JP3243905B2 - Object recognition method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object recognition method and apparatus for recognizing an object using three-dimensional information.

[0002]

2. Description of the Related Art A conventional object recognition method is described in, for example, Masatake Oshima, "Study of Object Recognition Using Three-Dimensional Information", Research Report No. 826 (July 1982), pp. 20-45, of the Electrotechnical Laboratory. Has been described. This object recognition method is for recognizing an object by the following procedure.

(A) Three-dimensional coordinate data of a point A on an object surface is measured over multiple points (FIG. 8A).

(B) The points A are grouped into units called surface elements B, and assuming that each surface element B forms a plane, a plane equation is obtained (FIGS. 8B and 9). (Λ, μ, ν)
Is the direction cosine and p is the distance perpendicular to the plane from the origin, the plane equation is expressed by the following equation (51).

Λx + μy + νz = p (51)

(C) The basic regions R1 to R9 are generated by integrating the surface elements which are close to each other and have similar plane equations as shown in FIG.

[0007] A surface element serving as a nucleus for forming a basic region is evaluated by m in the following equation (52).

M = w / n + S (52)

Here, w is a constant, n is the number of nucleus candidates in the vicinity of 8, and S is the average value of the standard deviation in the vicinity. When m is small, the surface element has many candidates nearby and forms a smooth surface. At each time point, a surface element that gives the minimum value of m is selected as a nucleus.

[0010] As shown in FIG. 10, a surface element B is integrated around a nucleus B1 to enlarge an area B2. If the plane equation of the target surface element is similar to the plane equation applied to the points constituting the region, the surface element is integrated into the region. This determination is made by the following equation (53).

D _k ² = (λ _k −λ _r ) ² + (μ _k −μ _r ) ² + (ν _k −ν _r ) ² <d _t ² (53)

[0012] _{| p k -p r | <p} t ··· (54)

Here, _k , _r , and _t are subscripts representing a surface element, a region, and a threshold value, respectively. Note that B3 in FIG.
Indicates an inspection area.

(D) As shown in FIG. 8D, FIG. 8E, and FIG. 8B, a larger area is generated, and the scene is described and recognized based on the positional relationship.

[0015]

In the conventional object recognition method, a polyhedron having a small surface irregularity and a high surface accuracy, that is, a high flatness, can be effectively recognized. However, FIG.
For an object M composed of a surface having irregularities as shown in FIG. 11, a meaningless region and a disjointed region are generated as shown in FIG. Therefore, there is a problem that a surface having irregularities cannot be classified, and as a result, an object cannot be recognized. The reason is explained as follows (g) and (h).

(G) Nuclei cannot be generated.

FIG. 12 shows a surface 60 having irregularities.
This is obtained by applying a plurality of planes (plane elements B) given by Expression (51). Of these surface elements B, the smoothest surface element according to equation (52) is defined as a core B1. However, the surface element B adjacent to the core B1 does not form a smooth surface due to unevenness. Therefore, the direction 62 of the normal vector of the nucleus B1 does not always represent the direction 64 of the normal vector of the surface 60 having irregularities. Therefore, the obtained core B1 has no meaning.

(H) An attempt to create a region cannot be made. Alternatively, a meaningless region is generated.

Here, as shown in FIG.
In order to integrate the region into the nucleus B1 ₁ that was forcibly obtained by the formula (53), the plane elements B1 ₂ , B
1 Select the _3. That is, B1 ₁ , B1 ₂ and B1 ₃ are:
It is assumed that the direction cosines (λ, μ, ν) are all equal. However, even the nucleus B1 ₁ in the direction 62 ₁ and surface elements B1 _2, B1 ₃ direction 62 _2, 62 ₃ are the same, the distance from the origin p
₁ , p ₂ and p ₃ increase as the distance from the origin increases. Therefore, as is apparent from the equation (54), the plane elements B1 ₂ and B1 ₃ cannot be integrated into the area. As a result, the regions cannot be integrated while remaining isolated.

If _dt in equation (53) is increased to expand the range that can be integrated into an area, erroneous area integration as shown in FIG. 14 occurs. In this figure, the direction 62 of the normal vector of the nucleus B1 _{1 on} the surface 60 ₁ having irregularities
_1, it is considered the same as the direction 62 ₄ of the normal vector of the surface elements B1 ₄ in the plane 60 ₄ having other irregularities. Moreover, the distances p ₁ and p ₄ from these origins are also considered to be the same. Thus, equation (53), by (54), surface elements B1 ₄ would be integrated in the area of nuclear B1 _1. However, as is apparent from FIG. 11, the region generated by this method does not correspond to each surface of the polyhedron, and becomes completely meaningless. That is, d _{t in} equation (53)
Even if the number of surface elements to be integrated is increased by increasing the number of uncertainties, an uncertain element increases in the distance p from the origin, so that a correct area cannot be generated after all.

[0021]

SUMMARY OF THE INVENTION An object of the present invention is to provide an object recognition method and apparatus capable of accurately recognizing an object having a surface having irregularities.

[0022]

The object recognizing method according to the present invention has been made in order to achieve the above object.
By dividing the three-dimensional coordinate data of many points on the surface having irregularities in the object into a plurality of points, and projecting the normal vectors of these surface elements onto a spherical surface having a radius of 1, the plurality of points A large number of surface elements as small planes are generated based on these, and a set of those having similar normal vectors is obtained for these surface elements. For each of the surface elements of these sets, along the normal vector from the center of each of the surface elements Then, the distance between a plane including an adjacent surface element in the same set is determined, and when these distances are equal to or smaller than a predetermined value, it is determined that the surface element and the adjacent surface element are on the same surface. Things. So
Then, the constant value is set to the maximum undulation of the surface.
It is the width.

Here, "same plane" means the same plane or curved surface.

The object recognition device according to the present invention is for use in carrying out the object recognition method according to the present invention.

[0025]

The uneven surface of an object is constituted by a number of points, each of which is represented as a three-dimensional coordinate. The three-dimensional coordinate data is divided into a plurality of points to generate a large number of minute planes formed by these points. These planes are called plane elements. That is, the entire object surface is constituted by a large number of surface elements obtained from the three-dimensional coordinate data. Next, by projecting the normal vector of each surface element onto a spherical surface having a radius of 1, a set of similar surface vectors for each surface element is obtained. Each set is composed of surface elements oriented in a certain range. However, this set includes surface elements that have different levels in the same or similar directions, that is, different planes. Therefore, for each surface element, the distance from the center of each surface element to the plane including adjacent surface elements in the same set along the normal vector is determined, and these distances are fixed values (that is, the maximum value of the surface). If the amplitude is equal to or less than the amplitude of the undulation, it is determined that the surface element and the adjacent surface element are on the same plane.

[0026]

FIG. 1 is a flowchart showing an embodiment of an object recognition method according to the present invention, and FIG. 2 is a block diagram showing an embodiment of an object recognition apparatus according to the present invention. Hereinafter, an embodiment of the present invention will be described with reference to these drawings.

The object recognizing device according to the present invention includes a three-dimensional coordinate data input unit 10 for inputting three-dimensional coordinate data of a large number of points on an object surface, dividing the three-dimensional coordinate data into a plurality of points, and Surface element generating means 12 for generating a large number of surface elements as minute planes based on a plurality of points; normal vector classifying means 14 for obtaining a set of elements having similar normal vectors for the surface elements; An inter-element distance calculating means 16 for obtaining a distance between the element and a plane including adjacent element elements in the same set;
If the amplitude is equal to or less than the maximum undulation of the surface, the surface element and the adjacent surface element are determined to be on the same surface, and the same surface determination means 18 is combined with the same surface to combine And an object shape output section 20 for outputting the shape of the object surface.

The three-dimensional coordinate data input unit 10 obtains three-dimensional coordinate data of the surface of an object by a known method such as a light cutting method, and is constituted by a slit light source, a CCD camera, a moving stage, and the like. I have. The object shape output unit 20 is a CRT, a printer, or the like. The data processing unit 2 includes a surface element generation unit 12, a normal vector classification unit 14, a surface element distance calculation unit 16, and a same plane determination unit 18.
2. The data processing unit 22 is realized by a computer or the like.

FIG. 3 is a perspective view showing surface elements constituting an object formed of a surface having irregularities, FIG. 4 is a distribution diagram showing normal vectors of the surface elements, and FIG. FIG. 6 is a perspective view showing a set, FIG. 6 is an explanatory view showing a method for calculating a distance between surface elements, and FIG. 7 is a perspective view showing a state in which an object formed of a surface having irregularities is recognized by this embodiment. Hereinafter, description will be made with reference to FIGS.

The three-dimensional coordinate data of a number of points on the surface of the object M obtained by the three-dimensional coordinate data input unit 10 is divided into a plurality of points by the surface element generating means 12 (step 1).
01). The plurality of points is, for example, a total of 64 points of 8 × 8. Each point may be partially or entirely overlapped and divided. Subsequently, the plurality of points are regarded as planes by a method such as the least squares method, and each of these planes is set as a plane element B (step 102).

The normal vector classifying means 14 obtains a set of the surface elements B having similar normal vectors (step 103). That is, when the normal vector of each surface element B is projected onto a spherical surface having a radius of 1, a distribution as shown in FIG. 4 is obtained. The normal vector of the high-density portion in the space indicates the direction of the surface having irregularities. Therefore, the surface elements of the normal vector of the high density part are selected in order,
When surface elements having a normal vector similar to the direction are integrated with the surface element, a set of surface elements having similar directions (including the same direction) such as R1 to R3 shown in FIG. Can be classified.

However, as is apparent from FIG.
These sets R1 to R3 include all surface elements having the same direction as a surface. That is, R1 has three surfaces, R
2 and R3 each include two faces. Therefore, it is necessary to classify these into sets of planes with different levels. Here, if the vertical distance from the origin to each plane element is used as in the prior art, an erroneous area is generated. Therefore, in the present invention, attention is paid to how the plane elements are connected.

That is, the inter-plane element distance calculating means 16 obtains the distance between each plane element of the set and a plane including adjacent plane elements in the same set (step 104). Subsequently, the same plane determination unit 18 determines that the distance is a constant value (eg,
In other words , if the amplitude is equal to or smaller than the maximum undulation of the surface, it is determined that the surface element and the adjacent surface element are on the same plane (step 105). That is, one face element is sequentially extracted from the set of face elements. Then, as shown in FIG. 6, from the center ρ (a, b, c) of the plane element, the center A ₀ (x ₀ ,
y _0, z ₀₎ the distance h to be calculated by the following equation (1).

H = | ax ₀ + by ₀ + cz ₀ | / √ (a ² + b ² + c ² ) (1)

Here, the distance h is within a certain range (hmax).
, That is, if the distance h satisfies the following equation (2), ρ
It is determined that A ₀ is present on the same plane of.

H <hmax (2)

If this process is performed until there are no more isolated surface elements, it is possible to classify the surface region into a region corresponding to the surface having each unevenness as shown in FIG. Since Place this hmax amplitude maximum waviness with the surface, allowing the most accurate classification. FIG. 7 is output by the object shape output unit 20 such as a CRT or a printer.

[0038]

According to the present invention, the direction of a surface having irregularities can be accurately recognized by obtaining a set of surface elements constituting an object surface having similar normal vectors. In addition, the distance between each plane element and the plane including the adjacent plane element in the same set is obtained. If these distances are equal to or smaller than a certain value, the plane element and the adjacent plane element are on the same plane. As a result, it is possible to accurately classify a surface having unevenness having a similar direction for each surface having different levels. In addition to this, the constant value
To the maximum swell amplitude of the surface.
The most accurate classification is possible.

[Brief description of the drawings]

FIG. 1 is a flowchart showing one embodiment of the present invention.

FIG. 2 is a block diagram showing one embodiment of the present invention.

FIG. 3 is a perspective view showing a surface element constituting an object having a surface having irregularities.

FIG. 4 is a distribution diagram showing a normal vector of a surface element in the embodiment.

FIG. 5 is a perspective view showing a set of normal vectors for each similar surface element in the embodiment.

FIG. 6 is an explanatory diagram illustrating a method of calculating a distance between surface elements according to the present embodiment.

FIG. 7 is a perspective view showing a state in which an object having a surface having irregularities is recognized according to the embodiment.

FIG. 8 is an explanatory diagram of a conventional example, and FIGS.
(B), FIG. 8 (c), FIG. 8 (d), FIG. 8 (e), FIG.
The state of processing data in the order of (f) is shown.

FIG. 9 is a perspective view showing the concept of a surface element.

FIG. 10 is a perspective view showing a state in which plane elements are integrated around a nucleus to enlarge an area in a conventional example.

FIG. 11 is a perspective view showing a state where an object having a surface having irregularities is recognized by a conventional example.

FIG. 12 is an explanatory diagram showing the drawbacks of the conventional example.

FIG. 13 is an explanatory diagram showing a drawback of the conventional example.

FIG. 14 is an explanatory view showing a drawback of the conventional example.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Three-dimensional coordinate data input part 12 Surface element generation means 14 Normal vector classification means 16 Surface element distance calculation means 18 Same plane determination means 20 Object shape output part M Object consisting of a surface having irregularities

Continuation of the front page (56) References Masatake Oshima, “Study of Object Recognition Using Three-Dimensional Information”, Research Report of the Electrotechnical Laboratory, July 1982, No. 826, p. 20-45 Masatake Oshima, Yoshiaki Shirai, "Object Recognition Using Three-Dimensional Information", IEICE Transactions D, May 1982, Vol. J65-D, NO. 5, p. 629-636 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06T 1/00-17/50

Claims (2)

(57) [Claims] 1. The method according to claim 1, further comprising the steps of:
The three-dimensional coordinate data of a point is divided into a plurality of points, and based on these points, a large number of surface elements as minute planes are obtained.
Generate the numbers and project the normal vectors of these surface elements onto a sphere of radius 1.
The normal vectors are similar for the surface elements
Of each of these sets, from the center of each
Adjacent faces along the normal vector in the same set
The distance to the plane containing the element is found, and if these distances are less than a certain value, the plane element and the
Is determined to be on the same plane as the adjacent plane element ofAnd The constant value is defined as the maximum swell amplitude of the surface.
Was  An object recognition method characterized in that: 2. The method according to claim 1, wherein the surface of the object has a plurality of irregularities.
3D coordinate data input to input 3D coordinate data of points
And dividing the three-dimensional coordinate data into a plurality of points.
Based on these multiple points, many surface elements
Surface element generating means for generating numbers, and projecting normal vectors of these surface elements onto a spherical surface having a radius of 1
The normal vectors are similar for the surface elements
Normal vector classifying means for obtaining a set of objects to be processed, and for each surface element of the set, a normal line from the center of each surface element
Along the vector, the adjacent face elements in the same set
An inter-plane element distance calculating means for obtaining a distance to a plane including the plane element;
Same plane judgment to judge that the touching surface element is on the same plane
Means for outputting the shape of the object surface by combining the same surface
Object shape output unitWith The constant value is defined as the maximum swell amplitude of the surface.
Was An object recognition device characterized by the above-mentioned.