JP2017120551A - Autonomous traveling device - Google Patents

Abstract

An object of the present invention is to easily and quickly recognize a marker by further searching even when a marker cannot be well recognized, and to accurately estimate the position of an autonomous mobile device. [Solution] A PC 10 uses information about the initial position, orientation, and goal point inputted from an operation unit 11, and surrounding information detected by detection means such as a laser range finder 13 and an ultrasonic sensor 23 during autonomous driving. determines the moving direction and moving speed, and the drive/IO control section 20 drives and controls the motor 24. The vehicle autonomously travels to a desired location by manual operation in advance, and the PC 10 detects surrounding information using a detection means, and creates and stores map information. During autonomous driving, the camera 12 acquires image data at least near the goal point, detects markers, and calculates the self-position. If the marker cannot be detected, an area where the marker is likely to be located is specified from the map information, the image is enlarged for that area, and the marker is detected again. [Selection diagram] Figure 1

Description

  The present invention relates to an autonomous traveling device.

  An autonomous traveling device that autonomously travels to a place when a destination is specified is used in, for example, an automatic guided machine. In the autonomous traveling device, a technique is already known in which a camera is installed and marker detection is performed, self-position is estimated in units of several centimeters, and high-precision alignment is performed.

  For example, Patent Document 1 discloses a technique in which a predetermined figure is used as a marker for the purpose of estimating the position from the coordinates of the camera, and the marker is photographed by a camera attached to the autonomous mobile device. Yes. And the image data which image | photographed the marker is analyzed, and the position of an autonomous traveling apparatus is estimated by calculating | requiring the coordinate of a camera.

  However, the marker detection in such a conventional autonomous traveling device has a problem that it is not recognized when a figure constituting the marker cannot be recognized. Further, in order to increase the recognition accuracy, if an image search for detecting a figure is further attempted, there is a problem that the search becomes a full search and takes time.

  The present invention has been made to solve the above problem. Even when the marker cannot be recognized well, the marker can be easily and quickly recognized, and the position of the autonomous traveling device can be accurately determined. The purpose is to make estimation possible.

In order to achieve the above object, an autonomous traveling apparatus according to the present invention has an input means for an operator to input information, a detection means for detecting surrounding information, and a drive means for moving by driving a motor. Determining a moving direction and a moving speed based on the initial position and direction information input by the input means and each information of the goal point, and surrounding information detected by the detecting means during autonomous traveling, and the motor Control means for driving and controlling, manual traveling means for manually traveling a place where autonomous traveling is desired in advance, surrounding information is detected by the detecting means during traveling by the manual traveling means, and map information is created and stored Map information creating / storing means.
The detection means includes a camera that acquires at least image data in the vicinity of the goal point, and the control means detects a predetermined marker from the image data acquired by the camera and detects its own position. An autonomous traveling device having a function of calculating
If the control means cannot detect the marker, the pixel of the marker is searched for an area where the marker is likely to be found from the map information, and an area where the marker is likely to be specified is specified. On the other hand, an image enlargement process using the image data is performed, and the marker is detected again.

  According to the present invention, even when the marker cannot be recognized well, it is possible to recognize the marker easily and quickly by further searching, and it is possible to accurately estimate the position of the autonomous traveling device.

It is a block diagram which shows the structure of one Embodiment of the autonomous running apparatus by this invention. It is a flowchart which shows the flow of 1st Example of the process in connection with this invention performed by CPU by the side of PC10 in FIG. It is a flowchart which similarly shows the flow of 2nd Example of the process in connection with this invention. It is a flowchart which similarly shows the flow of 3rd Example of the process in connection with this invention. It is a figure which shows the example of a figure of the marker used for the 1st-3rd Example. It is a figure for demonstrating the corner position detection of a marker. It is a figure which shows an example of the map information used by each Example. It is a figure for demonstrating preserve | saving the marker positional information on the map information of FIG. 7, and inputting each information of the initial position and direction of an autonomous mobile device, and a goal point. It is a figure which shows the example of the marker corresponding coordinate point for recognizing the marker shown in FIG.

Hereinafter, embodiments for carrying out the present invention will be specifically described with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of an embodiment of an autonomous mobile device according to the present invention.
In this autonomous traveling device 30, a personal computer (hereinafter abbreviated as "PC") 10 and a drive / IO control unit 20 constituting a drive circuit and an IO control circuit are connected to each other so that signals and information can be exchanged. ing.

The PC 10 includes an operation unit 11 and a camera 12 and a laser range finder 13. The operation unit 11 is input means for an operator to input information, and the camera 12 and the laser range finder 13 are detection means for detecting surrounding information.
The camera 12 acquires at least image data in the vicinity of the goal point. The laser range finder 13 is a small distance measuring device (laser distance meter) that transmits and receives a laser and measures an accurate distance to an object.
The drive / IO control unit 20 is connected to a battery 21 serving as a power source, a plurality of ultrasonic sensors 23, a plurality of motors 24, and the same number of encoders 25 through power supply lines and signal lines.

The plurality of motors 24, their driving force transmission mechanisms, wheels and steering mechanisms are moving means for moving by driving the motors. The plurality of ultrasonic sensors 23 are also detecting means for detecting surrounding information.
The PC 10 uses the operator's input from the operation unit 11 as a trigger to create map information during travel by manual operation and store (save) the map information. During autonomous traveling, the PC 10 determines the travel route and travel speed using the map information, and issues a drive request to the drive / IO control unit 20.

The PC 10 also senses the peripheral information of the autonomous traveling device using all the ultrasonic sensors 23 via the drive / IO control unit 20 in order to create map information by manual operation and detect obstacles during autonomous traveling. These processes are executed by software processing by the CPU of the PC 10.
The drive / IO control unit 20 receives information on the movement route and the movement speed from the PC 10, drives and controls a plurality of motors 24 to execute wheel rotation, steering, and the like. 25 is detected and fed back to the PC 10.

  Further, the drive / IO control unit 20 makes a request for sensing peripheral information to the plurality of ultrasonic sensors 23. As a result, the drive / IO control unit 20 feeds back peripheral information obtained from each ultrasonic sensor 23 to the PC 10.

  For unexpected obstacle approach, the drive / IO control unit 20 requests the ultrasonic sensor 23 to sense peripheral information at regular intervals before driving / controlling the motor 24 according to route information from the PC 10. To grasp the surrounding information. For this reason, an instruction to decelerate or stop the motor 24 is directly given from the PC 10 side to the drive / IO control unit 20 side to avoid collision with an obstacle.

Since the drive / IO control unit 20 also has a CPU and is controlled independently of the process on the PC 10 side, the CPU process is exclusively used for the process on the PC 10 side and the drive control of the motor 24 is not awaited.
The battery 21 supplies electricity to operate each module.

A camera 12 for capturing a marker and capturing it as an image is connected to the PC 10 and acquires an image of, for example, 640 × 480 pixels as an RGB image at 30 FPS (frame / second) in real time.
When this autonomous traveling device reaches the vicinity of the goal point where high-accuracy position detection around the destination is necessary, the PC 10 acquires an image from the camera 12. Then, the CPU on the PC 10 side performs marker detection processing on the image to determine the route, and moves toward the destination where the marker is located.

In the autonomous traveling device 30, the PC 10 and the drive / IO control unit 20 perform the functions of a control unit, a manual traveling unit, and a map information creation / storage unit.
In the function as the control means, based on each information of the initial position and direction and the goal point input by the operation unit 11 which is an input means, and the surrounding information detected by the detection means during autonomous traveling, The direction and moving speed are determined, and the motor 24 is driven and controlled.
The function as the manual travel means is a function for causing a place where autonomous travel is desired to travel in advance by manual operation.

In the function as the map information creation / storage means by the PC 10 and the drive / IO control unit 20, the surrounding information is detected by the detection means during the travel by the manual travel means, and the map information is created and stored in the internal nonvolatile memory. To do.
The function of the PC 10 and the drive / IO control unit 20 as control means also detects a predetermined marker from the image data acquired by the camera 12 and calculates its own position. However, if the marker cannot be detected by the function of the control means, the pixel of the marker is searched for the area where the marker is highly likely from the stored map information, and the area where the marker is likely is specified. Then, an image enlargement process using the image data is performed on the area, and the marker is detected again.

Several embodiments will be described below. The hardware configuration of the autonomous mobile device is basically the same as the configuration shown in FIG. 1, and route determination or software processing by the CPU in the PC 10 is performed. This is an example in which the processing content of map generation is different.
The first embodiment will be described with reference to FIG. FIG. 2 is a flowchart showing the flow of the first embodiment of the process relating to the present invention, which is performed by the CPU on the PC 10 side for the image (still image) input from the camera 12.

In the first embodiment and the second and third embodiments described later, as shown in FIG. 5, a graphic (pattern) having a plurality of black square points in a white background in a square black frame is used as a marker. Use. Then, the image input from the camera 12 is searched for whether there is a continuation of black pixels that are supposed to constitute the marker.
In the flowchart of FIG. 2, the half-dotted portion, that is, the processes of steps S8 and S10 are characteristic processes according to the present invention, and the other processes are general processes of marker detection.

When the CPU of the PC 10 starts the process shown in the flowchart of FIG. 2, first, in step S <b> 1, a square extraction process is performed on the image data acquired from the camera 12. In this quadrangular extraction, each pixel of the acquired image data is simply binarized, a contour is extracted from the binary image data, and a linear approximation of the contour is performed. Then, the feature determination of the quadrangle is performed based on whether there are four corners, and the quadrangle (square black frame and black square point inside it) constituting the marker is extracted.
In the next step S2, the CPU determines whether or not a square constituting the marker has been detected. If it is determined that it has detected, the CPU determines in step S3 whether or not there are two or more squares in the black frame. As a result, if it is determined that there are two or more, the CPU proceeds to step S4 to perform perspective projection conversion.

In the perspective projection conversion, the CPU performs perspective conversion using the extracted four corners of the rectangle and converts the black square so that the marker is in front of the image.
Thereafter, the CPU confirms the marker direction in step S5. Therefore, normalization is performed to 100 × 100 pixels for a block of black pixels that seems to correspond to square points in the black frame. Then, the normalized image is divided into an upper left corner (0 °), an upper right corner (90 °), a lower right corner (180 °), and a lower left corner (270 °), and the number of black pixels of each divided image. Count.
As a result, the direction of the divided image with the largest count value of the number of black pixels is determined as the direction in which the marker is facing.

  Next, the CPU performs corner position detection in step S <b> 6 to increase the accuracy of the corner positions of the squares constituting the marker. That is, using a known cvFindCornerSubPix function, iterative processing is performed in order to detect a corner or saddle point with sub-pixel accuracy. Corner positioning with subpixel accuracy is based on the principle that each vector from the center of a neighboring area to a point located in that area is orthogonal to the image gradient. Repeat the calculation until the quantity falls within the given threshold.

Thereafter, in step S7, the CPU calculates the posture of the object (here, the camera 12) from the corresponding 3D-2D points. That is, when a set of point coordinates on an object and a corresponding projection point on an image is given from a camera matrix and a distortion coefficient, which will be described later, the camera coordinates (x, y, θ) are calculated using a known solvePnP function. To lower.
Camera coordinates are determined to minimize reprojection errors. That is, the sum of squares of the distance between the observed projection point and the point on which the array of point coordinates on the object in the object coordinate space is projected is determined to be minimum. Further, a rotation vector and a translation vector, which will be described later, are also obtained using the solvePnP function.

The source code is as follows.
cv :: solvePnP (worldPoints, imagePoints, intrinsicMatrix, distortionCoeffs, rotationVector, translationVector)
worldPoints: Coordinate points of both the square black frame and the inner square corner (fixed value by marker)
imagePoints: Coordinate points obtained from corner detection
intrinsicMatrix: Camera internal parameters (values obtained in advance)
distortionCoeffs: Distortion coefficient (value obtained in advance)
rotationVector: rotation vector
translationVector: Translation vector

Then, in the last step S8, the CPU calculates the camera self-position estimation information in consideration of the presence or absence of the enlargement process, and then ends the process of FIG. In this step S8, camera coordinates (x, y, θ) are obtained from the rotation vector and translation vector obtained in the object posture calculation in step S7. θ is a YAW angle (yaw angle), that is, a rotation angle around the vertical axis.
Also, it is determined whether or not enlargement processing has been performed in order to detect the marker. If enlargement processing is being performed, camera coordinates (x, y, θ) are calculated in consideration of the magnification used for the enlargement processing.
This camera coordinate is coordinate information of the camera 12 mounted on the autonomous mobile device, and corresponds to position information up to the marker. Therefore, calculating the camera coordinates (x, y, θ) in consideration of the magnification used for the enlargement process calculates the position information up to the marker in consideration of the magnification used for the enlargement process.

  The CPU of the PC 10 can detect the marker if it cannot be determined that the square constituting the marker has been detected in step S2 described above, or if it cannot be determined in step S3 that there are two or more squares in the black frame. This is the case. In that case, the process proceeds to step S9 to determine whether or not enlargement processing has been performed. As a result, when it is not determined that the enlargement process has been performed (determined that the enlargement process has not been performed), the process proceeds to step S10 to perform the enlargement process.

In the enlargement process, the marker information is searched from the stored map information by referring to the coordinate information (x, y, θ) of the marker, and the marker pixel is searched for an area where the marker is likely to exist. Identify such areas. Then, an image enlargement process is performed on the specified area using image data acquired by the camera. In this enlargement process, it is preferable to sharpen corners and enhance edges so that quadrilateral extraction is easy.
Thereafter, the markers are detected again in steps S1 to S3. If the marker is still not detected, and the process proceeds to step S9 and it is determined that the enlargement process has been performed, the process proceeds to step S11, where no marker is recognized, and the process is terminated.

In the present invention, in order to increase the detection accuracy of the marker when the marker cannot be detected, the position information of the marker is stored in the map information created in advance, and the image is efficiently displayed for the region where the marker is highly likely to exist. The enlargement process is performed.
For example, even if the entire marker is imaged by the camera 12, there are cases where the marker cannot be detected because the distance is long and the number of pixels constituting the marker is small. In such a case, the marker can be detected by efficiently enlarging the area where the marker is likely to be present. Thereby, the position of the camera 12 can be reliably calculated, and the self-position of the autonomous mobile device can be estimated with high accuracy.
As described above, even when the marker cannot be recognized, the marker can be recognized easily and quickly after the enlargement process, and the position of the autonomous mobile device can be estimated with high accuracy.

FIG. 3 is a flowchart showing the flow of a second embodiment of the process related to the present invention performed by the CPU on the PC 10 side for an image (still image) input from the camera 12.
The basic flow of the flowchart shown in FIG. 3 of the second embodiment is the same as that of the first embodiment shown in FIG. 2, but it is approximate that marker detection is performed using image data from the camera. The only difference is that it is performed only when it arrives at the goal point.

In order to prevent the marker from being only near the goal point, and to detect a figure similar to the marker when the autonomous mobile device is on the way to the goal point, it will not be recognized as a goal. In the example, the place where the marker is detected is limited to the vicinity of the goal point.
Therefore, in the flowchart shown in FIG. 3, when the CPU on the PC 10 side starts processing, it is first determined in step S0 whether or not an approximate goal point has been reached according to the route based on the map information. As a result, the processing is terminated without doing anything until it is determined that it has arrived at the approximate goal point. This is repeated at regular time intervals, and if it is determined in step S0 that an approximate goal point has been reached, the process proceeds to the process after the quadrangle extraction in step S1.
Since each process of step S1-S11 is the same as Example 1 demonstrated by FIG. 2, description is abbreviate | omitted.

  As described above, in the second embodiment, only when the goal point on the map is recognized based on the map information stored in advance by the CPU on the PC 10 side and the autonomous mobile device arrives in the vicinity of the goal point, The same processing as in the first embodiment is executed. That is, a predetermined marker is detected from the image data acquired by the camera 12 to calculate its own position. However, if the marker cannot be detected, the pixel of the marker is searched for an area where there is a high possibility of the marker from the map information, and the area where the marker is likely to be identified. Then, an image enlargement process using image data is performed on the area, and marker detection is performed again.

  Therefore, it is possible to accurately detect the marker, calculate the position of the camera 12, accurately estimate the self-position of the autonomous traveling device, and perform control to accurately travel to the goal point and stop. Moreover, even if there is a figure resembling a marker on the way of the autonomous traveling device toward the goal point, there is no risk of erroneously detecting it and recognizing it as a goal.

FIG. 4 is a flowchart showing the flow of a third embodiment of the process related to the present invention performed by the CPU on the PC 10 side for an image (still image) input from the camera.
The basic flow of the flowchart shown in FIG. 4 of the third embodiment is the same as that of the second embodiment shown in FIG. However, if the marker is not detected even if the marker detection is performed by performing the enlargement process in the direction in which the marker is expected to be near the goal point, the difference is that the detection method is switched to another detection method.

In the third embodiment, if the CPU of the PC 10 performs the image enlargement process in step S10 and no marker is detected even if the marker detection is performed again in steps S1 to S3, it is determined that there is no marker in step S11. After the recognition, the process at step S12 is performed. In step S12, the marker detection using the camera 12 is switched to the detection of ambient information using the laser range finder (LRF) 13. Then, this process ends.
The surrounding information is detected by the laser range finder 13 by preliminarily holding the shape of the goal by recognizing the shape of the plane using the laser range finder 13 shown in FIG. Control is performed so as to approach the direction in which the reflected wave is detected.

Marker detection using a camera can acquire several frames of images per second. For this reason, it is possible to switch to detection of surrounding information using the laser range finder 13 because the marker cannot be detected in a certain frame, and in the next frame in which the direction is determined, the marker detection using the camera is performed again to approach the goal. It is done.
Since the autonomous traveling apparatus shown in FIG. 1 also includes an ultrasonic sensor 23, switching to detection of ambient information using the ultrasonic sensor 23 instead of detection of ambient information using the laser range finder (LRF) 13 is performed. May be.

Or you may switch to the detection of the surrounding information using the laser range finder (LRF) 13 and the ultrasonic sensor 23. FIG. Using one or both of the laser range finder (LRF) 13 and the ultrasonic sensor 23, the PC 10 can calculate the distance to the surrounding obstacle and determine the travel route.
The autonomous traveling device shown in FIG. 1 includes the laser range finder (LRF) 13 and the ultrasonic sensor 23, but may include only one of them.
Step S0 in FIG. 4 is the same as that of the second embodiment described with reference to FIG. 3, and steps S1 to S11 are the same as those of the first embodiment described with reference to FIG.
This increases the possibility that the autonomous mobile device can accurately reach the goal point even when the marker cannot be detected from the image data acquired by the camera.

FIG. 5 shows an example of a marker graphic used in the first to third embodiments. The marker figure is a binary figure of white (background) and black, and a plurality of small black squares exist in the outer frame constituting the black. As shown in the flow charts of FIGS. 2 to 4, a plurality of black squares constituting the marker are extracted, and after the marker is recognized, perspective transformation is performed, and the camera 12 that captured the marker is captured. Predict camera coordinates (x, y, θ) indicating position and angle.
The corner position detection will be described with reference to FIG. Corner positioning is calculated based on the idea that each vector from the center q of the neighboring area shown in FIG. 6 toward the point p located in the area is orthogonal to the image gradient at the point p. It can be expressed by Equation 1.

Here, DI _pi is an image gradient at a point pi in the neighborhood region q. q is obtained as a value that minimizes this εi. By setting εi to 0, the simultaneous equations with Equation 2 are established.

Here, the image gradient is summed in the region near q. When the primary gradient is G and the secondary gradient is b, the relationship of Equation 3 is established with q.
Note that “T” in Equations 1 and 2 indicates a transposed matrix.

The center of the search window shown in FIG. 6 is reset to the new value q, and the calculation is repeated until the change amount of the value falls within the given threshold value. Find the position (coordinates).

The position of the camera is a point P whose three-dimensional position is known and its observed coordinates p on the image in the process of “calculation of object posture from corresponding three-dimensional and two-dimensional points” in step S7 in FIGS. It can be obtained from a plurality of correspondences. This problem is known as the Perspective-n-Pont Problem.
The external parameters (rotation vector and translation vector) of the camera are estimated by inputting the array of points with known three-dimensional positions, the array of observation coordinates on the image, the internal parameter matrix of the camera, and the distortion coefficient vector, respectively.
The internal parameter matrix and the distortion coefficient vector of the camera are parameters that can be obtained by performing calibration processing for each camera in advance.

The camera matrix is a camera internal parameter matrix inherent to the camera and can be represented by a 3 × 3 vector.
The distortion coefficient vector is a vector having 4, 5 or 8 elements. If this vector is NULL or empty, the distortion factor is considered to be zero.

The camera position (x, y, θ) is obtained from the external parameters (rotation vector and translation vector) of the camera.
The rotation vector and the translation vector are calculated by the following vectors.
Rotation vector = (r1, r2, r3)
Translation vector = (h1, h2, h3)

x = h2 / 1000 [mm]
y = h1 / 1000 [mm]
θ is obtained using the same quaternion (quaternion) as in the prior art using the rotation vector r2 and converted to a YAW angle (yaw angle).
x and y can be obtained from the above equation if the marker is detected without performing the enlargement process, and when the enlargement process is performed, the enlargement ratio is obtained by multiplying the coefficient prepared in advance. Since θ is irrelevant to the enlargement process, the calculation formula does not change depending on the presence or absence of the enlargement process.
x = h2 * KX / 1000 [mm] KX: coefficient according to enlargement ratio y = h1 * KY / 1000 [mm] KY: coefficient according to enlargement ratio

FIG. 7 shows an example of the map information used in each embodiment described above. This is an example of map information necessary for the autonomous traveling device to autonomously travel. This map information can be expressed two-dimensionally as shown in FIG.
The autonomous traveling device 30 is manually traveled by manual travel means. During the manual travel, the PC 10 shown in FIG. 1 uses the laser range finder 13 to scan the distance to the obstacle on the plane where the laser range finder 13 is installed. Accordingly, surrounding information is detected, and a location where the autonomous traveling device 30 can travel (white region in FIG. 7) and a location where the autonomous traveling device 30 cannot travel (gray region) are identified and created as map information. The PC 10 stores the map information in the memory.

In the initial state, all images are only images where the vehicle cannot travel (gray area). However, when the autonomous traveling device 30 is traveling manually, the laser range finder 13 scans and the laser light hits the surrounding obstacles and returns. Measure distance by time. Based on the information, map information is created by expanding a portion (white region) where the autonomous traveling device 30 can travel.
With respect to this map information, as shown in FIG. 8, position information (x, y, θ) of a plurality of markers is overwritten and saved.

FIG. 8 is a diagram for explaining that the position information of the marker is overwritten and saved in the map information of FIG. 7 and the initial position and orientation of the autonomous traveling device 30 and the goal point information are input. The position information of the markers A to C is expressed as follows.
Marker A (x1, y1, θ1)
Marker B (x2, y2, θ2)
Marker C (x3, y3, θ3)
The position information when there is another marker X is expressed as follows.
Marker X (xx, yx, θx)

Each of these markers is installed at a predetermined distance from a plurality of selectable goal points of the autonomous mobile device.
The marker position information (x, y, θ) is coordinate information (x, y, θ) of a marker set in advance from a certain origin. The number of markers and the coordinates from the origin for each marker are Since it is determined, it can be stored in the ROM of the PC 10. Therefore, the pattern of each marker may be the same pattern as shown in FIG.

Here, an example of marker pattern discrimination information will be described. For example, in the case of the marker shown in FIG. 5, as shown in FIG. 9, the coordinate point of each corner at the lower left of the black square outer frame and the six small black square patterns inside it is recognized. .
Therefore, the coordinate points of each corner indicated by the circled numbers 1 to 12 in FIG. 9 are stored in the ROM of the PC 10 as information on the marker corresponding coordinate points (x, y, z) as shown in Table 1, for example. .
The marker corresponding coordinate point (x, y, z) of each corner has the origin (0, 0, 0) as the marker corresponding coordinate point of the circled number 1.

If the coordinate points of these 12 corners in the pattern detected by the PC 10 coincide with the stored marker corresponding coordinate points, it can be recognized as a predetermined marker.
Since each marker has a z-axis coordinate of 0, the z-axis is assumed to be the same height as the camera, and is attached to a wall or a pillar at the same height as the camera 12 of the autonomous mobile device 30. Assumes something.

At the time of autonomous traveling, as shown in FIG. 8, the initial position of the autonomous traveling device 30 and the facing direction F are input on this map, the goal point G to be autonomously traveled is set, and autonomous traveling is requested. As a result, the autonomous traveling device 30 autonomously travels to the goal point by the built-in PC 10 determining a route based on the map information and the self-location information.
The marker coordinate information (x, y, θ) is used to grasp which marker on the map is approached. The camera coordinates (x, y, θ) are images obtained by acquiring the relative coordinates of the closest marker from the camera when the closest marker is recognized by the camera and the marker detection process is performed. Calculated from data.

  The marker is used when the autonomous traveling device 30 travels autonomously, approaches a marker near the goal point, and stops at a goal point opposite to the set marker with high accuracy (error number cm). An area where there is a high possibility that a marker is present is imaged by a camera mounted on the autonomous mobile device 30, and the PC 10 performs a marker detection process based on the image data. When the PC 10 detects the marker, the distance to the detected marker is determined from the camera coordinates, and the drive / IO control unit 20 is controlled to move in the direction of the marker.

  In this embodiment, a plurality of markers are identified by position information set in advance for map information. However, as another method of identifying a plurality of markers, the size, number, arrangement, etc. of the small black squares inside the square black frame are changed for each marker, and the autonomous traveling device discriminates them to change the markers. Can also be identified.

  Moreover, this invention is applicable not only when an autonomous running apparatus detects a marker in the vicinity of a goal point, but also when detecting a marker for each important point to pass or stop during autonomous running. Thereby, even when the marker cannot be recognized well, it is possible to recognize the marker easily and quickly by further searching, and it is possible to accurately estimate the position of the autonomous traveling device.

As mentioned above, although embodiment of this invention was described, the specific structure of each part of the embodiment, the content of a process, etc. are not restricted to what was described there.
Further, the present invention is not limited to the above-described embodiment, and it is needless to say that the present invention is not limited in any way except for having the technical features described in the claims.
Furthermore, the configuration examples, operation examples, modification examples, and the like of the embodiments described above may be changed or added as appropriate, or some of them may be deleted, and any combination may be implemented as long as they do not contradict each other. is there.

10: Personal computer (PC) 11: Operation unit 12: Camera
13: Laser range finder 20: Drive / IO control unit
21: Battery 23: Ultrasonic sensor 24: Motor
25: Encoder 30: Autonomous traveling device

JP 2014-21624 A

Claims (5)

An input means for an operator to input information;
Detecting means for detecting surrounding information;
Driving means for moving by driving the motor;
Based on each information of the initial position and direction and the goal point input by the input means, and surrounding information detected by the detection means during autonomous traveling, a moving direction and a moving speed are determined, and the motor is Control means for driving control;
Manual travel means for manually traveling in advance a place where you want to travel autonomously;
Map information creating / storing means for detecting surrounding information by the detecting means during traveling by the manual traveling means, and creating and storing map information;
Have
The detection means has a camera that acquires at least image data in the vicinity of the goal point,
The control means is an autonomous traveling device having a function of detecting a predetermined marker from the image data acquired by the camera and calculating its own position,
When the control unit cannot detect the marker, the pixel of the marker is searched for an area where the marker is highly likely from the map information to identify the area where the marker is likely to exist. On the other hand, an autonomous traveling device that performs image enlargement processing based on the image data and detects the marker again.   2. The control unit according to claim 1, wherein, when the enlargement process is performed and the marker is detected, the position information to the marker is calculated in consideration of a magnification used for the enlargement process. The autonomous traveling device described in 1.   The control means recognizes the goal point on the map based on the map information, and detects the marker by image data acquired by the camera only when it arrives in the vicinity of the goal point. If the marker cannot be detected, the pixel of the marker is searched from the map information for an area where the marker is highly likely to be calculated. 2. The method according to claim 1, further comprising: identifying an area where the marker is likely to be detected, performing an image enlargement process on the area based on the image data, and detecting the marker again. 2. The autonomous traveling device according to 2.   The autonomous traveling device according to any one of claims 1 to 3, wherein the detection unit includes at least one of a laser range finder and an ultrasonic sensor.   Means for switching from marker detection using the camera to detection of ambient information using the laser range finder or ultrasonic sensor when the control means fails to detect the marker even after performing the enlargement process; The autonomous traveling device according to claim 4, comprising: