than the threshold value Th2 from the plane on which the host vehicle is mounted, the data processing devices 12 and 42 are side grooves. Judge that. On the other hand, as shown in FIG. 36 (b), the data processing devices 12 and 42, when the three-dimensional position of the small area is within the range of the threshold Th2 from the plane on which the host vehicle is placed, It is determined that the line is a white line or a road pattern.

  Note that, in the above-described road component recognition processing, the data processing devices 12 and 42 determine that the road component is a gutter when the three-dimensional position of the small area is lower than the threshold Th2 from the plane on which the vehicle is placed. However, when the position of the region between the lateral edges is lower than the plane on which the vehicle is placed and is aligned vertically with respect to the plane, the data processing apparatus determines that the road component is in the side groove. It may be determined.

  In addition, when there is a parallel running vehicle C2 that is running parallel to the side of the host vehicle C1, the data processing devices 12 and 42 perform processing for determining the presence of the parallel running vehicle C2. Since the parallel running vehicle C2 enters the field of view of the imaging devices 10 and 40 as shown in FIG. 37, when the parallel running vehicle C2 is directly beside the host vehicle C1, the inside of the frame shown in FIG. 38 is photographed. Image data is obtained.

  The data processing devices 12 and 42 obtain image data captured by the imaging devices 10 and 40 (S301). As shown in FIG. 39A, the image data includes an upper image portion and a lower image portion in which the side surface of the parallel running vehicle C2 is photographed. Next, the data processing devices 12 and 42 extract the horizontal edge from the image data, and obtain the result shown in FIG. 39B (S302). The result of FIG. 39B includes lateral edges corresponding to the roof, window frame, body line, lower surface of the body, and the like.

  Next, as shown in FIG. 39C, the data processing devices 12 and 42 assume a straight line extending in the vertical direction in the horizontal edge extraction result, and pixels at a plurality of intersections of the horizontal edge and the straight line. Find the position. Then, the data processing devices 12 and 42 obtain an intersection corresponding to the same part of the parallel running vehicle from the upper image portion and the lower image portion using the epipolar constraint (S304). As a result, the intersection corresponding to the roof of the upper image portion and the intersection corresponding to the roof of the lower image portion are obtained as a pair of intersections having a corresponding relationship. Similarly, with respect to the window frame, the body line, the lower surface of the body, and the like, the intersection of the upper image portion and the intersection of the lower image portion are obtained as a pair of intersection points that correspond to each other.

  Next, the data processing devices 12 and 42 calculate the three-dimensional position of the intersection in the vehicle coordinate system based on the pixel position of each pair of intersections. The data processing devices 12 and 42 determine whether or not the vehicle is a parallel running vehicle based on the calculated three-dimensional position of each intersection. The distance and the relative angle of the parallel running vehicle with respect to the host vehicle are output. In order to determine whether or not the vehicle is a parallel running vehicle based on the three-dimensional position of each intersection, it is determined that the height of a large number of intersections is greater than or equal to a predetermined threshold value, It can be determined from.

  In addition to the recognition processing described above, the data processing devices 12 and 42 perform processing for recognizing the types of all road components and obstacles, and output the recognition results (S309).

[Improvement of image pickup device arrangement]
Next, in the automatic traveling device 3 according to the above-described embodiment, an improved arrangement of the imaging device 40 attached to the vehicle and a usage method thereof will be described. As shown in FIG. 40, four imaging devices 40A, 40B, 40C, and 40D are attached to the vehicle. Specifically, the vehicle, the imaging device 40C for capturing the image pickup apparatus 40A for capturing a front side of the region R _A of the vehicle, the image pickup device 40B for photographing a left region R _B of the vehicle, the right area R _C of the vehicle and, An imaging device 40D that captures an area _RD on the rear side of the vehicle is provided. Here, each of the imaging devices 40A to 40D may generate stereo-viewed image data, or may be a monocular camera that generates normal image data that is not stereo-viewed.

As described above, when the four image pickup devices 40A to 40D are provided in the vehicle, the image pickup areas of the two image pickup devices are partially located at the left diagonal front, right diagonal front, left diagonal rear and right diagonal rear of the vehicle. Is duplicated. Therefore, in the image data generated by the two imaging devices, the data processing device can calculate the position of the road component by using the image data of the portion where the imaging regions overlap as stereo image data. it can. For example, as shown in FIG. 40, the diagonally forward left of the vehicle, and the imaging region R _A of the imaging device for photographing the front of the vehicle, overlap with the imaging region R _B of the imaging apparatus for capturing a left side of the vehicle Yes. Further, in the left oblique rear of the vehicle, it overlaps with the imaging region R _B of the imaging apparatus for capturing a left side of the vehicle, and the imaging region R _D of an imaging device for photographing the rear of the vehicle. The data processing device 42 uses the image data generated by the two imaging devices as stereo-viewed image data when there are road components such as guide rails and gutters in the overlapping imaging regions in this way, The position of road components and gutters can be calculated.

Furthermore, the data processing apparatus can estimate the position of the road structure in the non-overlapping region based on the calculated position of the road structure. For example, when a straight guardrail or a side groove is detected at the left front of the vehicle, and when a straight guardrail or a side groove is detected at the left rear of the vehicle, the left front straight line and the left rear rear are detected. If straight line is collinear, in the region R _B on the left side of the vehicle it can be estimated that there is such as a guide rail or groove in the position of the straight line.

  In addition, when the four imaging devices 40A to 40D are provided in the vehicle as described above, the data processing device 42 processes the image data obtained by the four imaging devices 40A to 40D, so that FIG. A map around the vehicle, as shown, can be generated. This map shows the locations of road components (guardrails, gutters, etc.) and obstacles (trees, triangular cones, etc.) around the vehicle, and because there are road components and obstacles. The area where the blind spot is invisible to the vehicle is shown in black.

  In order to generate such a map, the data processing device 42 performs a process of calculating a blind spot area. This process will be described with reference to FIGS. 42 (a) and (b). Fig.42 (a) is the figure which looked at the imaging device 40, the obstruction, and its blind spot from the horizontal direction, and FIG.42 (b) is the figure which looked at them from the upper direction. As shown in FIG. 42A, the data processing device 42 calculates a distance L that is followed by an area that becomes a blind spot on the plane on which the host vehicle is placed. Further, as shown in FIG. 42 (b), the data processing device 42 calculates the width W of a region that becomes a blind spot on the plane on which the host vehicle is placed. Then, the data processing device 42 sets a range surrounded by the calculated distance L and width W as a blind area.

It is a block diagram which shows the map preparation apparatus for automatic driving | running | working. It is a figure for demonstrating the position calculation of a road component. It is a figure for demonstrating coordinate transformation. It is a figure which shows a world plane rectangular coordinate system. It is a figure which shows a world plane coordinate system. It is a graph which shows the measurement error of the position of a road composition. It is a figure which shows a series of measurement points of the driving | running | working locus | trajectory of a vehicle. It is a figure which shows the driving | running | working locus | trajectory of a vehicle. It is a figure which shows the weighting with respect to a driving | running | working locus. It is a figure which shows the running locus after weighting. It is a figure which shows another weighting with respect to a driving | running | working locus. It is image data which shows a driving locus map. It is image data which shows a running locus weight addition map. It is a figure for explaining calculation of a target run locus. It is image data which shows a road component map. It is image data which shows a road component map. It is image data which shows the road component map of each time. It is image data which shows a road component weight addition map. It is image data which shows the presence position of a road component. It is a figure which shows the hierarchical structure of a weight addition map. It is a figure for demonstrating the information memorize | stored in a map database. It is a figure for demonstrating the information memorize | stored in a map database. It is a flowchart which shows the process of a map production apparatus. It is a block diagram which shows an automatic traveling apparatus. It is a figure for demonstrating the setting of a search range. It is a figure for demonstrating correction | amendment of the own vehicle position. It is a figure for demonstrating correction | amendment of the own vehicle position. It is a flowchart which shows the process of an automatic traveling apparatus. It is a figure which shows the modification of an imaging device. It is a figure which shows arrangement | positioning of an imaging device. It is a flowchart which shows the process of a data processor. It is a figure for demonstrating calculation of the position of a road component. It is a figure which shows the position of the extracted horizontal direction edge. It is a figure which shows the positional relationship of a vehicle and a horizontal direction edge. It is a figure for demonstrating the detection of a side groove. It is a figure for demonstrating the detection of a side groove. It is a figure for demonstrating the detection of a parallel running vehicle. It is a figure for demonstrating the detection of a parallel running vehicle. It is a figure for demonstrating the detection of a parallel running vehicle. It is a figure which shows arrangement | positioning of an imaging device. It is a map which shows the road structure around a vehicle, an obstruction, and those blind spots. It is a figure for demonstrating calculation of the range of a blind spot.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Automatic travel map creation apparatus, 3 ... Automatic travel apparatus, 10 ... Imaging apparatus, 12 ... Data processing apparatus, 14 ... Vehicle position / orientation / attitude detection part, 16 ... Image data capture part, 18 ... Road component Recognizing unit, 20 ... position calculating unit, 22 ... maximum error calculating unit, 24 ... measurement information writing unit, 26 ... measuring point information storage unit, 28 ... same road information extracting unit, 30 ... same road information combining unit, 32 ... map Data writing unit, 34, 46 ... map database, 40 ... imaging device, 42 ... data processing device, 44 ... vehicle position / orientation / attitude detection unit, 48 ... image data capturing unit, 50 ... road component position calculating unit, 52 ... Search distance setting unit, 56 ... In-image position estimation unit, 58 ... Error effect calculation unit, 60 ... Minimum search frame setting unit, 62 ... Recognition object type setting unit, 64 ... Object recognition unit, 66 ... Own vehicle Position correction unit, 68 ... Command output unit.

Claims (6)

Storage means for storing a map in which an area along a road is divided into a number of small areas, and a weight associated with the weight indicating the presence of a measurement target is associated with each small area;
Using a plurality of position data of the measurement object obtained by vehicle equipped with the measuring instrument runs a multiple of the same road, a small area corresponding to the position data, adds a weight that indicates the presence of the measurement object And the same road information combining means for specifying the position of the measurement target based on the weights associated with a large number of small areas of the map ;
Equipped with a,
The weight added to the map has a distribution that is maximum at the position of the measurement target and decreases as the distance from the measurement target position increases.
The automatic travel map creation device is characterized in that the distribution of weights added to the map is a distribution corresponding to a position error of a measurement target including at least an error caused by a reception state of GPS radio waves . 2. The automatic travel map creation device according to claim 1, wherein the position error of the measurement target further includes an error caused by camera distance resolution. The position error of the measurement target further includes at least one error of a lateral acceleration detection sensor, a steering angle detection sensor, a roll sensor, a pitch sensor, a yaw sensor, and a vehicle height sensor. Automatic map creation device. The weight distribution G added to the map is calculated by the following equation:
(Where u is the horizontal position and σ = maximum error Err / 3)
The automatic travel map creation device according to any one of claims 1 to 3.   The map for creating an automatic travel according to any one of claims 1 to 4, wherein the measurement target is a travel locus of a vehicle.   The automatic travel map creation device according to any one of claims 1 to 4, wherein the measurement object is a road component.