JP2002183740A - Detection device for precise three-dimensional position attitude of circule characteristic of component - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect three-dimensional position attitude of a circle characteristic such as 'an end point of a cylindrical part' or 'a hole' of an industrial component. SOLUTION: This detection device for the precise three-dimensional position attitude of the circle characteristic of the component limits a set state that the hole of the target component to be detected is nearly not inclined to an optical axis of a camera, sets only a portion entering a set radial range to an arc characteristic when extracting the arc characteristic, sets an edge segment from which an arc is extracted to a circle candidate point sequence, and calculates the circle characteristic on the basis of the set circle candidate point sequence. The detection device has a rough arc radius-setting means setting the arc radius range such that the detection device can deal with all range wherein the target component is possibly present at an initial state, and a precise arc radius-setting means setting the arc radius range in consideration of a respective small range above and below the rough detection result found by the rough arc radius-setting means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a precision three-dimensional position / orientation detecting apparatus for a circular feature of a part. The present invention relates to so-called object recognition processing, and particularly provides means for extracting features used in a model-based matching method.

[0002]

2. Description of the Related Art In an object recognition process using a model-based matching method, a model which is a set of geometric feature data (feature elements) such as a straight line and an arc which constitute an object,
The target object is recognized by comparing the feature elements of the target object extracted from the image. This comparison is performed by extracting straight lines, arcs, and circles constituting the target object as characteristic elements, and performing correlation (feature matching) of these characteristic elements of the target object with the characteristic elements of the model.

[0003] To extract a characteristic element, image data (edge image data) in which an edge portion is emphasized by applying sharpening processing such as a differential spatial filter to the image data is created, and an edge line composed of a series of edge pixels is created from the image data. This is performed by extracting minute data (edge chain).

However, it is actually very difficult to extract a circle feature from the above feature elements. This is because the “circle” is apparently “oval” when the object is obliquely inclined or located at a position away from the center of the image, even if it is circular when viewed from the front. Therefore, there is a method of using a closed curve as a circle feature candidate when detecting a circle feature (Japanese Patent Application No. Hei 8-100974, hereinafter referred to as “Proposal II”).
1 ").

In this method, a closed curve is classified into a "large closed curve" and a "small closed curve" according to the number of points constituting closed curve data which is a set of points having three-dimensional coordinates as data. In this case, the equivalent ellipse is calculated, and the specifications of the equivalent ellipse are extracted as feature data. In the case of a "small closed curve", a rectangle whose sides are parallel to the axis is calculated, and the center of the rectangle is calculated. The position is extracted as feature data.

However, a closed curve may be made of spots, shadows, irregularities, or the like on the surface or background of a component, and it may be difficult to distinguish a closed curve alone from a portion such as a “hole” or “end point of a cylinder”. In such a case, the number of times of matching increases, so that the processing time increases, or erroneous recognition may occur due to matching with an originally incorrect closed curve. Therefore, there is a method of detecting the three-dimensional position and orientation of the circular feature at the end point of the hole or the cylinder even when the component is tilted, and detecting the circular feature in accordance with various installation states of the component (Japanese Patent Application No. Hei 10-26139). No. 9-136452, hereafter referred to as "Proposal # 2".

This method extracts "holes" and "ends of a cylindrical portion" of an industrial product. Therefore, if a part is tilted, even if the parts do not become circular on the image, the "holes" or "cylindrical parts" are obtained. Notice that the `` end point of the part '' becomes a circle when viewed from the front, extract the point sequence part that is likely to be a circle from the edge line segments obtained from the input image, and based on the three-dimensional position data of the point sequence A circle approximation is performed on point sequence data on a plane to be calculated. Incidentally, as the arc approximation method, it is general to employ a method of approximating a circle equation using approximation means such as a least square method, calculating a center position and a radius, and determining this edge line segment as an arc. However, there is a method in which the processing for approximating an edge line segment in the image data to a circle or an arc can be performed quickly and always passes through the starting point and the ending point (Japanese Patent Application No. Hei 6 (1994) -197686).
No. 113217 or Japanese Patent Application No. Hei 6-113218, which is hereinafter referred to as “Proposal No. 3”.

According to this method, when an edge line segment in image data is approximated to a circle or a circular arc, a start point and an end point, which are two end points of a target edge line segment, and another one point, that is, a total of three points. It assumes a passing circle or arc. Then, after assuming a circle or an arc, if it is determined that all the element points of the edge line segment are on the circumference, the edge line segment is recognized as an assumed arc, and from the circumference, If it is determined that there are points separated by a certain distance or more, those edge line segments are not approximated to an arc.

Furthermore, as a three-dimensional position / posture detecting device for a circular feature of a component, when an oil or dust adheres to the surface of the circular feature on the component, the extracted edge line segment is cut small and does not become a closed curve. Also, there is a method capable of detecting a circular feature and responding to various installation states of components (Japanese Patent Application No. 11-2).
No. 32453 (hereinafter, this will be referred to as "previous proposal # 4").
This method is limited to the installation state in which the hole of the target component to be detected is not substantially inclined with respect to the optical axis of the camera, and when extracting arc features, only those within a set radius range are taken as arc features and extracted. The edge line segment that is the base of the arc is set as a circle candidate point sequence, and a circle feature is calculated based on the set circle candidate point sequence.

[0010]

However, if the arc radius range set in the prior art proposal # 4 is too large, the possibility of an erroneous recognition result increases. On the contrary, if the arc radius range set is too small, there is no possibility of complete recognition. Recognition becomes impossible, and setting of this arc radius range becomes difficult.

[0011]

According to a first aspect of the present invention, there is provided a precision three-dimensional position / orientation detecting apparatus having a circular feature of a component, wherein a hole of a target component to be detected is positioned with respect to an optical axis of a camera. When extracting arc features, only those that fall within the set radius range are taken as arc features, and the edge line segment that is the source of the extracted arc is set as a circle candidate point sequence, In a precision three-dimensional position and orientation detection device for a circular feature of a component for which a circular feature is calculated based on a set of circle candidate point sequences, an arc is formed so as to be able to correspond to an entire range in which the target component may exist in an initial state. A rough arc radius setting means for setting a radius range, and a precise arc radius setting means for setting an arc radius range in consideration of a small range up and down around a rough detection result obtained by the rough arc radius setting means. To And butterflies.

According to a second aspect of the present invention, there is provided a precision three-dimensional position / orientation detecting apparatus having a circular feature of a component, wherein a hole of a target component to be detected is substantially not inclined with respect to an optical axis of a camera. When extracting arc features, only those that fall within the set radius range are taken as arc features, and the edge line segments that are the basis of the extracted arcs are set as a circle candidate point sequence, and the set circle candidate points In the precision three-dimensional position and orientation detection device for a circular feature of a component that calculates a circular feature based on a column, in a continuous recognition / removal state, a depth of at least one inspection object is considered up and down from the previous detection result. Rough arc radius setting means for setting an arc radius range; and precision arc radius setting means for setting an arc radius range in consideration of a small range up and down around a rough detection result obtained by the rough arc radius setting means. Have It is characterized in.

According to a third aspect of the present invention, there is provided a precision three-dimensional position and orientation detecting apparatus for circular features of a part according to the present invention. In, means for measuring the three-dimensional position of the extracted point sequence, means for calculating a plane based on the position data of the point sequence, and a circle approximation for the point sequence data on the calculated plane, It is characterized by comprising means for selecting one that matches the setting data from among them, and means for detecting the three-dimensional position and orientation of the circle feature with respect to the selected circle.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (1) Basic Concept An object of the present invention is to detect a three-dimensional position and orientation of a circular feature such as a "hole" or "end point of a cylindrical portion" of an industrial part. In the present invention, the installation part is limited to an installation state in which the hole of the target part to be detected is not substantially inclined with respect to the optical axis of the camera, and the edge line segment extracted due to oil or dust attached on the part is cut small and the closed curve is closed. Even if it does not occur, an arc is extracted on the screen by approximating the edge line segment with an arc as shown in Proposal # 3, and a circle is formed based on the arc instead of the closed curve as shown in Proposal # 2. A candidate point sequence is set, and finally a circular feature is detected.

(2) Automatic Setting of Arc Radius Range In the present invention, the prior proposal No. 4
Instead of the quantitative arc radius range setting method in
The arc radius range is automatically set based on the target component height, the previous detection result, and the rough detection result. When extracting arc features,
Only those that fall within the automatically set radius range are considered arc features.

(2.1) Setting of Arc Radius Range in Initial State In the initial state, the position of the target component is indefinite, so that the radius of the arc is adjusted so as to be able to cover all ranges in which the target component may exist. Set the range. The position detected here is a rough detection result.

(2.2) Setting of Arc Radius Range in Continuous Recognition / Extraction State In order to output the recognition result having the highest position of the center of the detected circular feature, the target to be detected next is output. It is presumed that the position of the component will be less than or equal to the height position substantially equal to the previous detection result. From this, the setting of the arc radius range in the continuous recognition / removal state is performed as shown in FIG.
An arc radius range (hereinafter, referred to as a rough detection setting range) is set above and below the previous detection result, taking into account the depth A of one inspection object and α. The rough detection setting range is indicated by an allowable error between the maximum arc radius and the minimum arc radius on the image plane with respect to the lens focal point, as shown in FIG. The position detected here is a rough detection result.

(2.3) Setting of Arc Radius Range after Rough Detection Result The results detected in (2.1) and (2.2) are detection results including errors because the arc radius range is large. Therefore, as shown in FIG. 2, in order to obtain a precise detection result,
Depth B of small range β up and down around the rough detection result
Is set in consideration of the above (hereinafter, referred to as a precision detection setting range). This fine detection setting range is, as shown in FIG.
It is indicated by the tolerance between the maximum and minimum arc radii.

(3) Setting of a circle candidate point sequence In the present invention, a circle candidate point sequence using the arc feature in the prior proposal # 4 is set for the circle candidate point sequence. That is, assuming that the hole of the target component to be detected is in a state where the hole of the target component to be detected is not substantially inclined with respect to the optical axis of the camera, the image of the hole of the target component captured by the camera has a small edge line segment. However, since the arc is a part of a circle, a circle feature can be extracted regardless of whether the curve is a closed curve or not. Therefore, as the arc approximation, when approximating a point sequence in the image data to a circle or an arc, it is assumed that an arc passes through two endpoints of the target point sequence and a total of three other points. If another point in the point sequence is on the circumference of the assumed arc, it is determined that all element points in the point sequence are on the circumference of the assumed arc, and the target point sequence is assumed. Recognize as an arc.

(4) Calculation of circle feature data The three-dimensional position measurement of the point sequence, the three-dimensional circle calculation, the circle feature confirmation, and other circle feature data calculation processes are shown in FIGS.
5 or the processing after the "point sequence data extraction unit" of the apparatus shown in FIG. 26 is used. That is, means for measuring the three-dimensional position of the extracted point sequence, means for calculating a plane based on the position data of this point sequence, and circle approximation for the point sequence data on the calculated plane, It is provided with a means for selecting one that matches the setting data from among them, and a means for detecting the three-dimensional position and orientation of the circle feature with respect to the selected circle.

As an example, in the processing shown in FIG. 4, after extracting a point sequence portion likely to be a circle from the edge data obtained from the input image, three-dimensional position measurement of the extracted point sequence portion is performed.
After calculating the plane on which the three-dimensional point sequence data rides (hereinafter referred to as “circular plane”) by the least square method or the like, map all the points of the point sequence data that is a circle candidate onto the circular plane, and A circle approximation is performed on the point sequence data, and the diameter of the extracted circle is set in a predetermined range (in order to confirm whether the circle extracted from the input image is correct as an actual target part). (Hereinafter referred to as “circle set value”), and sets data indicating a circle feature to be included in the circle set value.

The processing shown in FIG. 5 is different from the processing shown in FIG. 4 in that the epipolar line is formed by a straight line formed from several points before and after the point to be measured in the three-dimensional position measurement of the extracted point sequence. A measurement point selection process for determining an angle and comparing the angle with a set value to determine whether or not to measure the three-dimensional position of the point is added. Further, the processing shown in FIG. 6 is different from the processing shown in FIG. 4 in that the vicinity of a circle is performed on the point sequence data on a plane calculated based on the three-dimensional position data of the extracted point sequence, and the extracted circle is concentric with the extracted circle. Extract the nearby circular arc above, calculate the circular plane again from the three-dimensional position data of the circle and the point sequence of the nearby circular arc, and perform circle approximation again on the point sequence data that constituted the circle on the circular plane , A neighborhood arc extraction process for detecting the three-dimensional position and orientation of a circular feature is added.

(5) Precise measurement of circle feature data When a circle candidate point sequence is set by using the method of the preceding proposal # 4, the amount of original data used for calculating the circle feature data is small. In some cases, the inclination of the circle feature differs greatly from the actual one. Therefore, by performing (2.3), measurement points are narrowed down, and precise tilt measurement of the circular feature is performed.

(6) Embodiment of the Present Invention FIG. 3 shows an example of an apparatus for precisely detecting the three-dimensional position and orientation of a circular feature according to the present invention. This device is a device for detecting the three-dimensional position and orientation of a circular feature of the embodiment (6.12) of the prior proposal # 4. Also, two types, one for rough detection and one for precision detection, are provided.

That is, as shown in FIG. 3, a point sequence data extraction unit 10 and a circle feature data calculation unit 20 for rough detection, and a point sequence data extraction unit 30 and a circle feature data calculation unit 40 for precision detection are combined. It is provided. The rough detection point sequence data extraction unit 10 includes an arc extraction unit 11 and an extracted arc virtual circle candidate point sequence setting unit 12. Arc extraction unit 11 for rough detection
When extracting arc features, in the initial state, (2.1)
As described above, the arc radius range is set so as to be able to cover all the ranges in which the target component may exist, and only those falling within the set radius range are regarded as arc features.

In the continuous recognition / removal state, (2.
As described in 2), the rough detection setting range A is set in consideration of the depth of one inspection object + α with respect to the previous detection result as the center of the previous detection result, and only those that fall within the set rough detection setting range A are arcs. Features. The rough detection setting range A is set in advance by the rough arc radius range calculation unit 50 using the positional relationship shown in FIG. 1, and the value is stored in the memory 70.
The extracted arc virtual circle candidate point sequence setting unit 12 for rough detection considers a virtual circle having the same data as the center and radius of the extracted arc, calculates the position of a point on the circumference of the virtual circle, A point sequence obtained as the circumference of the virtual circle is set as a circle candidate point sequence.

Based on the set circle candidate point sequence, the circle feature data calculation unit 20 calculates the circle feature as described in (4). On the other hand, the point sequence data extraction unit 30 for precise detection
Includes an arc extraction unit 31 and an extraction arc virtual circle candidate point sequence setting unit 32. When extracting the arc feature, the arc detection unit 31 for precision detection sets the precision detection setting range in consideration of the depth of the small range β up and down around the rough detection result, as described in (2.3). B is set, and only those that fall within the set precision detection set range B are arc features.

The precision detection setting range B is set in advance by the precision arc radius range calculation unit 60 using the positional relationship shown in FIG. 2, and its value is stored in the memory 70. The extracted arc virtual circle candidate point sequence setting unit 32 for precise detection considers a virtual circle having the same data as the center and radius of the extracted arc, calculates the position of a point on the circumference of the virtual circle, A point sequence obtained as the circumference of the virtual circle is set as a circle candidate point sequence. Based on the set circle candidate point sequence, the circle feature data calculation unit 40
Calculates the circular feature as described in (4).

[0029]

As described above, according to the present invention, the following effects can be obtained as described in detail with reference to the embodiments. (1) In an input image, even if oil or dust adheres to the surface of a component, the position and orientation of a circular feature such as a “hole” or “end point of a cylinder” on the component can be accurately detected. (2) By extracting the circle feature, it is possible to detect what position the target component is three-dimensionally located. (3) Since the arc radius range for extracting the arc feature is divided into two stages, setting is easy, and an accurate recognition result can be obtained.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing the setting of a rough arc radius range in a continuous recognition / removal state according to the present invention.

FIG. 2 is an explanatory diagram showing a setting of a precise arc radius range after a rough detection result in the present invention.

FIG. 3 is a block diagram illustrating a precision three-dimensional position and orientation detection apparatus according to an embodiment of the present invention.

FIG. 4 is a block diagram illustrating an example of an apparatus for detecting a three-dimensional position and orientation of a circular feature.

FIG. 5 is a block diagram showing an example of an apparatus for detecting a three-dimensional position and orientation of a circular feature.

FIG. 6 is a block diagram illustrating an example of an apparatus for detecting a three-dimensional position and orientation of a circular feature.

[Explanation of symbols]

 A, B Depth 10, 30 Point sequence data extraction unit 11, 31 Arc extraction unit 12, 32 Extracted arc virtual circle candidate point sequence setting unit 20, 40 Circle feature data calculation unit 50 Rough arc radius range calculation unit 60 Precision arc radius range Calculation unit 70 Memory

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06T 1/00 315 G06T 7/00 300D 7/00 300 G01B 11/24 K F term (Reference) 2F065 AA04 AA27 AA37 AA53 BB05 DD00 DD13 FF01 FF04 JJ03 JJ05 JJ09 JJ26 QQ03 QQ25 QQ31 RR02 SS03 TT02 UU04 UU05 5B057 AA01 BA02 CA08 CA13 CA16 CE15 DA06 DC03 DC08 DC09 DC16 5L096 BA03 FA04 FA06 FA68 FA69 JA09

Claims (3)

[Claims] 1. Limiting an installation state in which a hole of a target component to be detected is not substantially inclined with respect to an optical axis of a camera, and extracting an arc feature, only those falling within a set radius range are extracted as an arc feature. In the precision three-dimensional position / posture detection device for a circular feature of a component for which the edge line segment that is the base of the circular arc is set as a circle candidate point sequence and a circular feature is calculated based on the set circle candidate point sequence, The rough arc radius setting means for setting the arc radius range so as to be able to cover the entire range where the target part may exist, and a small vertical scale centering on the rough detection result obtained by the rough arc radius setting means A precision three-dimensional position / orientation detecting apparatus for a circular feature of a component, comprising: a precision arc radius setting means for setting an arc radius range in consideration of a range component. 2. Limiting an installation state in which a hole of a target component to be detected is not substantially inclined with respect to an optical axis of a camera, and extracting an arc feature, only those falling within a set radius range are extracted as an arc feature. In the precision three-dimensional position and orientation detection device for a circular feature of a component for which the edge line segment that is the base of the circular arc is set as a circle candidate point sequence and a circular feature is calculated based on the set circle candidate point sequence, In the take-out state, rough arc radius setting means for setting an arc radius range in consideration of the depth of one or more inspection objects above and below the previous detection result, and rough detection calculated by the rough arc radius setting means A precision three-dimensional position / orientation detecting apparatus for a circular feature of a component, comprising: a precise arc radius setting means for setting an arc radius range in consideration of a small range up and down with a result as a center. 3. A precision three-dimensional position and orientation detecting apparatus for circular features of a part according to claim 1 or 2, wherein a means for measuring a three-dimensional position of the extracted point sequence is provided, based on the position data of the point sequence. Means for calculating a plane, means for performing a circle approximation on point sequence data on the calculated plane, and means for selecting the one that matches the set data from the approximated circle, and three-dimensional position and orientation of a circle feature with respect to the selected circle And a means for detecting the position of the component.