JPS6142221B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an automatic inspection method for surface defects in cold- or hot-rolled steel sheets. This method improves the difficulty in evaluating the appropriate defect level using visual inspection, and enables the application of automatic surface inspection equipment in place of visual inspection in the inspection process. The aim is to provide an inspection method.

Surface defects on rolled steel plates such as stainless steel plates, carbon steel plates, various galvanized steel plates, and electrical steel plates are not only a factor that determines the quality of the finished product's surface, but are also inspected during the process to determine the type of defect. Therefore, it is possible to identify the process in which the defect occurs, determine the working conditions for the next process depending on the degree of the defect, and become a factor in determining the destination and use of the material, etc.
Since this is an important point in quality control, careful visual inspection has traditionally been carried out. However, the types and degrees of surface defects on steel sheets vary, and their identification and judgment requires great skill.Also, since this inspection is usually performed while the steel sheet is moving, that is, between runs, it is extremely difficult for the inspection worker to do so. There were problems unique to sensory testing, such as intense concentration and intense fatigue.

For this reason, various methods for automating steel plate surface defect inspection have been developed, and surface defect (or surface flaw) inspection devices are also commercially available. Conventional optical surface inspection equipment mainly consists of a light projector that illuminates the surface of the material to be inspected, and a light that is reflected from the surface of the material to be inspected and that receives information representing the characteristics of defects and converts it into an electrical signal. By processing the electrical signals obtained from the light receiving section and the light receiving section,
It consists of a signal processing section that determines the presence or absence of a defect and the size of the defect. Normally, the inspection is performed by squeezing either the light beam of the light emitter or the field of view of the light receiver (or both) very tightly in order to obtain basic information from changes in light reflectance at minute portions of the surface of the steel plate. The apparatus was configured to measure surface reflection in an extremely narrow range (field of view), and to sequentially move or scan the field of view in order to inspect the entire surface of the steel plate. This scanning method uses the so-called flying spot scanning method, in which the light emitter is shaped like a narrow beam, scans the surface at high speed, and detects the reflected light with a wide-field receiving system, and the light emitter illuminates a wide area uniformly. The typical method is a flying image scanning method in which the light receiving section is used as a micro field of view and scanning this field, and a special method is the flying point flying image method in which both the light emitting section and the light receiving section are scanned, or either one. It is well known that there are fixed types that do not scan either.

The capabilities required for such optical defect inspection equipment are (1) the ability to discover defects, (2) the ability to judge the harmfulness of the discovered defects, that is, the defect level, and (3) the ability to identify the type of defect. This is the main thing.

As mentioned above, in the conventional inspection apparatus of this type, the basis of information is based on the reflectance at minute portions on the surface of the steel plate, and the presence of flaws is estimated from changes in the obtained reflectance signal. Therefore, usually, the output signal of the photodetector is amplified by an appropriate preamplifier, and then either as it is or after extracting only the output change component by a differentiator, the level is discriminated using an appropriate threshold value, so that it can be distinguished from the noise component on the surface. This involved distinguishing between defects and determining the defect level. However, there are many types of surface defects on steel sheets, and their shape, color, frequency of occurrence, etc. vary greatly depending on the type of steel sheet to be inspected and the processing steps up to the inspection line. In particular, the criteria for determining the defect level, which is the ultimate purpose of inspection, differs for each type of flaw depending on the purpose of use of the steel plate, and since standards established based on visual judgment are usually used, the criteria for determining the defect level differ depending on the type of flaw. Since the surface reflectance changes of the defects are different from each other, it has been extremely difficult to quantitatively determine the defect level of various types of defects.
For example, if the level setting value of a determination circuit based on a certain type of defect is used to determine the level of other types of defects, there is a very high possibility of an error.

The present invention has been made for the purpose of accurately determining the defect level, and the details thereof will be described below.

As already mentioned, in order to determine the level of surface defects, it is important to know the type of the defects. Defect types are empirically classified based on the nature and cause of the defect, and there are many types, but when the purpose is to determine the defect level, similar types are grouped together into several groups as described below. Is possible. It is clear that if it is possible to identify which group a defect belongs to, the determination can be made for each group, making it relatively easy to determine the defect level. In addition, in distinguishing between these types of defects, the shape, color, location, etc. of defects are visually observed, and judgments are made in an instant by empirically examining the shape, color, and location of the defects, and this function is in the field of so-called pattern recognition. Therefore, unlike conventional optical inspection methods, these flaws cannot be detected using only partial reflection characteristics, but instead can be detected as a flat collection of flaws.
That is, it is necessary to recognize it as an image, and in this way, it becomes possible to determine the type of flaw in a way similar to visual inspection.
The present invention detects defects as two-dimensional images and performs signal processing instead of the conventional one-dimensional method of detecting defects, in order to provide a defect inspection apparatus with performance close to that of visual inspection.

There are a wide variety of defects that occur on the surface of steel sheets, but they can be classified from several viewpoints.
For example, defects can be classified into collective defects and single defects based on their density. The former is a collection of minute defects such as rough scale defects on hot-rolled steel sheets, and although it is not a problem when looking at individual defects, the characteristics of the collection, such as the density and breadth of the collection, are defects. It is like determining the degree of The latter is a case where even if the defects are densely packed to some extent, the degree of crowding does not matter, but is determined by the largest (most influential) defect contained therein.
Generally, such flaws often exist singly. Other types of flaws include oyster flaws, which consist of multiple deep roots, and misalignment flaws, where many short oyster-like flaws line up in a row in the rolling direction. It can be considered as an intermediate defect or as a collective defect. Whether a defect is treated as a collective defect or as an individual defect is usually determined by the type or use of the steel sheet, that is, by the purpose of the inspection.

Another type of defect classification is classification by defect shape, which is applied to single defects. In other words, linear defects where the defects are long in the rolling direction of the steel plate and short in the width direction, widthwise defects which are spread in the width direction and short in the length direction, dotted defects which are short on both sides, and long or short defects on both sides. This includes planar defects that have a large area. Further, these defects may be collected to form one collective defect. The characteristics of the defects described above are all in the planar (two-dimensional) distribution state of optical reflection characteristics at the defective portion. In other words, it is possible to identify the type of defect by extracting the light reflection characteristics of the defective part in a two-dimensional plane. It is easy to judge the degree (level).

In the present invention, based on the characteristics of the above defects,
Detects the defect image two-dimensionally and performs signal processing,
The defect type is determined by extracting the characteristics of the defect, and the two-dimensional Fourier transform method is used as the signal processing method to determine the defect type.
A power spectrum in a two-dimensional frequency space is obtained from information regarding periodicity and directionality, and the defect type is identified from the power spectrum pattern in the two-dimensional frequency space. The details of the present invention will be explained below based on the drawings.

Figures 1 and 2 show the application of the present invention to the surface inspection of stainless steel plates at the exit side of a continuous annealing and pickling line, where various surface defect images of hot-rolled coils are subjected to two-dimensional Fourier transformation, and the power spectrum is The results of measuring the two-dimensional distribution are shown. The defect in Figure 1 has power almost independently in the rolling direction and width direction,
There are few correlated components. That is, on the two-dimensional power spectrum plane, most of the power is on straight line 1, where the frequency Fr in the rolling direction is zero, and on straight line 2, where the frequency Fw in the width direction is zero, which indicates that the size of the defect is This indicates the fact that a large number of defects have no periodicity in both the width direction and the rolling direction, and are scattered dots or island-like defects. FIG. 2 shows that the power spectrum has periodicity in the rolling direction and is a defect with no periodicity in the width direction.

Moreover, the characteristics of a defect can also be determined by the ratio of the energy E within a specific frequency band to the total on the two linear axes. That is, for example, scale defects,
A collection of fine defects has energy concentrated in a relatively high frequency component, while a defect with thick linear defects such as oyster scratches has energy concentrated in a low frequency band in the board width direction. Therefore, by looking at the ratio of energy in a relatively high frequency band to the total energy and the ratio of energy in a relatively low frequency band to the total energy on the frequency axis 1 in the board width direction of the power spectrum, it is possible to detect scale flaws and cracks. It becomes possible to extract the characteristics of the flaw.

As parameters for identifying types of surface flaws, Japanese Patent Application No. 52-142516 (Japanese Unexamined Patent Publication No. 54-74793) filed at the same time.
No.) As detailed in ``Steel plate surface defect inspection method'', it is convenient to use DFw, DFr and Rw, Rr defined by the following formulas.

Parameters DFw and DFr indicate the ratio of each frequency component in the rolling direction (indicated by subscript r) and the width direction (indicated by subscript w), and if normal, DFw = 0, DFr =
However, when the power spectrum in the rolling direction or the width direction is stronger than the other power spectrum, DFw<0 and DFr<0 depending on the degree. parameters
Rw and Rr indicate the strength of the power spectrum in each direction of width and rolling, Rw<(35~)40, Rr
<(35~)40 is normal, Rw(35~)40, Rr
(35~) If it is 40, it is abnormal, that is, there is a flaw. Note that the determination values 35 to 40 are determined experimentally using sample data or online data.

In addition, the width (length in the sheet width direction), length (length in the rolling direction), and area (percentage of the entire image) of the defect image captured as an image also have characteristics, and these can be combined with the power spectrum described above. Identification of specific defect types is possible. FIG. 3 shows a procedure for identifying surface defects using these parameters.

In FIG. 3, numerals 31 to 35 are logic decision blocks, which pass if true and branch if false. In these blocks 31 is DFw
= 0 or not, that is, whether the change component in the sheet width direction axis is large. 32 is Rw40 or not, that is, whether the defect characteristic frequency characteristic band component is large in the sheet width direction. 33 is DFr = 0 or not. In other words, whether the change component in the rolling direction axis is large or not, 34 is Rr40 or not, that is, whether the defect characteristic frequency band component in the rolling direction is large or not, and 35 is the logical judgment criterion whether the defect area is large or not. ing. Due to this flow, defect A (rough skin scale), defect B (oyster flaw), defect C (co-slip flaw), defect A', A'' (rough skin scale flaw that is weak in the width direction), defect B', B'' (weak scratches in the width direction), defects D, E (undefined defects) 5
It is possible to determine seed defects.

90% of natural defects that normally occur with this method
It is possible to have the ability to automatically identify the above defect types, and represents a significant improvement over the prior art. Furthermore, in the method of the present invention, information is added regarding the position where the defect occurs, that is, whether the defect occurs at the tip, middle, or tail in the longitudinal direction of the strip, or whether it is at the edge or the center in the width direction. , it is possible to improve identification accuracy.

FIG. 4 shows a basic configuration diagram of a defect inspection device based on the present invention. In the figure, 41 is a steel plate which is a material to be inspected, and 42 indicates defects occurring on this surface. 43 is a light source for illumination. 44 is an optical system that converts the reflected light from the flaw into an image such as a light image;
45 is an imaging section that converts the image converted by the optical system 44 into an electrical signal; 46 is a two-dimensional signal processing section that processes the obtained image information to extract defect characteristics; and 47 is a two-dimensional signal processing section. An identification/determination unit 48 for identifying feature parameters DFw, DFr, Rw, and Rr for determining the type of flaw from the processed information obtained by the dimensional signal processing unit 46 determines the size of each of the obtained feature parameters. A level determination unit 49 determines the type and extent of defects according to the flow shown in FIG. 3 based on the level correlation of each feature parameter; This is an output display device that outputs defect information DI to an inspection information processing device.

In FIG. 4, an image of the defect is focused on an imaging section 45 by an optical system 44, converted into an electrical signal 4a, and sent to a signal processing section 46. In the signal processing section 46, the defect image information is subjected to necessary preprocessing such as removal of average values and trends, and then subjected to two-dimensional Fourier transform processing. The output 4b of the signal processing unit 46 is a frequency spatial image corresponding to the defect image, and preselected feature parameters DFw, DFr, Rw, and Rr are calculated based on this two-dimensional spectral distribution pattern, and a signal 4c of the magnitude is calculated. Each level discrimination section 48
Output to. The level determination unit 48 performs level determination for each characteristic parameter, and sends the signal 4d with several levels of determination to the defect determination unit 49. The defect determining section 49 determines the type and size of the defect from the correlation between the levels of each parameter, and the result 4e is outputted to the outside by the display output section 50.

The selection of the characteristic parameters and the level setting in the parameter determination section differ depending on the steel plate manufacturing process to which the defect inspection is applied. In addition, as for signal processing methods, in addition to two-dimensional Fourier transform, there are also two-dimensional Hadamard transform, two-dimensional correlation function, etc., so it is effective to use two or more of these signal processing methods together, depending on the object to be inspected. Just choose an appropriate method.

An example of commonly used image-to-electrical signal converters is an image pickup tube such as a vidicon used in a television camera, or an image conversion device using a solid-state image sensor that has a large array of photodiodes, charge transfer devices (CCD), etc. Resolution is limited, usually 1
It is only possible to break down the screen into 500 x 500 pixels, or at most 1000 x 1000 pixels. Therefore, when the width of a cold-rolled steel plate reaches approximately 1,500 mm, viewing the entire width on one screen may overlook minute defects of several mm or less, or it may be difficult to perform the above-mentioned analysis on defects. Since the numbers of pixels do not correspond, an incorrect conclusion will be obtained. For this reason, the size of the viewing area (hereinafter referred to as the field of view) of the imaging unit (image converter) when it is stationary is limited to a maximum of about 100 mm x 100 mm. Therefore, in order to inspect the entire width of a steel plate, it is necessary to sequentially move (scan) the field of view in the width direction of the steel plate or to arrange a plurality of imaging units in the width direction of the steel plate so that the field of view covers the entire width. An example of a field scanning mechanism for this purpose is shown in FIG.

In FIG. 5, 56 is a rotating mirror, 57 is a drive motor, and 58 is an auxiliary mirror for projecting the light beam from the rotating mirror 56 onto the stationary optical system 54. Due to the rotation of the rotating mirror 56, the field of view of the imaging unit 55 changes to the steel plate 51.
The entire surface of the plate can be inspected by running the steel plate 51 (in a direction perpendicular to the width direction of the plate).
In these scanning mechanisms, it is important to form a good image of the defect on the imaging section over the entire scanning width. For this reason, it is necessary to appropriately design the entire optical system including the auxiliary mirror 58, rotating mirror 56, and optical system 54, and by making the scanning width relatively small, it is easy to obtain a good image that does not cause any practical problems. can be obtained. Therefore, in order to inspect wide steel plates, the above scanning width should be limited to about 200 mm to 500 mm.
By arranging a plurality of these in parallel in the width direction of the base board, it is possible to inspect the entire board width.

In the signal processing unit 46 of FIG. 4, the image signal 4a is normally converted into digital information and then stored in a storage device (not shown), and the signal processing is performed digitally (this processing is usually performed using a small electronic computer). (performed by the system). When carrying out a running inspection of a steel plate, images of the surface of the steel plate including defective parts are sent to the signal processing section 47 every moment. For this reason, since the amount of digital signal processing for the defect image is quite large, it is difficult in terms of time to process all the images of the entire surface of the board, and it is preferable to make this possible because it increases the cost of the inspection equipment. do not have. For this reason, it is preferable that the signal processing is performed only on the image of the defective portion on the steel plate. For this reason, it is necessary to pre-select whether or not to perform signal processing on the steel plate surface image obtained on the imaging unit 45, and it is also necessary to accumulate the selected image information. For this preliminary selection, it is effective to use a conventional optical inspection device in combination. An example of this is shown in FIG.

Figure 6 shows an example of an optical system similar to Figure 5,
In this figure, 69 is a semi-transmissive mirror, which guides a part of the light reflected from the steel plate surface going to the image converters 64 and 65 to the photoelectric conversion element 70 through the optical system 68, and inputs this output signal to the defect selection signal processing unit 67. Then, it is determined whether or not the digital signal processing is necessary.
The signal processing unit 67 performs determination based on the change in the amount of reflected light at the defective portion or its differential value, and has a function similar to the signal processing in a conventionally known surface defect inspection device. 61 is the material to be inspected, 62 is the defect, 6
3 is a projection light, and 66 is a rotating mirror.

The present invention can also be implemented with a configuration different from that shown in FIG. That is, the two-dimensional signal processing unit 46 shown in FIG. 4 typically employs digital signal processing to enable a variety of processing, but in this case, processing time is required and processing delays may occur on ultra-high-speed lines. There is sex. In such a case, part of the digital processing can be performed optically, thereby reducing processing time. FIG. 7 shows an example in which optical two-dimensional Fourier transformation is adopted. In this figure, 73 is a coherent light source typified by a laser, etc., and the illumination light source 4 in FIG.
Install in place of 3. It is well known that the diffraction pattern of reflected light on the surface of a steel plate irradiated with coherent light is a Fourier transform of surface reflection characteristics, and its intensity distribution becomes a two-dimensional power spectrum. That is, the steel plate 7 is placed on the imaging surface 78 placed on the reflected optical path.
A power spectrum image of the defect 72 on 1 is obtained. Therefore, the image converter 75 converts the power spectral image into an electrical signal, and the subsequent signal processing can be simplified.

As described above, according to the present invention, it is possible to distinguish the types of surface defects, which could not be achieved with conventional surface defect inspection equipment, and the accuracy of defect level determination is improved.In addition, the classification information for each defect type allows for quality control in pre- and post-processes. It has become possible to fully automate surface flaw inspection instead of conventional visual inspection.

[Brief explanation of the drawing]

Figures 1 and 2 are graphs showing actual measurement examples of power spectrum distributions, Figure 3 is a flowchart showing procedures for determining defect types, and Figures 4 to 7 are explanations showing an overview of the defect inspection apparatus according to the present invention. It is a diagram. In the drawing, 41, 51, 61, 71 are steel plates, 42,
62 and 72 are defects, and 43, 63, and 73 are light sources.

Claims (1)

[Claims] 1. Light is projected onto the surface of the steel sheet to be inspected, and the reflected light is focused to perform photoelectric conversion and two-dimensional Fourier transformation, and the characteristics of defects are detected on the surface by taking into consideration the power spectrum distribution in the frequency space. The parameter is extracted as a parameter for identifying the defect type, and the type of defect is identified based on this.
1. A method for inspecting surface defects on a steel sheet, characterized in that a threshold level for a defect signal is determined according to the identified defect type to determine the degree of harmfulness of the defect. 2. Claim No. 2, characterized in that non-coherent light is projected onto the surface of a steel plate, the reflected light is imaged on an imaging section, photoelectrically converted, and two-dimensional Fourier transform is performed on the obtained electrical image signal. The method for inspecting surface defects of a steel plate according to item 1. 3. A surface defect on a steel plate according to claim 1, characterized in that coherent light is projected onto the surface of the steel plate and a two-dimensional Fourier transform image obtained on an imaging plane placed on a reflected optical path is photoelectrically converted. Inspection method.