JPH07271979A - Image recognizing method and device therefor - Google Patents

Abstract

PURPOSE:To improve a recognition ratio and to reduce erroneous recognition by increasing feature elements to be extracted from an image. CONSTITUTION:The image 2a of a model object is fetched by a camera 1a, and straight lines and circular arcs of feature elements are extracted by a feature extractor 3a. From these straight lines and circular arcs, a pair extractor 4 combines a pair of specified straight lines and extracts a straight line pair where the relative location is defined as numerical information. In the pair extractor 4, a pair of circular arcs and a pair of a straight line and a circular arc are extracted in addition to the pair of straight lines. A memory 5 preserves each feature element. An object to be an deciding object fetches an image 2b by a camera 1b and extracts stright lines and the circular arcs by a feature extractor 3b in the same way as the model object. A pair generator 6 extracts a pair of straight lines, a pair of circular arcs and a pair of the straight line and the circular arc from these stright lines and circular arcs. A corresponding device 7 recognizes the existence and the location of the object by making each feature element of the object to be the deciding object correspond to that of the model object and deciding the difference.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition method and an apparatus for recognizing an object image by comparing characteristic elements of an object with characteristic elements of a model image and recognizing the difference.

[0002]

2. Description of the Related Art Conventionally, as a method of recognizing an object, a picked-up image is compared with a model image and the difference is determined. The images are compared with each other by a method of converting each pixel into binary data for comparison, and by extracting a straight line and an arc forming an image as characteristic elements and using characteristic correlation. In particular, the feature correlation expresses the contour of an object by subjecting the image to a filtering process and expressing it as a set of straight lines and arcs.

[0003]

SUMMARY OF THE INVENTION In the above conventional method,
It is difficult to fully represent a specially shaped object with feature elements consisting only of straight lines and arcs. In the case of an image with a lot of noise, a straight line or a circular arc may be shorter than the actual object or may be divided. In particular, when the arc is divided, the directivity also changes significantly. For this reason, the recognition rate is lowered and erroneous recognition is caused.

The present invention has been made in view of the above circumstances, and an object thereof is to provide an image recognition method and apparatus for increasing the number of characteristic elements to be extracted to improve the recognition rate and prevent erroneous recognition. To do.

[0005]

According to the present invention, in order to solve the above-mentioned problems, straight lines and arcs forming an image are digitized into characteristic elements, and those of an object with respect to characteristic elements of a model image are compared. In the image recognition method for recognizing an object image based on the difference, a method of adding relative positional relationship between paired straight lines, relative positional relationship between arcs and relative positional relationship between straight lines and arcs to a feature element is adopted. .

In order to carry out this method, the image pickup units 1a and 1b for capturing an object as an image, and the feature extraction for extracting the characteristic elements by digitizing the straight lines and arcs constituting the image from the captured images 2a and 2b. A pair extraction unit 4 and a pair generator for extracting the characteristic elements quantifying the relative positional relationship between the sections 3a and 3b and the straight lines forming a pair, the relative positional relationship between the arcs, and the relative positional relationship between the straight lines and the arcs. The image recognition apparatus is configured from 6, the storage unit 5 that stores the characteristic elements of the model image, and the associating unit 7 that compares the characteristic elements of the determination target object with the model image and determines the difference.

[0007]

According to the method of the present invention, in addition to the straight lines and arcs that make up the image, the relative positional relationship between the straight lines, the relative positional relationship between the circular arcs, and the relative positional relationship between the straight lines and the circular arcs are also digitized to obtain the image. It is a characteristic element. Then, after extracting the characteristic element of the object, it is compared with the characteristic element of the model image extracted in advance. For example, when the difference is within a predetermined error, the difference is determined to be the same, and the difference is determined to determine whether or not the object is the same as the model object.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. In FIG. 1, the processing apparatus according to the embodiment of the present invention includes a camera 1a that captures an image. Camera 1
In the image 2a of the model object captured and captured in a, the feature extractor 3a extracts straight lines and arcs of the feature elements by, for example, performing filtering processing on the image.

The pair extractor 4 further combines two specific straight lines from the straight lines and arcs extracted by the feature extractor 3a, and extracts a straight line pair whose relative position is used as numerical information. The pair extractor 4 extracts not only pairs of straight lines but also pairs of arcs and pairs of straight lines and arcs. That is, in the conceptual configuration of the characteristic element that describes an object as numerical information, a straight line and a circular arc are first extracted, and therefore, as shown in FIG. 2, a straight line description configuring section P1 and an arc description configuring section P2. It is roughly divided into. Next, since mutual pairs are extracted in the respective description constituent parts P1 and P2, the description constituent part P3 of a straight line pair and the description constituent part P4 of an arc pair are located. Further, since the mutual pair of the straight line and the circular arc is also extracted, the description constituent part P5 of the pair of the straight line and the circular arc is located so as to straddle the description constituent parts P1 and P2.

Numerical information of straight line pairs of straight lines is shown in FIG.
As shown in (A), the lengths L1 and L of the two line segments A and B
2. Coordinates (x, y) of the midpoint of the line segment connecting the midpoints of the line segments A and B, the angle deg between them, and the coordinate point (x to the horizontal line from the intersection on the extension line of the line segments A and B). , Y) of each angle θ. Also, the numerical information of the arc pair is shown in FIG.
As shown in, the distance dist between the centers of the two arcs C1 and C2, dist,
The radii r1 and r2 of the arcs C1 and C2, the coordinates (x, y) of the midpoint between the arcs C1 and C2, the angle deg between them, and the angle θ with respect to the horizontal line of the straight line connecting the arcs C1 and C2 are taken. As shown in FIG. 3C, the numerical information of the straight line / arc pair includes the coordinates L of the length L of the line segment D, the midpoint of the line connecting the midpoint of the line segment D and the center point of the arc C ( x, y), radius r of arc C, angle θ of the line connecting the midpoint of line segment D and the center point of arc C with respect to the horizontal line
And each value of.

Here, for example, if the feature extractor 3a extracts nL straight lines and nA arcs, the pair extractor 4 can extract _nL C ₂ pair straight lines, and a model is used for selecting the pair. As appropriate. Pair arcs,
Similarly, it is assumed that _nA C ₂ and (nL × nA) pairs of straight lines and arcs can be extracted, and that the pair is appropriate as a model. The memory 5 stores each characteristic element.

On the other hand, the object to be judged to be recognized is imaged by the camera 1b in the same manner as the model object, the image 2b is taken in, and the straight line of the characteristic elements is extracted by the characteristic extractor 3b. The arc and are extracted. The pair generator 6 extracts a straight line pair, a circular arc pair, and a straight line and circular arc pair from the straight line and the circular arc in the same manner as above. The correlator 7 recognizes the existence and position of the object by associating each characteristic element of the object to be determined with that of the model object and determining the difference.

The method of the present embodiment uses straight lines forming an image,
In addition to arcs, relative positional relationships between straight lines, relative positional relationships between arcs, and relative positional relationships between straight lines and arcs are also characteristic elements of an image. Then, after the feature elements of the object to be judged are extracted by the feature extractor 3b and the pair generator 6, the feature elements of the image of the model object previously extracted and stored in the memory 5 are compared in the correlator 7. . In the case of a straight line pair among the characteristic elements, the error value calculated by the following equation is used as the criterion, and the pair with the smallest error value is used as the correspondence result.

[0014]

[Equation 1]

In the case of an arc pair, the error value calculated by the following equation is used as the criterion, and the pair having the smallest error value is used as the correspondence result.

[0016]

[Equation 2]

In the case of a pair of a straight line and a circular arc, the error value calculated by the following equation is used as a criterion, and the pair having the smallest error value is used as the correspondence result.

[0018]

[Equation 3]

In this way, since the difference between a large number of corresponding parts is represented by a numerical value, the judgment object is recognized by judging the difference with the model object according to the ratio of whether each is within a predetermined error range.

[0020]

As described above, according to the present invention,
Even if the object has a special shape that cannot be described only by a straight line or a circular arc, adding a pair to the feature element increases the analysis capability and makes it possible to recognize the object. In addition, even if the image contains a lot of noise and the extracted straight lines or arcs become shorter or fragmented, using pairs not only allows you to compare individual straight lines and arcs Since the relative positional relationship between straight lines and arcs forming a pair is also used as the correspondence reference, the recognition rate is improved and erroneous recognition is reduced.

[Brief description of drawings]

1 is a block diagram of an apparatus for implementing the method according to the invention.

FIG. 2 is a descriptive configuration diagram of a characteristic element of an object.

FIG. 3 is an explanatory diagram of numerical information of each pair.

[Explanation of symbols]

 1a, 1b ... Camera 2a ... Model image 2b ... Judgment target image 3a, 3b ... Feature extractor 4 ... Pair extractor 5 ... Memory 6 ... Pair generator 7 ... Correlator

Claims (2)

[Claims] 1. An image recognition method for quantifying straight lines and arcs constituting an image as characteristic elements, comparing the object with respect to the characteristic elements of a model image, and recognizing an object image based on the difference between the characteristic elements. In the image recognition method, the relative position relationship between straight lines forming a pair, the relative position relationship between arcs, and the relative position relationship between straight lines and arcs are included. 2. A relative positional relationship between a pair of straight lines and an image pickup unit that captures an object as an image, a feature extraction unit that extracts a characteristic element that digitizes straight lines and circular arcs that compose the object from the captured image. , A pair extraction unit that extracts a characteristic element that quantifies a relative positional relationship between arcs and a relative positional relationship between a straight line and an arc, a storage unit that stores a characteristic element of a model image, and a characteristic element of a determination target object. An image recognition apparatus, comprising: an associating unit that compares the model image and determines whether the model image is different.