Method for determining the quality of linear defects

Abstract

To provide a surface flaw detection method that can easily and reliably evaluate whether linear flaws occurring on the surface of an object are harmful or harmless, even if the photographed image is blurred.SOLUTION: There is provided a method of photographing a surface of a steel billet 1 with a CCD camera 2 and determining the quality of linear flaw based on the flaw width and brightness of the linear flaw in the photographed image, that detects a photographing angle θ of the CCD camera 2 with respect to the above surface, geometrically deformation-corrects trapezoidal distortion of the photographed image at the photographing angle θ into a square, and corrects respectively a flaw width W and brightness ΔL of a flaw image Pi in the deformation-corrected photographed image using a flaw width correction value and a flaw brightness correction value prepared in advance for the photographing angle θ in a computer 4.SELECTED DRAWING: Figure 1

Description

【Technical Field】 【0001】 The present invention Method for determining the quality of linear defects relates to, particularly, a Linear scratches Method for determining whether something is good or bad suitable for detecting linear defects on the surface of steel slabs. 【Background Art】 【0002】 In the detection of linear defects on the surface of a steel slab, the width of the defect serves as an index for evaluating whether the defect is harmful or not. When detecting linear defects occurring on the surface of a specimen of a certain length such as a steel slab, it is reasonable to provide only one imaging means and swing it to photograph the entire surface of the specimen. However, if the imaging means is not directly facing the surface of the specimen, distortion corresponding to the imaging angle occurs in the photographed image, causing the width of the defect in the image to fluctuate, and accurate determination of the defect width cannot be made. 【0003】 Therefore, Patent Document 1 discloses a method that enables accurate determination of the defect width by performing image conversion by utilizing the occurrence of trapezoidal distortion corresponding to the imaging angle in the rectangular frame image at the time of facing directly. 【Prior Art Documents】 【Patent Documents】 【0004】 【Patent Document 1】 Japanese Unexamined Patent Application Publication No. 2017-49152 【Summary of the Invention】 【Problems to be Solved by the Invention】 【0005】 However, in addition to the occurrence of distortion in the photographed image when the imaging angle changes, the distance between the imaging means and the surface of the specimen changes, causing the focus of the imaging means to deviate and resulting in blurring in the photographed image, thereby changing the defect width. Also, the luminance corresponding to the defect depth also serves as an index for evaluating whether the defect is harmful or not, but there is a problem that when blurring occurs in the photographed image, the luminance also changes and accurate evaluation of the linear defect cannot be performed. 【0006】 Therefore, the present invention solves these problems and allows for the simple and reliable evaluation of whether linear defects on the surface of a subject are harmful or harmless, even if the captured image is blurred. Linear defects No judgment method The purpose is to provide. [Means for solving the problem] 【0007】 To achieve the above objective, the first invention involves photographing the surface of a subject (1) with an imaging means (2) and detecting linear defects in the captured image. Image of the defect (Pi) The quality of the linear defect is determined based on its width and brightness. A method which involves detecting the shooting angle (θ) of the imaging means (2) with respect to the surface, and geometrically deforming and correcting the trapezoidal distortion of the captured image at the shooting angle (θ) into a quadrilateral, and the deformed The aforementioned In the captured image The aforementioned The defect width (W) and brightness (ΔL) of the defect image (Pi) are used in advance. The method is characterized by correcting the image using a designated scratch width correction value and a scratch brightness correction value relative to the shooting angle. 【0008】 In this first invention, the defect width and brightness of the defect image in the deformation-corrected captured image are corrected to those of the camera facing directly using pre-prepared defect width correction values ​​and defect brightness correction values ​​for the shooting angle. This eliminates blurring of the captured image due to changes in the distance between the imaging means and the surface of the subject, making it possible to easily and reliably evaluate whether linear defects are harmful or harmless. 【0009】 In this second invention, the defect width correction value and the defect brightness correction value are substituted into the formula for a normal distribution to generate a defect image (Pi) of a linear defect in which the brightness of the cross-section follows the normal distribution. 【0010】 In this second invention, by generating a defect image of a linear defect, it is possible to intuitively determine whether the defect is harmful or harmless. The symbols in parentheses above are for reference only, indicating the correspondence with the specific means described in the embodiments described later. [Effects of the Invention] 【0011】 As described above, according to the present invention Method for determining the quality of linear defects even if blurring occurs in the captured image, it is possible to simply and surely evaluate whether the linear flaw generated on the surface of the subject is harmful or harmless. 【Brief Description of the Drawings】 【0012】 [Figure 1] It is a diagram showing the device configuration for implementing the method of the present invention in the first embodiment. [Figure 2] It is a diagram showing the relationship between the distance from the camera to the surface of the steel sheet and the swing angle of the camera's neck. [Figure 3] It is a diagram showing the swing angle of the camera and the captured image corresponding thereto. [Figure 4] It is a diagram for explaining the projective transformation of the captured image. [Figure 5] It is a diagram showing the trapezoidally deformed captured image and the captured image obtained by projective transformation of this image into a rectangle. [Figure 6] It is a graph showing an example of the calibration function f(θ). [Figure 7] It is a graph showing an example of the calibration function g(θ). [Figure 8] It is a graph showing the luminance distribution when crossing the flaw image of the linear flaw and the luminance distribution after correction in the second embodiment. [Figure 9] It is a diagram showing the flaw image drawn according to the luminance distribution of FIG. 8. 【Modes for Carrying Out the Invention】 【0013】 Note that the embodiments described below are merely examples, and various design improvements made by those skilled in the art without departing from the gist of the present invention are also included in the scope of the present invention. 【0014】 (First Embodiment) In FIG. 1, the Method for determining the quality of linear defectsShows the equipment configuration for carrying out. In Fig. 1, above a steel strip 1 with a certain width and length T as the specimen, a CCD camera (hereinafter simply referred to as the camera) 2 as the imaging means is installed. On both sides of the camera 2, lighting devices 3A and 3B are provided to illuminate the surface of the lower steel strip 1. The camera 2 can swing its head in a certain angular range θm in the longitudinal direction of the steel strip 1 respectively forward and backward by a mechanism not shown in the figure, whereby with one camera 2, the surface of the necessary range of the steel strip 1 can be photographed. Note that a rotary encoder not shown in the figure is attached to the camera 2 as an angle detector. Also, as shown in Fig. 1, the photographed image of the camera 2 and the angle information of the rotary encoder are sent to the computer 4, and the processing described below is performed. 【0015】 According to the swing angle θ of the camera 2, the distance d from the camera 2 to the steel strip surface changes exponentially as shown in an example in Fig. 2. For this reason, the phenomenon that the nearer ones look larger and the farther ones look smaller becomes more pronounced as the swing angle θ increases. As shown in Fig. 3, at the swing angle of 0°, the photographed image of the steel strip 1 at the distance d0 directly below the camera 2 is a rectangle obtained by cutting out a steel strip 1 with a certain width and a certain length. In contrast, when the swing angle of the camera 2 is θm and the distance to the surface of the steel strip 1 is dm, the steel strip 1 in the photographed image is deformed into a trapezoid in which the overall width shrinks and the side closer to the camera 2 is relatively long and the side farther away is relatively short. 【0016】 Since the width of the linear flaw generated on the surface of the steel strip 1 also changes along with this deformation, in order to accurately detect the flaw width, as shown in Fig. 4, the coordinates (x11, y11) to (x14, y14) of the four points of the trapezoidal figure at the angle θ are calculated from the swing angle information of the camera 2, and a known projective transformation for converting these four points into the four points (x21, y21) to (x24, y24) of the rectangular figure when the camera 2 is facing the surface of the steel strip 1 is performed. 【0017】 In this way, the trapezoidal captured image (Figure 5(1)) is projected (Figure 5(2)) to a rectangular shape as it would be when camera 2 is directly facing the surface of steel piece 1. However, even after projecting and transforming it into a rectangular image, the width of the defect image Pi is still larger than its actual width, and its brightness is reduced. This is because the distance d between camera 2 and the surface of steel piece 1 changes according to the panning angle θ of camera 2, causing the entire captured image of the surface of steel piece 1 to become blurred as it moves out of the camera 2's focus. 【0018】 Therefore, in this embodiment, calibration functions f(θ) and g(θ) are prepared in advance to correct this. An example of the calibration function g(θ) is shown in Figure 6. As shown in Figure 6, the pan-tilt angle θ of the camera 2 and the change in the defect width W due to blurring of the captured image are plotted, and an appropriate approximation formula is defined for each plotted point to form the calibration function g(θ). An example of the calibration function f(θ) is shown in Figure 7. As shown in Figure 7, the pan-tilt angle θ of the camera 2 and the change in the defect brightness ΔL due to blurring of the captured image are plotted, and an appropriate approximation formula is defined for each plotted point to form the calibration function f(θ). 【0019】 The defect image Pi (Figure 5(2)) of the linear defect in the captured image after projection transformation onto a rectangular shape is for a panning angle θ. Therefore, the defect width W and defect brightness ΔL are corrected to those for the panning angle θ when the camera is facing directly, i.e., 0°, by referring to the graph (or table) of the calibration functions f(θ) and g(θ). The harmfulness or harmlessness of the defect is determined by whether or not the corrected defect width Wf and defect brightness ΔLf each exceed predetermined thresholds. 【0020】 (Second Embodiment) Incidentally, the luminance distribution when crossing a linear defect image Pi can be approximated by a normal distribution. For example, by substituting the defect width Wf and defect luminance ΔLf corrected with a calibration function into equation (1) below from the measured values ​​and angle information in Figure 8(1), and drawing a normal distribution curve with a and b as constants (Figure 8(2)), and then drawing a linear defect image Pi with such a luminance cross-sectional distribution (Figure 9), it is possible to intuitively determine whether the defect is harmful or harmless. 【0021】 TIFF0007875434000001.tif29149 [Explanation of Symbols] 【0022】 1...Steel piece (subject), 2...CCD camera (imaging device), 3...Illumination device, 4...Computer.

Claims

[Claim 1] A method for determining the quality of a linear defect by capturing an image of the surface of a subject with an imaging means and determining the quality of the linear defect based on the defect width and brightness of the defect image in the captured image, characterized in that the imaging angle of the imaging means relative to the surface is detected, the trapezoidal distortion of the captured image at that imaging angle is geometrically deformed and corrected into a quadrilateral, and the defect width and brightness of the defect image in the deformed and corrected captured image are corrected using pre-prepared defect width correction values ​​and defect brightness correction values ​​for the imaging angle, respectively. [Claim 2] The method for determining the quality of a linear defect according to claim 1, wherein the defect width correction value and the defect brightness correction value are substituted into a formula for a normal distribution to generate a defect image of the linear defect in which the brightness of the cross-section follows the normal distribution.

Images