JPH10222680A - Method for recognizing position - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly attain position confirmation without depending on the variation or attitude of an objective shape by calculating the distance of a line segment connecting two points at a constant distance position along an outline, and detecting a place where two points whose line segment distance is less than the threshold value of the constant distance are present on the outline as an object angle changing part. SOLUTION: An image pickup means 1 image picks-up an object, and outputs a video signal, and a concentration storing means 2 stores the video signal as concentration picture data. A binarizing means 3 converts the concentration picture data into black and white binary picture data by a constant threshold value, and an outline detecting means 4 scans the binary picture data, and detects the outline part of the object. A two point straight distance calculating means 5 calculates the distance of a segment connecting two points with a constant distance along the outline, and a candidate point calculating means 6 detects a place where two points whose line segment length is less than a constant distance are present on the outline as the candidate point of an angle changing part.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a position recognition method for a visual recognition device for a robot, and more particularly to a position recognition method for detecting a position of a recognition target from an image captured by an image capturing means.

[0002]

2. Description of the Related Art In recent years, a position recognition device for detecting the position of an object from an image picked up by an image pickup device has been widely used as a visual recognition device for a robot in various production facilities.

A conventional example of a position recognizing method for recognizing the position of an object by the position recognizing device will be described with reference to FIGS.

FIG. 11 shows a basic configuration of a conventional position recognition device. The basic configuration of the conventional example is a
The grayscale image storage means 44, the grayscale image binarization means 45, the binarized image labeling processing means 46, and the center of gravity detection means 47 for detecting the center of gravity of the labeled object
And edge detecting means 48 for detecting the edge of the object in an arc shape from the position of the center of gravity, and the coordinates obtained by the edge detecting means 48 are converted into the coordinate system of the robot and used as the gripping position of the robot.

[0005]

However, in the above conventional example, after the position of the center of gravity 50 is detected as shown in FIG.
When there are a plurality of objects on the arc 51 for edge detection (for example, the edge 52 and the edge 53), the robot should grab the edge 53 when the robot should originally grab the edge 53, Since the center position of the arc 51 is determined based on the position of the center of gravity of the object, there has been a problem that stable position detection cannot be performed due to the influence of the posture and noise of the object.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems and to provide a position recognizing method capable of accurately recognizing the position of a characteristic shape of an object without depending on the variation or posture of the object shape. I have.

[0007]

In order to solve the above-mentioned problems, a position recognition method according to the present invention has two substantially linear portions each having a fixed length on a contour and a constant interval from each other. And between the two straight portions, approximately 1
An object including an angle change that causes an angle change of 80 ° is imaged by an imaging unit to obtain a grayscale image, the grayscale image is binarized, a contour of the object is detected, and a constant distance is set along the contour. The distance of a line segment connecting two points at a position is sequentially calculated, and the location where two points on the contour where the distance of the line segment is equal to or smaller than a threshold of a certain distance exists as the object angle changing unit. It is characterized by detecting.

According to the position recognition method of the present invention, when determining a candidate for an angle changing portion, if a line segment connecting two points on the contour is inside the contour of the object, it is determined as a candidate for the angle changing portion. Is preferred.

According to the position recognition method of the present invention, when a plurality of candidates for the angle change portion are detected, the distance between a plurality of line segments connecting two points on the contour before and after each of the positions is constant. It is preferable to determine the angle changing part from the candidates by confirming that there is.

Further, in the position recognition method of the present invention, when a plurality of candidates for the angle changing portion are detected, the threshold value of the distance of a line segment for detecting each of the candidate portions is detected. Based on another threshold value, the angle change portion may be determined based on the magnitude of the central angle formed by each of the other portions whose line segments are equal to or less than a predetermined distance and each of the portions with the center of gravity of the object as a center. It is suitable.

Further, in the position recognition method according to the present invention, when a plurality of candidates for the angle changing portion are detected, a distance between two straight lines at the plurality of positions is calculated, and the plurality of distances along the straight line are calculated. It is preferable to determine the angle change portion from the plurality of candidates by confirming the continuity of the interval having a constant length continuously at the position.

Further, according to the position recognition method of the present invention, when calculating the continuity of the interval between two straight lines, two straight lines are calculated from an approximate straight line obtained by connecting the position of the object and the center of the line segment. It is preferable to calculate the inclination angle of the second straight line, and calculate the length of a second line segment that is orthogonal to the approximate straight line and intersects the two straight lines.

The position recognizing method of the present invention is capable of accurately recognizing the position of an object without depending on the variation in the shape or posture of the object, and has the following operation.

According to the position recognition method of the present invention, an outline of an object is detected for an object which is composed of two straight lines, has a steep angle change, and has a constant width and length.
When calculating the distance of a line segment connecting two points located at a certain distance along the contour, the distance of the line segment in the contoured portion with irregularities becomes shorter, and the distance of the line segment in the linear contour portion becomes longer. A sharp angle change part can be detected as a peak of a distance waveform.

In addition, the position recognition method of the present invention can be used to detect a steep angle change portion only by detecting a peak of a distance waveform to determine whether a concave portion of an object is a convex portion. Is detected, and if this line segment is within the object, it can be recognized as a convex portion.

The position recognizing method of the present invention focuses on the fact that the shape of the object is composed of two straight lines, and since the distance of the line segment before and after the angle changing portion is constant, the position of the angle changing portion is determined. Can be detected.

Further, in the position recognition method of the present invention, when a plurality of candidate points are detected as an angle changing part, the distance waveform is centered on the center of gravity of the contour point and other candidate points having a distance waveform peak separately from the candidate points. The angle change part can be determined by the angle.

Further, the position recognition method of the present invention focuses on the fact that the shape of the object has a linear portion having a constant width and length.
The angle changing unit can be determined by calculating the interval between the straight lines of the book and determining whether the interval has a constant length continuously.

Further, in the position recognition method of the present invention, when calculating the width of the straight line portion of the object, the inclination angle of the straight line portion is detected from an approximate straight line consisting of the position of the object and the midpoint of the line segment, and the approximate angle is calculated. By calculating the length of a line segment that is orthogonal to the straight line and intersects the two straight lines, it is possible to detect the continuity of the width without receiving an error due to the influence of noise, and a position that is strong against variations in the shape of the target object Recognition can be performed.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described with reference to FIGS. In the present embodiment, a chrysanthemum cutting having a variation in shape is set as an object, and a main stem position consisting of a straight line portion in the cutting is detected. It is assumed that the cutting which is the object has not a plurality of shapes to be detected but only a part thereof.

FIG. 1 is a block diagram showing a main part of the position recognition apparatus according to this embodiment. In FIG. 1, an imaging means 1 captures an image of an object and outputs a video signal, and a density storage means 2 stores the video signal as density image data and stores the image signal.
The grayscale image data is converted into black and white binary image data by a constant threshold value by a value converting means 3, and the outline image detecting means 4 scans the binary image data to detect a contour portion of an object. . The two-point straight-line distance calculating means 5 calculates the distance of a line segment connecting two points located at a certain distance along the contour, and the candidate point calculating means 6 determines two points whose line segment length is equal to or less than the certain distance on the contour. Is detected as a candidate point of the angle changing unit. In other words, the main stems of the cuttings are almost parallel 2
Since they form a right angle, and these change a sharp angle of about 180 ° along the contour, this is detected as an angle change portion. The straight-line distance calculating means 7 for calculating a straight-line distance along the contour before and after the location detected as the angle change portion determines whether a straight line of a certain length exists, and determines the angle at which a straight line of a certain length exists. When there is only one changing part, the position is set as a candidate point of the main stem position of chrysanthemum. The candidate point angle calculating means 8 calculates the inclination angle of the main stem from the middle point sequence of a straight line connecting two points along the contour existing before and after the candidate point of the angle changing portion, and the candidate point width calculating means 9 calculates the angle of the main stem. The distance connecting the two points on the contour existing before and after the changing portion is defined as the width of the main stem, and the predetermined main width is determined as the final main stem position depending on the existence of a predetermined length or more. The Y coordinate and the inclination angle θ of the main stem of the object obtained by the candidate point angle calculation means 8 are detected.

FIG. 2 is a view showing a part of the contour points of the object by using chain codes in the vicinity of 8, where the distance in the oblique direction 11 is 1. When calculated as 4, the point at a fixed distance interval 6 from A becomes B, and the calculated distance of the chain code label A is A
-B is the straight line distance.

FIG. 3 is a graph showing a relationship between a chain code label number, which is a contour point of an object, and a linear distance between two points which are at a fixed distance along the contour. A point C on the graph lower than the fixed distance threshold value 14 is defined as an angle changing part.

FIG. 4 shows a part of the graph 13 in FIG. 3 in which the fixed front and rear labels from the angle changing unit 15 are not calculated as the dead zone 16, and the distances between the labels in the sections indicated by front and rear 17a and 17b are used. Determine the presence of linearity.

FIG. 5 shows an example of the chrysanthemum cutting outer shape D used in the present embodiment. The portion 18 detected as an angle changing portion corresponds to the convex portion of the object, and 18a and 18b are chain codes before and after the portion 18. In the position of the label number 19 is position 18a,
This is a straight line connecting the positions 18b. 20 corresponds to a concave portion, 2
0a and 20b are the positions of the chain code label numbers before and after the concave portion 20, and 21 is a straight line connecting the positions 20a and 20b.

FIG. 6 is a graph showing the relationship between the chain code label number of the object contour used in the present embodiment and the linear distance between two points on the contour.
If 2, 23 is equal to or more than a certain threshold value, it is detected as the petiole position of the chrysanthemum cutting ear.

FIG. 7 shows the angular relationship between the main stem candidates 26 and 27, which are the angle change points of the outer shape of the object used in the present embodiment, and the petiole positions 28, 29 and 30. Main stem candidate 26 and center of gravity 2 centered on center of gravity 25 of rectangle 24
5 and the line segment connecting the center of gravity 25 with the front and rear petiole of the main stem candidate 26 or the main stem candidate points 30, 28 and the center of gravity 25.
1, and θ2 are calculated. Similarly, main stem candidate 2
7 between the front and rear petiole or main stem candidate points 28 and 29
3, and θ4 are calculated.

FIG. 8 shows the main stem portion of the object used in the present embodiment.
The middle point 32c of the straight line when each of 34b is connected by a straight line
The inclination 35 of the approximate straight line obtained by connecting 34c is the inclination of the main stem, and the position 36 of the chain code label number closest to the approximate straight line 35 is the detection point of the main stem.

FIG. 9 shows a main stem portion of an object used in the present embodiment, and positions 37a to 39a and 37 of chain code label numbers at predetermined intervals in front and rear from a detection point 36 of the main stem.
The distances 37 to 39 obtained by connecting b to 39b are defined as the width of the main stem portion, and it is determined whether or not the width within a predetermined range is equal to or more than a predetermined number.

FIG. 10 shows the main stem portion of the object used in the present embodiment.
The approximate straight line 41 is orthogonal to the approximate straight line 41 at a point 42 at a predetermined distance from the main stem position 40 in the approximate straight line 41 of the main stem, and the length of a line segment intersecting the outline of the object at the points 42a and 42b is detected as the width of the main stem portion. By doing so, the width can be detected while allowing an error in the length of the line segment based on the contour points around the noise.

Next, the position recognition operation in the present embodiment will be described with reference to FIGS. First, a video signal obtained by imaging an object by the imaging means 1 shown in FIG. 1 is stored as grayscale image data by the grayscale storage means 2, and the grayscale image data can be cut out from the background by the binarization means 3. The image data is converted into binary image data of black and white with a fixed threshold value. With respect to the binary image data, when the background is black, the object is sequentially scanned clockwise in the vicinity of 8 around the position where the white image data as the object is first found, and then the white image data is found. Label the direction of the position with the chain code. Next, the two-point straight-line distance calculation means sequentially calculates the straight-line distance between two points at a constant distance on the contour. In the case of FIG. 2, the chain code in the front-rear, left-right direction is 1, and the chain code in the oblique direction is 1.4.
When the constant distance on the contour is set as 6, the point B (1 + 1 + 1.4) where the distance of the chain code from the point A becomes 6 or more.
The linear distance 12 of (+ 1.4 + 1.4 = 6.2) is calculated. When this is performed for all points on the contour of the object and graphed, the state shown in FIG. 3 is obtained. Here, the position C between two points on the contour having a predetermined threshold value 14 or less becomes an angle changing portion. Since the position at which the object is to be detected is composed of two straight lines, the straight line distance for a certain label excluding the front and rear dead zones from this angle change portion should be detected over a long distance. Therefore, sections 17a and 17b in FIG.
If the straight line distance does not exist for a certain length, it is better to judge that the main stem candidate position does not exist in order to prevent erroneous recognition.

It is assumed that sections 17a and 1a of FIG.
Even if the straight line distance of 7b is a certain length, there is a part corresponding to the concave part of the object as shown at 20 in FIG. 5, so that a line segment 21 connecting the angle changing part and the front and rear fixed label positions 20a and 20b is two. It is possible to determine a concave portion or a convex portion according to the background portion of the value image data or the portion inside the contour of the object.

If only one angle change portion as a main stem candidate is detected in the above process, the process can proceed to the process of evaluating the width of the main stem portion. Here, a plurality of main stem candidates are detected. In this case, a petiole position other than the main stem candidate is detected, and one main stem candidate is determined from the petiole positions before and after or the angular relationship with other main stem candidate positions with reference to the main stem candidate position. From the graph of FIG. 6, the width 2 from peak to peak of the linear distance waveform
If the lengths 2 and 23 have a certain length or more, they are detected as petiole portions, and at the same time, whether the concave portion of the object is a convex portion is determined by the above method. When 26 and 27 in FIG. 7 are detected as main stem candidates, the petiole positions 28, 29 and 30 are detected by the above-described method, and the angle θ is set around the center of gravity 25 of the circumscribed rectangle of the object.
1, θ2, θ3, and θ4 are calculated. Here main stem candidate 26
Θ1 + θ2 with reference to angle θ and angle θ with reference to 27
The final candidate is determined based on whether 3 + θ4 is equal to or greater than a certain threshold.
If only one of a plurality of candidate points has an angle equal to or greater than a certain threshold, it is a candidate point. If all angles are less than a certain threshold, it is better to determine that no candidate point exists. Alternatively, if a plurality of candidates are equal to or more than a certain threshold, it is possible to determine by a width evaluation of the next processing.

In the width evaluation, the midpoints 32c to 34c of the line segments connecting the positions 32a to 34a and 32b to 34b of the chain code label numbers at constant intervals before and after are connected from the angle changing portion 31 as the main stem candidate in FIG. Approximate line 35
Is the inclination of the main stem, and the position 36 of the chain code label number closest to this approximate straight line 35 is the detection point of the main stem. Lines 37 to 39 connecting the positions 37a to 39a and 37b to 39b of the chain code label numbers which are present at a certain distance from the detection point 36 of the main stem in FIG. It is determined whether the width exists over a predetermined length. In addition, when noise such as 43 in FIG. 10 is present, a straight line 41 indicating the inclination of the main stem is orthogonal to the approximate straight line 41 at a point 42 at a certain distance from the main stem position 40, and 42a, 42b
By detecting the length of the line segment that intersects as the width of the main stem, it becomes possible to detect the width by allowing the error of the line segment length based on the contour points around the noise, and to determine the angle as the main stem candidate. By judging whether or not the changing portion has a predetermined width and a fixed length, it is possible to accurately detect the main stem position and the inclination angle of the main stem.

[0035]

The position recognition method according to the present invention includes two linear portions, and includes an angle changing portion which causes an angle change of approximately 180 ° along the contour between the two linear portions. Upon detecting the position of the object, the contour points of the object are detected, and the distance of a line connecting two points located at a certain distance along the contour is sequentially calculated, and the distance of the line becomes less than a certain distance. When the position where two points on the contour are present is detected as the position of the target, and the width of the two straight lines and the continuity of the width are observed from the detected angle change portion, so that there is a variation in the target. This provides an effect that the position of the object can be accurately recognized even when the posture is not constant.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a position recognition method according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a chain code that is a contour point of an object and a straight-line distance between two points at a fixed distance along the contour according to the present embodiment.

FIG. 3 is a graph showing a relationship between a chain code label number, which is a contour point of an object, and a linear distance between two points at a fixed distance along the contour.

FIG. 4 is a partially detailed view of the graph of FIG. 3;

FIG. 5 is a diagram showing an object and a main stem position detected as an angle changing unit.

FIG. 6 is a diagram showing a method for detecting a petiole other than the main stem candidate for the object.

FIG. 7 is a diagram showing an angular relationship with another petiole when a plurality of main stem candidates are detected for an object.

FIG. 8 is a diagram showing a main stem portion of an object and its inclination.

FIG. 9 is a diagram showing a main stem portion of an object and its width.

FIG. 10 is a diagram showing a detection width when noise is present in a main stem portion and a straight portion of an object.

FIG. 11 is a diagram showing a configuration of a conventional position recognition method.

FIG. 12 is a diagram showing the position of the center of gravity of an object and a circular arc for edge detection in a conventional position recognition method.

[Explanation of symbols]

A, B Two points located at a fixed distance from each other along the contour C Angle changing part D Object 10, 11 Chain code 12 Linear distance between constant contour points 13 Graph of chain code label number and linear distance 14 Constant distance threshold Value 16 Dead zone width 17a, 17b Average straight line distance 18,20 Main stem candidate position 18a, 18b Front and rear contour points of main stem candidate 18 20a, 20b Front and rear contour points of main stem candidate 20 22,23 Petiole calculation threshold 24 Target Object circumscribed rectangle 25 Center position of circumscribed rectangle 26,27 Main stem candidate point 28,29,30 Petiole candidate point θ1-θ4 Angle with other petiole with main stem candidate point as axis 31 Temporary main stem candidate point 32a- 34a, 32b-34b Contour points at fixed intervals from main stem candidate point 32c-34c Midpoint of straight line connecting outline points before and after main stem candidate point 35 Checked from point sequence of 32c-34c End points of the approximate straight line 36 straight 42a connecting the contour points before and after the predetermined intervals approximate line 35 from the contour points 37 to 39 principal stem candidate points closest, the width perpendicular to 42b slope 41 which is

Continuation of the front page (72) Inventor Yutaka Matsuzaki 1006 Kadoma Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd.

Claims (6)

[Claims] 1. The contour includes two substantially straight portions having a fixed length and a fixed distance from each other, and approximately 180 along the contour between the two straight portions. Imaging an object including an angle changing unit that causes an angle change of ° by an imaging unit, binarizing the captured image and detecting an outline of the object,
A distance of a line segment connecting two points located at a fixed distance position along the contour is sequentially calculated, and a position where two points on the contour where the distance of the line segment is equal to or less than a threshold of the fixed distance is determined. A position recognition method, wherein the position is detected as an angle change part of the object. 2. The position recognition method according to claim 1, wherein, when determining a candidate for the angle changing unit, if a line segment connecting two points on the outline is inside the outline of the object, the position is recognized as a candidate for the angle changing unit. 3. When a plurality of angle change portion candidates are detected, by confirming that the distance of a plurality of line segments connecting two points on the contour before and after each of the positions is constant, The position recognition method according to claim 1, wherein an angle changing unit is determined from the candidates. 4. A line based on another threshold value different from a threshold value of a distance of a line segment at the time of detecting each of the candidate positions when a plurality of angle change portion candidates are detected. The position recognizing method according to claim 1, wherein the angle change portion is determined by a magnitude of a central angle formed by the other portions each having a distance equal to or less than a predetermined distance and each of the portions centering on the center of gravity of the object. 5. When a plurality of candidates for an angle changing unit are detected at a plurality of locations, an interval between two straight lines at the plurality of locations is calculated, and the interval is a constant length continuously at a plurality of positions along the straight line. The position recognizing method according to claim 1, wherein the angle changing unit is determined from the plurality of candidates by confirming the continuity of the interval having the following. 6. When calculating the continuity of the interval between two straight lines, an inclination angle of the two straight lines is calculated from an approximate straight line obtained by connecting the position of the object and the midpoint of the line segment, 6. The position recognition method according to claim 1, wherein a continuity of the interval is detected by calculating a length of a second line segment that is orthogonal to the approximate straight line and intersects the two straight lines.