JPH0829358A - Product inspection method by image processing - Google Patents

Abstract

PURPOSE:To detect a three-dimensional defect by processing an image obtained by photographing an inspection product with one single-lens camera and detecting the three-dimensional shape of the inspection product. CONSTITUTION:The image of an inspection product 1 carried by a belt conveyor 2 is shot by one single-lens camera 3. An image processing part 6 obtains the three-dimensional shape of the inspection product 1 by processing an image at a time (t) and that at a time (t+1). Then, the three-dimensional defect of the inspection product is detected by comparing the three-dimensional shape of the inspection product 1 with that of a conforming product.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for inspecting a product by image processing, which is a method for inspecting a product by processing an image obtained by a single monocular camera so as to detect a three-dimensional defect. Is.

[0002]

2. Description of the Related Art An image processing apparatus is used to inspect products flowing on a line. In this image processing device, a product is photographed by a camera, an image signal obtained by photographing is quantized by an A / D converter, and the quantized image information (digital image information) is stored in a memory. Then, the computer reads the digital image information stored in the memory and performs signal processing (described later) to detect a defect in the product.

A signal processing procedure performed by a computer for detecting a defect of a product is performed as follows. Here, the image signal is an image signal obtained by shooting with a single monocular camera.

First, an inspection product is photographed, digitized, and the contour of the product is obtained from the digital image information stored in the memory. The contour is obtained by subjecting the digital image information to differential filtering and thresholding. By tracking the contour of the image, a closed region corresponding to the product can be obtained. In advance, a contour obtained by photographing a normal product and performing image processing (this is referred to as a “reference contour”) is stored in a memory. Matching (comparative inspection) is performed between the reference contour and the contour of the inspection product (this is referred to as “inspection contour”). As a result of the matching, if there is no difference between the reference contour and the inspection contour, it is determined as a good inspection product, and if there is a difference, it is determined as a defective product.

[0005]

In the prior art of detecting a product defect by performing image processing on an image signal taken by a single monocular camera, the defect is detected only by matching the reference contour and the inspection contour. Therefore, it was not possible to detect a three-dimensional defect such as "dent" or "protrusion" of the product.

In view of the above-mentioned conventional technique, the present invention is a method of performing image processing on an image signal obtained by shooting with a single monocular camera to perform product inspection, and image processing capable of detecting a three-dimensional defect. The purpose is to provide a product inspection method according to.

[0007]

The structure of the present invention for solving the above-mentioned problems is such that an inspection product moving in one direction is photographed by a single monocular camera, and the photograph is taken on the projection surface of the camera. It is a product inspection method by image processing that judges the quality of the inspection product by processing the image formed on the
The motion vector obtained from the inspection product image on the projection surface of the inspection product at (t) and the inspection product image on the projection surface of the inspection product at time (t + 1), and the cubic of the inspection product at time (t). The relational expression between the original coordinates and the two-dimensional coordinates on the projection plane of the inspection product image, the three-dimensional coordinates of the inspection product at time (t), and the three-dimensional coordinates of the inspection product at time (t + 1) The relational expression and the three-dimensional coordinates of the inspection product at time (t + 1),
Obtain the three-dimensional coordinates of the inspection product from the relational expression with the two-dimensional coordinates on the projection plane of the inspection product image, and compare the obtained three-dimensional coordinate value with the preset three-dimensional coordinate value. The quality of the inspected product is determined by.

Further, the structure of the present invention for solving the above problems is as follows.
In a three-dimensional coordinate system in which the x-axis is defined in the vertical direction and the z-axis in the horizontal direction, the inspection product moving at the speed v in the x-axis direction,
The X and Y axes are defined in the horizontal plane according to the x and y axes,
A product inspection apparatus by image processing, wherein a single monocular camera having an optical axis on the z axis and a projection surface on the X and Y axes is imaged, and the quality of the inspected product is determined by processing the imaged image. The motion vector V _m obtained from the inspection product image on the projection surface of the inspection product at time (t) and the inspection product image on the projection surface of the inspection product at time (t + 1), and the inspection at time (t) Relational expression (1) between the 3D coordinates of the product and the 2D coordinates on the projection plane of the inspection product image, the 3D coordinates of the inspection product at time (t), and the inspection product at the time (t + 1) Relational expressions (3), (4), and (5) with the three-dimensional coordinates of
From the relational expressions (8) and (9) between the three-dimensional coordinates of the inspection product at (t + 1) and the two-dimensional coordinates of the inspection product image on the projection surface, the relational expressions (10), (11) and (12) are obtained. The three-dimensional coordinates x _i (t), y _i (t), and z _i (t) of the inspection product shown in are obtained, and the obtained three-dimensional coordinate values are compared with preset three-dimensional coordinate values. The quality of the inspected product is determined by.

[Equation 2]

[0009]

In the present invention, the image is obtained by photographing with one monocular camera,
By processing the inspection product image at time (t) and the inspection product image at time (t + 1), the 3D shape of the inspection product can be detected, and the 3D shape of the inspection product and the 3D shape of the non-defective product can be compared. By doing so, it is possible to detect a three-dimensional defect in the inspection product.

[0010]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows the system configuration of an embodiment of the present invention. As shown in the figure, the inspection product 1 is placed on the belt conveyor 2 and conveyed in the direction of arrow A. The single-lens single camera 3 is arranged above the belt conveyor 2 which moves at a constant speed v, photographs the inspection product 1 from above, and outputs an image signal E. The image signal E is quantized into a digital image signal e by the A / D converter 4, and the digital image signal e is stored in the memory 5. The image signal processing unit 6 reads the stored digital image signal e and performs the signal processing described below to detect defects (including not only two-dimensional defects but also three-dimensional defects) of the inspection product 1. Then, the processed image is displayed on the monitor 7.

The three-dimensional coordinate axes x, y and z are taken in the directions shown in FIG. That is, the conveyance direction A of the inspection product 1 is the same as the x direction, the direction orthogonal to the x direction in the horizontal plane is the y direction, and the optical axis direction (vertical direction) of the camera 3 is the z direction.

Next, a signal processing procedure by the image signal processing unit 6
Will be described with reference to FIGS. 2 and 3 are
The projection image (Image Plan) shown in both figures.
e) S _IIs the focus from the camera's focus F as shown in FIG.
A plane that is parallel to the xy plane at a position separated by a point distance f
Is. In addition, the three-dimensional coordinate axes are lower case x, y, z
Shown, projection plane S_IThe upper coordinate axes are indicated by capital letters X and Y.

2 and 3, 1 (t) is the inspection product at time t, 1 (t + 1) is the inspection product at time (t + 1), F is the focus of the camera 3, and f is the focal length. , Respectively.

The image signal processing unit 6 first obtains a motion vector V _m from the image at time t and the image at time t + 1. That is, as shown in FIG. 2, a point q at which an arbitrary point q of the inspection product 1 (t) is projected on the projection surface S _{I at} time t is Q (t), and at time t + 1, the inspection product 1 (t + 1) is detected.
Q (t + 1) is a point on the projection plane S _I
Then, the vector having the point Q (t) as the starting point and the point Q (t + 1) as the ending point becomes the motion vector V _m . This motion vector V _m is obtained for each of a large number of corresponding points between the image at time t and the image at time t + 1.

Next, the difference filter processing and the threshold processing are performed on the images at time t and time t + 1 to obtain contours corresponding to the inspection products 1 (t) and 1 (t + 1). Then, the motion vector starting from the closed area surrounded by the contour showing the inspection product image (the motion vector obtained from the characteristic point of the inspection product 1) is left, and the motion vector starting from the outside of this closed area (the characteristic of the belt conveyor 2). The motion vector obtained from the point) is deleted.

As shown in FIG. 3, the x-axis direction of the inspection product 1
Direction, y-axis direction, z-axis movement amount is T _x, T_y, T_zWhen
Then, the focal length of the camera 3 is f, and the inspection at time t
Feature point q in product 1 (t)_iLet the three-dimensional coordinates of (t) be (x
_i(t), y_i(t), z_i(t)) and the point q_iprojection of (t)
Surface S_IUpper point Q_iSet the coordinates of (t) to (X_i(t), Y_i(t))
Then, the point q_i(t) and point Q_ibetween (t) and
The relationship of (1) is established. There are many feature points,
Q on behalf of one_iIt is indicated by (t).

[0018]

(Equation 3)

Further, the inspection product 1 (t + 1) at time t + 1
The three-dimensional coordinates of the feature point q _i (t + 1) in x _i (t + 1), y _i
If (t + 1) and z _i (t + 1), then the point q _i (t) and the point q _i (t +
The relation of the following equation (2) is established between 1) and.

[0020]

[Equation 4]

The moving speed of the belt conveyor 1 is v,
Since the directions of the x-axis, the y-axis, and the z-axis are determined as shown in FIG. 1, T _x = v, T _y = 0, and T _z = 0. Therefore, the following expressions (3) to (5) are established.

[0022]

(Equation 5)

From the above equations (2) to (5), the following equation (6)
The simultaneous equations shown by (9) are obtained.

[0024]

(Equation 6)

The above equation (6) to the three-dimensional coordinates x _i (t) of the feature point q _i by solving the simultaneous equations shown in (9) (t),
y _i (t) and z _i (t) can be obtained as shown in the following equations (10) to (12).

[0026]

(Equation 7)

In the equations (10), (11) and (12), X _i (t + 1) −X _i (t) is the above-mentioned motion vector V _m .

In this way, each feature point q _i of the inspection product 1
The three-dimensional coordinate of is calculated and used as the inspection three-dimensional coordinate.

On the other hand, even for a normal inspection product having no defect, the above-described processing is performed to obtain the three-dimensional coordinates of each feature point in advance. This three-dimensional coordinate is stored as a reference three-dimensional coordinate.

Then, the reference three-dimensional coordinates and the inspection three-dimensional coordinates are compared, and if there is no difference between the two, it is determined to be a good product, and if there is a difference between the two, it is determined to be defective.

[0031]

According to the present invention as described in detail with reference to the embodiments, the images at different times obtained by one monocular camera are processed to obtain the three-dimensional coordinates of each point of the inspection product. Therefore, by comparing the three-dimensional coordinates of each point of the inspection product with the three-dimensional coordinates of each point of the normal product, the three-dimensional defect of the inspection product can also be detected.

[Brief description of drawings]

FIG. 1 is a system configuration diagram showing an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an image processing state of the present invention.

FIG. 3 is an explanatory diagram showing an image processing state of the present invention.

FIG. 4 is a configuration diagram showing an arrangement state of the embodiment.

[Explanation of symbols]

 1 Inspection Product 2 Belt Conveyor 3 Camera 4 A / D Converter 5 Memory 6 Image Signal Processor 7 Monitor

Claims (2)

[Claims] 1. An inspection product moving in one direction is photographed by a single monocular camera, and an image formed on the projection surface of the camera is photographed to determine the quality of the inspection product. It is a product inspection method by image processing for determination, which is obtained from the inspection product image on the projection surface of the inspection product at time (t) and the inspection product image on the projection surface of the inspection product at time (t + 1). The relational expression between the motion vector, the three-dimensional coordinates of the inspection product at time (t) and the two-dimensional coordinates of the inspection product image on the projection plane, the three-dimensional coordinates of the inspection product at time (t), and the time ( t + 1)
From the relational expression with the three-dimensional coordinates of the inspection product at, the relational expression between the three-dimensional coordinates of the inspection product at time (t + 1) and the two-dimensional coordinates on the projection plane of the inspection product image, A product inspection method by image processing, characterized in that the quality of an inspection product is determined by obtaining three-dimensional coordinates and comparing the obtained three-dimensional coordinate values with preset three-dimensional coordinate values. 2. An inspection product which is moving at a speed v in the x-axis direction within a three-dimensional coordinate system in which the x-axis and the z-axis are defined in the vertical direction in the horizontal plane. The X and Y axes are defined in
A product inspection apparatus by image processing, wherein an image is processed by a single monocular camera whose optical axis is z axis and projection plane is aligned with X and Y axes, and the quality of the inspected product is determined by processing the captured image. Motion vector V _m obtained from the inspection product image on the projection surface of the inspection product at time (t) and the inspection product image on the projection surface of the inspection product at time (t + 1), and the inspection at time (t) Relational expression (1) between the three-dimensional coordinates of the product and the two-dimensional coordinates of the inspection product image on the projection plane, the three-dimensional coordinates of the inspection product at time (t), and the time (t + 1)
Relations with the three-dimensional coordinates of the inspection product in
(5) and the relational expressions (8) and (9) between the three-dimensional coordinates of the inspection product at the time (t + 1) and the two-dimensional coordinates of the inspection product image on the projection surface.
From, the three-dimensional coordinates x _i (t), y _i (t) and z _i (t) of the inspection product shown by the relational expressions (10), (11) and (12) are obtained,
A product inspection method by image processing, characterized in that the quality of an inspection product is judged by comparing the obtained three-dimensional coordinate value with a preset three-dimensional coordinate value. [Equation 1]