JP3192736B2 - Recognition method for stairs of moving objects - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for recognizing stairs of a moving object, and more specifically, to a mobile robot moving in a three-dimensional view of the outside world by recognizing objects protruding from a road surface such as stairs at a high speed. It relates to a device that obtains location information.

[0002]

2. Description of the Related Art Recently, various autonomous mobile robots such as a leg type, a wheel type, and a crawler type have been proposed. In order to move such a robot, a method of moving a predetermined route according to a preset movement pattern and a mobile sensor equipped with a visual sensor for recognizing the outside world are provided, and the robot is made to recognize and judge the surrounding conditions. It is possible to use a moving method. When using the former method, it is necessary to prepare accurate position information such as map information in advance. When the latter method is used, such position information is not required, but the position information must be obtained by quickly recognizing an object protruding from a horizontal road surface such as a stairway. As the latter prior art, for example, the one described in JP-A-59-79377 is known. In this example, only one visual sensor is provided, images are obtained at different positions while moving, and binarization processing is performed to measure the width of the stairs.

[0003]

The above-mentioned prior art does not recognize the existence of the stairs itself, and focuses on measuring the main specifications on the assumption that the stairs exist. It is not easy to specify what the object is when recognizing the outside world, so it is equipped with dedicated hardware and performs complex image processing such as spatial differentiation to extract the outline of the object. There is a need.

Accordingly, an object of the present invention is to easily recognize an obstacle from a horizontal road surface such as a stair and obtain its position information, and it is possible to generally use a general computer system without special hardware or the like. It is an object of the present invention to provide a method for recognizing a staircase of a moving object which can be easily constructed even on the above.

[0005]

Means for Solving the Problems In order to solve the above problems, the present invention is configured as follows. Signs described below
With description of those, in the one of claims, have one or more visual sensors 36, the moving mobile (legged moving performs movement control information obtained therefrom with reference
Polyhedron having a substantially horizontal surface definitive robot) (staircase 5
Oite the recognition method 0), by and against the image obtained from the visual sensor performs a predetermined binarization, the
A substantially horizontal plane of the polyhedron (specifically, a plane parallel to the floor) 50a
Pixels exhibiting the found the substantially border region partitioning the horizontal plane (Black
Color tape 52) and the background in (wall surface 50b, etc.) of the pixel and the resultant (control unit the binary image of the outside world, such as with different output value 26, S1, S10, S12) , wherein
For substantially horizontal, and measuring the area in the obtained binary image (control unit 26, S1, S14), and recognizes the polyhedron based on the area of the substantially horizontal plane determined (control
Control unit 26, S1, S16 to S20) . In claim 2, one or more
Has a plurality of visual sensors 36, and information obtained therefrom.
The moving body (leg type movement)
Polyhedron (staircase 5) having a substantially horizontal plane in a robot
0) In the recognition method, the information is obtained from the visual sensor.
By performing a predetermined binarization process on the image,
A substantially horizontal plane of the polyhedron (specifically, a plane parallel to the floor) 50a
Is indicated in the boundary area (black
Color tape 52) and pixels in the background (wall surface 50b, etc.)
Produces a binary image of the outside world with different output values
(Control units 26, S1, S10, S12)
For the obtained binary image, the x direction and the y direction
From the above, the substantially horizontal plane in the obtained binary image
A plurality of feature points indicating positional features are obtained (control unit
26, S2, S102), based on the obtained feature points,
Obtain the shape of the substantially horizontal plane in the obtained binary image
(Control unit 26, S2, S104, S106, S1
08), the characteristic point and the shape of the obtained substantially horizontal plane
A plurality of predetermined portions of a substantially horizontal plane are obtained (control unit 2
6, S3, S110) and from the determined predetermined part
Calculating the three-dimensional position (three-dimensional information) of the polyhedron (control
Control unit 26, S4, S5, S6, S7) it since
It was configured as follows. According to a third aspect of the present invention,
The corresponding point (end point) is determined for each horizontal plane, and the three-dimensional position is determined.
(Control unit 26, S5, S6)
did.

[0006]

For example, in the first aspect, the polyhedron is substantially horizontal.
Since the polyhedron is recognized based on the area of the binary image of the surface ,
It has a simple configuration and does not require dedicated hardware or the like, and does not require accurate position information such as pre-made map information. As a result, it is possible to find the stairs and go up and down only by roughly instructing the moving route including the stairs in the working environment of the mobile body such as the factory or the office.
In such a limited working environment, it is not a problem to give a sufficiently high brightness to the background on the upper surface (substantially horizontal plane) of a polyhedron such as a staircase.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below by taking a bipedal legged mobile robot as an example of a moving object. FIG. 1 is an explanatory skeleton diagram showing the entire robot 1.
Each of the left and right legs has six joints (axes) (for convenience of understanding, each joint (axis) is exemplified by an electric motor that drives it). The six joints (axis) are, in order from the top, joints (axis) 10R, 10L (R on the right side, R,
Let L be the left side. The same applies hereinafter), joints (axes) 12R and 12L in the waist pitch direction (x direction), joints (axes) 14R and 14L in the same roll direction (y direction), and joints (axes) 16R and 16L in the knee pitch direction. , Joints (axes) 18R, 18L in the pitch direction of the ankle, and joints (axes) 20 in the same roll direction
R, 20L, and a foot 22R, 2
2L is attached, and at the top is the body (base) 2
4, a control unit 26 comprising a microcomputer is stored therein.

In the above description, the hip joint is a joint (axis) 10R.
(L), 12R (L), 14R (L), the ankle joint is composed of joints (axes) 18R (L), 20R (L), and the leg links are for the left and right feet, respectively. Given 6 degrees of freedom, these 6x
By driving 2 = 12 joints (axes) at appropriate angles, a desired movement can be given to the entire foot, and the user can walk in a three-dimensional space arbitrarily. The thigh links 28R, 2 are located between the hip joint and the knee joint.
8L, the lower leg link 30R, 3 between the knee joint and the ankle joint
Connected at 0L. These joints are mainly composed of an electric motor and a speed reducer for boosting the output of the electric motor as described above. Details of the joints are described in the application proposed by the present applicant (Japanese Patent Application No. 1-324218). JP-A-3-184772
), And the description itself is not the gist of the present invention.

Here, the body 2 of the robot 1 shown in FIG.
4 has an image sensor 3 on a plate 34 having rigidity.
6 are fixed. The image sensor 36 is composed of a CCD (Solid State Imaging Device) camera having a known light receiving unit and a storage unit, and its output is sent to an image processing unit 38 composed of a microcomputer. Image processing unit 38
Has an image memory composed of 512 × 512 pixels as shown in FIG. The image sensor 36 is firmly fixed so as not to move on the plate 34 and
Is mounted on a shaft 40, and is configured to be able to pivot the shaft 40 via a mechanism (not shown) in response to a command from the control unit 26 to swing the image sensor 36 in the y-axis direction in FIG. . The robot 1 shown in FIG. 1 has a six-axis force sensor 4 for detecting a force component in x, y, and z directions transmitted to the robot via a foot and a moment component around the direction.
4. The control unit 26 includes an encoder 46 for detecting the angle of each joint and a tilt angle sensor (not shown) for detecting the tilt angle (speed) of the body 24 with respect to the road surface.
Receives these outputs, obtains information on the posture of the robot, obtains visual information from the image sensor 22, calculates the target joint angle (speed) degree in real time, and controls each joint to a target value to walk.

The gist of the present invention is to recognize protrusions such as stairs in a mobile robot that walks autonomously by obtaining information on the inside and outside worlds, and the following description focuses on this point. In the present invention, the stairs are as shown in FIG. The stairs 50 are composed of three steps, the height of which is set to be considerably lower than the position of the image sensor 36 when the robot 1 stands upright, so that the robot 1 can look down when approaching the vicinity thereof, that is, each step. The upper surface 50a of the stage is the lower stage, and the image sensor 36 of the robot 1
Is set so that it is not hidden from the field of view, and its upper surface 50a is painted white to give a sufficiently large brightness to the background. The edge of the upper surface 50a is bordered with a narrow black tape 52. Note that the wall surface 50b is left as it is.

FIG. 4 is an overall PAD showing the staircase recognition method.
FIG. 3 is a (Problem Analysis Diagram) diagram (structured flow chart) showing the operation of the image processing unit 38 described above. In the figure, first, a white area is extracted in S1. FIG. 5 is a subchart thereof.

Referring to FIG. 5, first, in step S10, a binarization process is performed. That is, a threshold value is determined for the original image as shown in FIG. 6 so that the upper surface 50a of each step of the stairs 50 is separated from the background, and binarization is performed to obtain an image as shown in FIG. . The determination of the threshold value is performed by an appropriate method. Next, in S12, the longest white line portion and its start point and end point are obtained for each scan line in the image memory shown in FIG. When the obtained white line portion is displayed, a quadrilateral as shown in FIG. 8 is obtained. Subsequently, it is checked whether or not the white line portions obtained in S14 are connected, the first line and the last line of the connected portion are stored, and the area thereof is obtained. Then S16
Then, the obtained white portion is sorted and rearranged in the order of the area. This is because when viewing the stairs 50 as shown in FIG. 3, the area of the lower step should be larger than the area of the upper step.
This is to prepare for the correspondence in triangulation described later using this. Subsequently, in S18, the portions smaller than the predetermined area are deleted, and the noises A and B (shown in FIG. 8) are deleted. Subsequently, the upper and lower ends of the remaining white portion are obtained in S20.

Returning to the PAD diagram of FIG. 4, the type of quadrilateral is determined in S2, and the end point is extracted in S3. still,
Here, the “end point” is used to mean the position of the corner of the quadrilateral.

FIG. 9 is a sub-chart of FIG.
At 00, the lost portion is searched. That is, the image sensor 36
Is mounted on a legged robot, and depending on the posture of the robot, the image may be tilted as shown in FIG. 10. In this case, as shown in FIG. 11, the steps may interfere with each other. In the upper row, which is in contact with the white line of the beginning row, check whether there is any white area.
The part is added to the lower white area as shown in FIG. Further, it is checked whether there is a white line portion connected to the lower white line portion in the line above it, and if it exists, it is added to the white region, if not, the search is terminated and the line is changed to the line at the beginning of the new white region. And
Also, regarding the end portion, it is checked whether there is a white portion in the lower line portion which is in contact with the white line portion of the end line of the white region, and if there is, the portion is added to the upper white region. Further, it checks whether there is a white line part connected to the upper white line part in the line below it, and if it exists, continues the search while adding it to the white area, if not, terminates the search and replaces the line with a new white area. End line.

Then, the program proceeds to S102, in which the starting point of the white line
Find the point where the maximum / minimum value of x is obtained from the end point,
The distance between the points obtained in S104 is calculated, and the mutual relationship of each point is obtained from the distance in S106. That is, for the two-dimensionally continuous white area registered in the work up to that point,
The following values are calculated using the lines at the beginning and end of the area and the start and end of the white line of each line on which the white area is placed. Note that, in the screen of FIG. 2, a straight line having a + in the right direction on the uppermost line is defined as an x-axis, and a straight line extending downward from the upper left end of the screen is defined as a y-axis. When the x and y axes are determined in this manner, the x, y, x + y, x− with respect to the start and end of each white line included in one white region, and the x, y coordinates of the start and end lines of the white line portion Find the coordinates that give the maximum and minimum of y. The positions and coordinates at which the maximum and minimum values of x, y, x + y, and x−y determined here are set as follows. xy coordinates of the maximum value of x: max x = (max xx , max xy) xy coordinate of the minimum value of x: mi n x = (min xx, min xy) xy coordinate of the maximum value of y: max y = (max yx, max yy) The xy coordinates of the minimum value of y: min y = (min yx, min yy)) The xy coordinates of the maximum value of x + y: max x + y = (max xpyx, max xpyy) The xy coordinates of the minimum value of x + y : Min x + y = (min xpyx, min xpyy) xy coordinates of the maximum value of xy: max x −y = (max xmyx, max xmyy) xy coordinates of the minimum value of xy: min x −y = ( min xmyx, min xmyy) The above is shown in FIG.

The following eight feature quantities are further calculated from the eight feature points obtained above. Top Right = (min yx-max xmyx) ² + (min yy-max xmyy) ² Top Left = (min yx-min xpyx) ² + (min yy-min xpyy) ² Bottom Right = (max yx-max xpyx) ² + (max yy-max xmyy) ² Bottom Left = (max yx-min xmyx) ² + (max yy-min xmyy) ² Max Up = (max xx-max xmyx) ² + (max xy-max xmyy) ² Max Down = (max xx-max xpyx) ² + (max xy-max xpyy) ² Min Up = (min xx-min xpyx) ² + (min xy-min xpyy) ² Min Down = (min xx-min xmyx) ² + (min xy − min xmyy) ^{2 In the} above, max, min, up, and down are top,
bottom, left, right are defined with respect to the y axis,
The eight features defined above are x, y,
It indicates the distance between the position coordinates giving the maximum and minimum values of x + y and x−y. These are (1,0), (0,1),
It is a quadrilateral contact point obtained from the straight line group of (1, 1) and (1, -1) and the upper surface 50a of each step of the stairs. Although the distance between the contacts takes various values, if the value is close to 0, that is, if the two contacts are very close or the same, FIG.
As shown in FIG. 3, the directions of two sides at the coincident contact point fall within a certain angle. When the aspect ratio of the image is 1: 1, the angle between the two sides is less than 135 degrees.
In FIG. 14 to FIG. 23, the arc drawn on the side where the two straight lines intersect with each other at the point where the quadrilateral touches each other is two points when each feature amount, for example, Top Right takes a value of 0 in FIG. The range in which the side exists is shown. From this relationship, FIG.
23, it is possible to find a case where the angle between the two sides is narrower. For example, To
In the case of FIG. 22 where p Right = 0 and Top Left = 0, the range where the two sides exist is within 90 degrees below the contact point. As will be described later, when searching for a narrower angle, not only the eight feature values but also a new combination may be selected from the position coordinates giving the maximum and minimum values of x, y, x + y, and x−y.

Subsequently, in S108, it is determined whether the quadrilateral is a trapezoid or a rhombus from the mutual relation of the points. More specifically, when the robot 1 shown in FIG. 1 looks down on the stairs 50, the obtained quadrilateral can be classified according to the following rules, and the discrimination is made accordingly. In this operation, the shape of the obtained quadrilateral is specified in order to extract an end point, which will be described later, and the operation is facilitated. (A) Trapezoid (FIG. 24) Min Down <δ and Max Down <δ and (Bottom Left <
δ or Bottom Right <δ) or | min xpyy-ma
If x xmyy | <ε and | min xmyy−max xpyy | <ε, it is trapezoidal. Here, δ and ε are positive relatively small integers. (B) Lower left upper right acute quadrangle (Fig. 25) The condition of (a) above was not satisfied, and Max Up <δ and Top Ri
If ght <δ and Min Down <δ and Bottom Left <δ, the lower left and upper right acute quadrilateral. (C) Upper left and lower right acute angles (FIG. 26) Conditions (a) and (b) are not satisfied, Min Up <δ and Top
If Left <δ and BottomRight <δ and Max Down <δ, an upper left and lower right acute quadrilateral. (D) Right acute angle (FIG. 27) or left acute angle (FIG. 28) The conditions of (a), (b) and (c) are not satisfied, and Min Up <δ and Min Down <δ, or Max Up <δ and Max Down < δ
If so, a right-angled square or a left-angled quadrilateral. (E) Others When the conditions (a), (b), (c), and (d) are not satisfied, a right-inclined quadrilateral or a left-inclined quadrilateral.

Subsequently, in step S110, the end points of the quadrilateral are extracted. Here, assuming that coordinates are determined with respect to the positions of the end points as shown in FIG. 29, the positions of the respective end points can be determined as follows for the determined type of the quadrilateral. (A) In the case of a trapezoid (FIG. 24) (trx, try) = (max xmyx, max xmyy) (tlx, tly) = (min xpyx, min xpyy) (brx, bry) = (max xpyx, max xpyy) ( blx, bly) = (min xmyx, min xmyy) (b) In case of lower left upper right acute angle (Fig. 25) (trx, try) = (max xmyx, max xmyy) (tlx, tly) = (min xpyx, min xpyy) (brx, bry) = (max xpyx, max xpyy) (blx, bly) = (min xmyx, min xmyy) (c) In the case of the upper left and lower right acute angles (FIG. 26) (trx, try) = (max xmyx, max xmyy) (tlx, tly) = (min xpyx, min xpyy) (brx, bry) = (max xpyx, max xpyy) (blx, bly) = (min xmyx, min xmyy) (d) Right acute angle (Fig. 27) Or, in the case of an acute left angle (FIG. 28) First, a check is made as to whether it is rising to the right or left. That is, when the line number of the start of a certain white area is St, the start of the white line of line i is Min [i], and the end is Max [i], Max [St + α] −Max [St] <Min [St] − Min [St +
α] rises to the right, otherwise rises to the left
(Note that α is a positive integer appropriately set). (Trx, try) = (min yx, min yy) (tlx, tly) = (min xx, min xy) (brx, bry) = (max xx, max yy) (blx, bly) = ( (max yx, max yy) If it rises to the left, (trx, try) = (max xx, max xy) (tlx, tly) = (min yx, min yy) (brx, bry) = (max yx, max yy) ( blx, bly) = (min xx, min xy) (e) Others (right inclination (FIG. 30), left inclination (FIG. 31)) After performing the same check, if it rises to the right, (trx, try) = ( min yx, min yy) (tlx, tly) = (min xx, min xy) (brx, bry) = (max xx, max xy) (blx, bly) = (max yx, max yy) Down <δ or Bottom Left <δ or Top Right <δ and Bottom Left <δ and Ma
If x Down <δ, then (tlx, tly) = (min xpyx, min xpyy). If the length of the white line in the starting line is longer than a certain length (for example, 10 dots), (trx, try) = (max xmyx, max xmyy) (tlx, tly) = (min xpyx, min xpyy) If the length of the white line in the last line is longer than a certain length (for example, 10 dots), (brx, bry) = (max xpyx, max xpyy) (blx, bly) = (min xmyx, min xmyy ). When falling to the left (trx, try) = (max xx, max xy) (tlx, tly) = (min yx, min yy) (brx, bry) = (max yx, max yy) (blx, bly) = ( min xx, min xy), but Max Down <δ or Bottom Right <δ or Top Left <δ and Bottom Right <δ and Min
If Down <δ, then (trx, try) = (max xmyx, max xmyy). If the length of the white line of the starting line is longer than a certain length (for example, 10 dots), (trx, try) = (max xmyx, max xmyy) (tlx, tly) = (min xpyx, min xpyy) If the length of the white line in the last line is longer than a certain length (for example, 10 dots), (brx, bry) = (max xpyx, max xpyy) (blx, bly) = (min xmyx, min xmyy ).

Referring again to the overall PAD diagram of FIG. 4 , subsequently, the parallelism of each stage is checked for the quadrilateral obtained in S4. That is, underlines (overlines) obtained by continuously max x (min x) of the figures in each row are compared to determine whether or not they are parallel to each other by a predetermined value or more. Since each stage is inherently parallel, non-parallel stages are deleted from the data. Thus, it is possible to determine whether or not the graphic obtained by mixing the noise is inappropriate.

The above processing is performed on each of the figures output by the two image sensors 36 to obtain the left and right figures as shown in FIG. 32, and then the corresponding left and right quadrilaterals are determined in S5. That is, first, the barycentric positions of the previously obtained end points for all the right and left quadrangles are obtained, and sorted (rearranged) by the Y coordinates of the barycenters. Next, the left and right quadrangles are associated with each other in the sorted order. At this time, the selection is made such that the Y coordinate of the center of gravity of the quadrilateral considered to correspond to approaches a predetermined value or less. If the corresponding quadrilateral does not exist, it is regarded that the quadrilateral was not obtained due to noise or the like, and the quadrilateral is not used in the stereoscopic triangulation described later (the previously detected graphic may be used. ).

Then, the process proceeds to S6, in which the Y coordinates of the four end points of the left and right quadrangles determined to correspond to each other are compared.
It is determined whether the distance is within a range less than a predetermined value, and triangulation is performed on the end points of the positive set. It should be noted that those whose distance is longer than a predetermined value are not regarded as corresponding end points for preventing erroneous measurement, and are not used in surveying. Further, as shown in FIG. 32 for each endpoint, keep numbered from left front counterclockwise, to facilitate correspondence. FIG. 33 shows the triangulation method. This method is known in the art, and the x and y coordinates of the target point (end point) L are obtained using similarity of a triangle. Assuming that the inter-camera distance 2d and the focal length f are known, the parallax X
1 and Xr are obtained, and then the target coordinates can be obtained by (x, y) = (d (X1 + Xr) / (X1-Xr), 2fd / (X1-Xr)). Note that this equation is limited to the case where the optical axes of the left and right lenses are parallel. Usually, appropriate correction is required due to convergence. However, since triangulation itself is not the gist of the present invention, a detailed description will be given. Description is omitted.

Then, the process proceeds to step S7, where the obtained position coordinates are converted into desired position coordinates, and the processing ends.

Since this embodiment is constructed as described above, it is possible to quickly recognize a stair and obtain its position coordinates. At that time, the corresponding points of the stereoscopic left and right images are obtained by using the characteristics of the stairs, so that the method can easily and quickly obtain the position information of the stairs while using a method that does not require dedicated hardware. be able to. The robot can calculate accurate three-dimensional position information of the stairs at any time as long as the robot can view the stairs, and does not need accurate position information such as map information created in advance. Due to this property, it is possible to find the stairs and go up and down only by roughly instructing the robot on the movement path including the stairs in the factory or office. When the moving environment of the robot is limited to factories, offices and the like, it is not troublesome to make the upper surface of the stairs brighter than the background. Note that lighting may be used instead of painting.

In the above embodiment, the Top Right
Although eight feature values are used, it is also possible to identify a shape of a pentagon or more by further increasing the feature value. In that sense, the staircase is taken as an example, but is not limited to this, and can be applied to all polyhedrons.

As for the parallelism check described in S4 of FIG. 4, a surface including the end point of each quadrilateral calculated by triangulation is considered, and if the surface is not parallel to the floor surface, it is deleted from the registration. May be. If there are a plurality of parallel quadrilaterals, those quadrilaterals are regarded as steps of the stairs, and the direction of the plane formed by each quadrilateral is determined without mounting the tilt sensor on the robot 1. The inclination of the robot 1 may be calculated.

Furthermore, as an embodiment, a bipedal legged mobile robot has been described as an example.
The present invention is applicable to a crawler type robot and the like, and can be applied to an unmanned traveling vehicle which is not a robot.

[0027]

Effects of the Invention Claim 1 term, multifaceted with one or a plurality of the visual sensor, substantially horizontal plane definitive the moving body moving performs movement control by the information obtained therefrom in Reference
Oite the recognition method of the body, by and against the image obtained from the visual sensor performs a predetermined binarization, the multi
Pixels exhibiting the substantially horizontal surface of the tetrahedron, to obtain a binary image of the outside world, such as with different output value from the pixel of the approximately the partition boundary region and background horizontal plane, about the substantially horizontal surface <br / > Te, and measuring the area in the obtained binary image, and so constructed as consisting of recognizing the polyhedron based on the area of the calculated <br/> meth substantially horizontal plane, only a simple structure A polyhedron can be recognized without requiring hardware or the like.

[0028] The second aspect wherein, polyhedron with one or a plurality of the visual sensor, substantially horizontal plane definitive the movable body that moves performs movement control information obtained therefrom with reference
Of Oite recognition method, by which to pair the images obtained from the visual sensor performs a predetermined binarization, the polygon
Pixels exhibiting the substantially horizontal plane of the body, to have an output value different from the pixel of the substantially horizontal surface segment boundary area and the background
Such to obtain a binary image of the outside world, scanned from the x and y directions in the coordinate axis for the binary image in which the obtained, wherein
The positional characteristics of the substantially horizontal plane in the obtained binary image are shown.
A plurality of feature points, and based on the obtained feature points,
The shape of the substantially horizontal plane in the obtained binary image is obtained,
From the characteristic points and the shape of the substantially horizontal plane obtained above,
Obtains a plurality of predetermined sites, and since it is configured as consists of calculating the three-dimensional position of the polyhedron from a predetermined region determined above, without the need for such special hardware a simple structure, polyhedron if only placed in a field of view, multi
The three-dimensional position coordinates of the plane can be calculated at any time, and accurate position information such as map information created in advance is not required. Thus, it is possible to find and move up and down the polyhedron ( stairs ) only by roughly instructing the moving body including the polyhedron such as the stairs to the mobile body in the factory or office.

[0029] The third aspect, wherein, since the generally determines the corresponding points for each horizontal plane and as configured to calculate the three-dimensional position, can be simplified further configuration.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing an entirety of a bipedal walking robot which is a moving object scheduled by a method of recognizing stairs of a moving object according to the present invention and a control device thereof.

FIG. 2 is an explanatory diagram of an image memory of the image sensor shown in FIG.

FIG. 3 is an explanatory perspective view showing a staircase planned by the method according to the present invention.

FIG. 4 is a flowchart (overall PAD diagram) showing a method according to the present invention performed by the image processing unit shown in FIG. 1;

5 is a sub-PAD diagram showing an extraction operation of a white area in the sub-chart of the PAD diagram of FIG. 4;

FIG. 6 is an explanatory diagram showing an original image as a part of the explanatory diagram showing an extraction operation of a white area in the PAD diagram of FIG. 5;

FIG. 7 is an explanatory view showing the binarized image in the same manner.

FIG. 8 is an explanatory view showing the operation of extracting the longest white line for each row in the same manner.

FIG. 9 is a sub-PAD diagram showing a quadrilateral type discrimination and end point extraction work in the sub-chart of the PAD diagram of FIG. 4;

FIG. 10 is an explanatory diagram illustrating a part of an explanatory diagram illustrating a search operation for a lost portion in the type determination of a quadrilateral in the PAD diagram of FIG.

FIG. 11 is an explanatory view similarly showing interference of each stage.

FIG. 12 is an explanatory view showing an operation of adding a missing part in the same manner.

FIG. 13 is an explanatory diagram showing a search operation of four end points in the type determination work of the PAD diagram of FIG. 9;

FIG. 14 is an explanatory diagram showing a case where Top Right of feature amounts used in the type determination work is zero.

FIG. 15 is an explanatory diagram showing a case where Top Left is zero similarly.

FIG. 16 is also an explanatory diagram showing a case where Bottom Right is zero.

FIG. 17 is also an explanatory diagram showing a case where Bottom Left is zero.

FIG. 18 is an explanatory view showing a case where Max Up is zero similarly.

FIG. 19 is an explanatory view showing a case where Max Down is zero as well.

FIG. 20 is an explanatory diagram showing a case where Min Up is zero as well.

FIG. 21 is an explanatory diagram showing a case where Min Down is zero in the same manner.

FIG. 22 is an explanatory diagram showing a case where Top Right and Top Left are both zero.

FIG. 23 is an explanatory diagram showing a case where Top Left and Min Up are both zero.

FIG. 24 is an explanatory view similarly showing a trapezoid.

FIG. 25 is an explanatory view showing a lower left upper right acute angle in the same manner.

FIG. 26 is an explanatory view showing an upper left and lower right acute angle in the same manner.

FIG. 27 is an explanatory view similarly showing a right acute angle.

FIG. 28 is an explanatory view similarly showing a left acute angle.

FIG. 29 is an explanatory diagram showing position coordinates of end points used in the end point extraction work of the PAD diagram of FIG. 9;

30 is an explanatory diagram showing a case where the feature amount used in the type determination work in the PAD diagram of FIG. 9 is inclined rightward, as in FIG. 14;

FIG. 31 is an explanatory view showing a case of similarly tilting to the left.

FIG. 32 is an explanatory diagram showing a work of associating quadrangles in the PAD diagram of FIG. 4;

FIG. 33 is an explanatory diagram showing triangulation in the PAD diagram of FIG. 4;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 legged mobile robot (moving body) 24 body part 26 control unit 34 plate 36 image sensor 38 image processing unit 50 stairs 50a Top surface of stairs

Continuation of the front page (51) Int.Cl. ⁷ Identification symbol FI G06T 7/00 G05D 1/02 K // G05D 1/02 G06F 15/62 415 15/70 460E

Claims (3)

(57) [Claims] [Claim 1 further comprising one or more of the visual sensor, the recognition method <br/> Oite polyhedron having a substantially horizontal surface definitive the movable body that moves performs movement control information obtained therefrom with reference, a. 2 given by pairs images obtained from the visual sensor
By performing binarization processing, it indicates the substantially horizontal surface of the polyhedron
To pixel, to obtain a binary image of the outside world, such as with different output value from the pixel of the approximately the partition boundary region and background horizontal surface, b. Determining the area in the obtained binary image for the substantially horizontal plane ; and c. Recognition method polyhedron mover including the recognizing, said polyhedron based on the area of the obtained substantially horizontal. 2. A has one or more of the visual sensor, the recognition method <br/> Oite polyhedron having a substantially horizontal surface definitive the movable body that moves performs movement control information obtained therefrom with reference, a. 2 given by pairs images obtained from the visual sensor
By performing binarization processing, it indicates the substantially horizontal surface of the polyhedron
To pixel, to obtain a binary image of the outside world, such as with different output value from the pixel of the approximately the partition boundary region and background horizontal surface, b. In the coordinate axis for the binary image in which the resulting scanned from the x and y directions, the substantially before Symbol resulting binary image
Obtaining a plurality of feature points indicating the positional features of the horizontal plane ; c. Based on the obtained feature points, the obtained binary image
Determining the shape of said substantially horizontal plane , d. From the obtained characteristic points and shape of the substantially horizontal plane,
Multiple pieces determine the predetermined portion of the plane, and e. And calculates the three-dimensional position of the polyhedron from a predetermined region determined above, recognizing method polyhedral mobile consisting. 3. The method for recognizing a polyhedron of a moving object according to claim 2, wherein said three-dimensional position is calculated by determining a corresponding point for each of said substantially horizontal planes .