JPH10121368A - Inspection apparatus for woven fabric - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection apparatus of a woven fabric, capable of simultaneously detecting abnormalities of warps and wefts under the same optical condition without depending on weave structure and automatically inspecting at a low cost. SOLUTION: This inspection apparatus of woven fabric is used for imaging woven fabric and carrying out inspection of woven fabric based on image data of the woven fabric and contains CPU 17 for calculating structure cycle of woven fabric from image data of woven fabric, setting a comparison region of image data based on the structure cycle of the woven fabric and detecting defect based on an amount of statistics of image data in a comparison region extracted by a characteristic amount extracting circuit 12.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a woven cloth inspection apparatus, and more particularly to a woven cloth inspection apparatus for automatically inspecting a woven cloth during weaving or a woven cloth for defects.

[0002]

2. Description of the Related Art Conventionally, as an apparatus or a method for inspecting the appearance of a woven fabric, an image of the surface of the woven fabric is taken by a camera, and image density data obtained from the taken image is compared with a threshold value to detect abnormalities in the appearance. An automatic inspection system that irradiates a woven cloth with laser light, receives the reflected light or transmitted light with a light-receiving element, compares the level of the received light amount with a threshold value, and detects an abnormality Methods and the like are known. However, the defect that can be detected by the inspection apparatus or the inspection method is limited to a relatively large defect such as a thread drop that can be easily determined visually by a person. There is also a problem that the detection accuracy is greatly reduced due to vibration, disturbance, or the like.

As a method for improving the detection accuracy, there is an invention disclosed in Japanese Patent Application Laid-Open No. 4-148852, for example. According to the present invention, light transmitted from a light source irradiating a woven fabric is received by a light receiving element through an optical slit arranged in a yarn direction of an inspection target, and abnormality is detected by comparing a received light waveform with a reference waveform. A method for doing so is disclosed. In this method, the presence or absence of a defect is determined based on the characteristic amount of the light transmitting portion, that is, the extracted woven fabric opening.

Japanese Patent Application Laid-Open No. 3-249243 discloses a method in which two pairs of known comb-shaped light-receiving sensors are used, and an abnormality is detected by comparing a difference between the outputs of the two with a preset threshold value. Have been. In this method, since the density information obtained by dividing the woven cloth narrow region into two is reflected on the two pairs of comb-shaped light receiving sensors, even if there is vibration or disturbance light, there is an effect that the difference value output between the two cancels out.

[0005] In order to ensure the same accuracy as the inspection performed by visual inspection, it is necessary to extract the characteristic of the yarn component itself in addition to the conventional characteristic inspection of the woven fabric opening. For example,
Visual inspection for defects such as warp pouring defects
The presence / absence of a defect is determined from the disorder of the periodicity of the vertical relationship of the yarn to be inspected on the yarn interlacing point. Therefore, in order to mechanically detect this defect, it is only necessary to extract only the yarn component in which the warp is above or below the weft on the yarn interlacing point coordinates. In order to detect this defect by image processing, binarization processing of the grayscale image is indispensable. However, in the conventional fixed binarization method, extraction of only the warp component above or below the weft is not possible in the target portion. This was not possible because the concentration was not uniform.

The applicant of the present invention discloses a woven cloth inspection apparatus which solves the above-mentioned problems in Japanese Patent Application No. 7-198171. According to the present invention, the light illuminated on the woven
An image is taken by a CD (Charge Coupled Device) element, and woven fabric information (yarn density, yarn inclination, woven structure, etc.) is calculated based on the image data obtained by the imaging, and is obtained from this woven fabric information. By comparing the correlation value over the entire image data between the region having the thread pitch width and the other region located at an integer multiple of this thread pitch width with the set threshold value, In this method, regions having different structures are detected with high accuracy over the entire width of the woven fabric under the same optical conditions without distinguishing between warps and wefts.

[0007]

SUMMARY OF THE INVENTION The above-mentioned JP-A-4-14-1
In the invention disclosed in Japanese Patent No. 48852, for example, in the case of a defect in which the opening portion of the woven fabric is not so different from a non-defective product, such as a defect in pouring a warp, the defect cannot be detected, and the detection accuracy is significantly reduced. There was a point. Further, in this method, it is assumed that the weave density is constant and the optical slit and the yarn in the inspection target direction are parallel. However, there are various woven densities of actual woven fabrics, and there is a problem that the optical slit needs to be replaced each time. Also, the actual woven yarn, especially the warp,
There is also a problem that the above conditions cannot be maintained due to the curvature at both sides of the woven fabric, and the detection accuracy is reduced.

In the invention disclosed in Japanese Patent Application Laid-Open No. 3-249243, similarly to the invention disclosed in Japanese Patent Application Laid-Open No. 4-148852, there are limitations on the defects that can be extracted and the density of the woven fabric changes. If the parallelism between the comb-shaped light receiving sensor and the yarn to be inspected cannot be maintained, there is a problem that the detection accuracy is reduced.

Further, both of the above-mentioned inventions have a problem that the same sensor cannot simultaneously detect the abnormality of the warp and the weft.

Further, in the invention disclosed in Japanese Patent Application No. Hei 7-198171, it has been recognized that there is little effect on defects of woven fabrics having different woven structures such as satin weave and twill weave other than plain weave. .

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the invention described in the claims is to simultaneously raise abnormalities of warp and weft simultaneously under the same optical conditions irrespective of the weaving structure. An object of the present invention is to provide a woven cloth inspection device capable of detecting with high accuracy and automatically inspecting at low cost.

[0012]

According to a first aspect of the present invention, there is provided a woven cloth inspection apparatus for imaging a woven cloth and inspecting the woven cloth based on image data of the woven cloth. A tissue cycle calculating means for calculating a tissue cycle of the woven fabric from the image data of the woven fabric; a comparison area setting means for setting a comparison area of the image data based on the tissue cycle; Defect extracting means for extracting a statistic from the image data and extracting a defect based on the statistic.

By calculating the texture period of the woven fabric and setting the comparison area of the image data based on the texture period, abnormalities of a plurality of types of woven fabrics having different woven densities, woven textures, etc. can be detected by the warp or weft. Without distinction, it is possible to detect an abnormality with high accuracy over the entire width of the woven fabric.

[0014]

FIG. 1 is a schematic block diagram of an inspection apparatus according to an embodiment of the present invention. The inspection apparatus includes a camera lens 3, a CCD camera 4, an A / D converter 5, a frame memory 6, a pre-processing circuit 7 for performing pre-processing such as differential emphasis, an FFT (Fast Fourier Transform) circuit 8, an inspection object. A density projection circuit 9 for density-projecting the yarn onto the coordinates of the yarn to be inspected, a binarization circuit 10 for binarizing the grayscale image, and integration of binding information for labeling the binary image in the rectangular area Circuit 11, a feature amount extraction circuit 12 for extracting various feature amounts of a labeled binary image, a filter circuit 13 for cutting an intermediate frequency region, and a logic operation on image data Image logic operation circuit 14, image bus 15, CPU bus 16, C
PU (Central Proccessing Unit) 17, ROM (Read
Only Memory) 18, RAM (Random Access Memory)
19 and input / output device 2 such as keyboard and display
Contains 0.

The light emitted from the light source 1 passes through the gap between the woven fabrics 2, is condensed by the camera lens 3,
An image is formed on a CD element. The light source 1 may be either a transmission type or a reflection type. Is preferred. When an area-type CCD camera is used as the light source 1, scattered light having a uniform surface illuminance is preferable. When a line-type CCD camera is used, the type of the light source 1 is a semiconductor laser, H
A light source that emits eNe laser or the like in a slit shape using a lens, a light source that emits halogen light from the side of the rod lens, and is irradiated in a slit shape by a special slit-shaped scattering paint provided on the rod, or a slit. Fiber optic illumination may be used. Among them, a light source using a rod is the most preferable because it has uniform orientation characteristics in the width direction and can obtain high luminance. In this embodiment, an example in which an area type CCD camera is used will be described. If the line type CCD camera is used on the premise that the woven fabric is run at a low speed, the A / D converter 5 is used as shown in FIG.
One-dimensional / two-dimensional converter 30 for converting one-dimensional image data into two-dimensional image data between the image data and the frame memory 6
Should be added.

The amount of light transmitted through the woven fabric is not particularly limited, but there is no particular problem as long as the amount of light is sufficient to accumulate sufficient charges within the CCD update cycle. The magnification of the camera lens 3 is determined based on the finest weaving density among the weaving densities of the product. In general, there is a tendency that the higher the woven cloth image magnification ratio, the easier the defect extraction is. However, as will be described later, since the size of the rectangular area from which the statistic is extracted needs to be adjusted to the weave tissue cycle, the number of yarns in the inspection target direction in the captured image is at least twice the number of yarns in the tissue cycle. It is assumed that there is.

The grayscale image data picked up by the CCD camera 4 is converted into 8-bit digital image data by an A / D converter 5 and then stored in a frame memory 6. As described above, when the CCD camera 4 is of a line type, the one-dimensional / two-dimensional converter 30 converts the one-dimensional image data into two.
The image data is converted into two-dimensional image data and stored in the frame memory 6. The stored image data is processed by the preprocessing circuit 7.
In order to effectively extract defects, pre-processing such as differential enhancement for enhancing the contour is performed. From the preprocessed image, the weave tissue period is determined in order to set the size of a rectangular area from which a statistic described later is extracted. This can be realized by the following method.

The short axis of the rectangular area is set in a direction perpendicular to the yarn to be inspected. The width of this short axis is set to the smallest value (in the woven cloth image captured under the same optical conditions,
While sequentially increasing the number of pixels in the short axis direction of the comparison area from the number of pixels corresponding to the pitch size of the yarn having the highest density of the inspection target yarn, a pair of comparison areas (two of the same size set in different areas) The correlation value of the image data in the comparison area) is obtained. It was found that the width of the short axis which became the maximum value in the obtained correlation values coincided with the weave tissue period. The width of the short axis is defined as the short axis size of the comparison area. Note that the method of obtaining the tissue cycle may be performed after conversion into a binary image instead of a grayscale image. Further, for example, a method of extracting a characteristic of a density waveform in a direction perpendicular to a yarn to be inspected to obtain a periodicity,
Although it is possible to calculate the weave tissue period by a method for obtaining the periodicity in the FT, the method is not particularly limited.

Next, in order to automatically calculate the average yarn pitch and the average yarn inclination amount in the inspection target yarn direction, the density projection circuit 9 performs density projection (density addition processing) on the inspection target yarn direction coordinates and performs one-dimensional processing. Generate density data. Fourier transform is performed on the density data by the FFT circuit 8,
Real number data and imaginary number data of the spectrum mode are obtained. From the real number data and the imaginary number data, the average yarn pitch (weaving density) and the amount of inclination in the inspection target yarn direction are obtained.
In addition, in order to obtain the amount of inclination, the above-described processing is performed by setting the image area to at least two divisions for the yarn in the inspection target direction, and based on each obtained imaginary data, that is, data of the phase component. The amount of inclination is obtained by an averaging method, a least square method, or the like.

The fast Fourier transform does not need to be performed at all times, and can be performed by software processing by the CPU 17. Further, this method has an advantage that it can be obtained with high accuracy even when defect information such as thread dropout is included in image data.

Other methods of calculating the yarn direction include, for example, a method of differentiating the vertical direction of the yarn to be inspected to emphasize its contour, and then following the peak value of the axial waveform, or a method of arranging substantially linearly. There is a method of obtaining the yarn direction by Hough transform for determining a straight line passing through as many of these points as possible from a large number of point sequences, but the method is not particularly limited.

Next, a binarizing circuit 10 for the image data
Performs a binarization process. The binarization is performed to extract the characteristic amount of the yarn component in the inspection direction. Especially,
In the case of warp, it is necessary to extract only the components above or below the point of intertwining with the weft. This processing will be described with reference to FIG.

First, density data in the X-axis direction of image data at a cut portion of a woven fabric is sequentially extracted to generate a raw waveform. Then, the filter waveform is generated by passing the raw waveform through the filter circuit 13. Filter circuit 13
Is a band elimination filter for cutting an intermediate frequency region including a previously determined periodic component in the yarn direction to be inspected. A threshold value is obtained by adding a slight offset value to the filter waveform obtained through the filter circuit 13. Then, the image data at the cutting section is binarized depending on whether or not the raw waveform is larger than the threshold value. This process is repeated while scanning the cutting section in the Y-axis direction to generate a first binary image. Similarly, a second binary image is generated by taking the cut portion in the Y-axis direction, binarizing the image data at the cut portion, and repeating the scan while scanning in the X-axis direction.

Since the first binary image has been subjected to the filtering process in the warp direction and the vertical direction, the first binary image is a binary image of only the warp component on the weft on the interlacing point. Is done. Further, as the second binary image, a binary image of only the weft component including the upper and lower portions of the warp on the intertwined point is generated. The above-described binarization method has an additional advantage that only the thread information can be reliably extracted even if there is variation in illumination, uneven illuminance, uneven weaving density of the woven fabric, or a difference in the level of yarn.

FIGS. 3A and 3B show an example of an image after binarization. The above-described first binary image and second binary image
If the rectangular area processing described later is performed based on the value image, defect information of the woven fabric can be obtained, but the binary image in the warp direction also includes an intermediate image in which the warp passes under the weft from above the weft. It is. Since this information is unnecessary for defect extraction of the woven fabric, in the case of the warp inspection, the first binary image and the second binary image are used.
By performing an AND logic operation on the value image and the image logic operation circuit 14, it is possible to convert only the warp image on the weft on the complete confounding point. An example of the converted binary image is shown in FIG.

A rectangular area for automatically extracting a statistic from the obtained texture period and the yarn direction in the yarn direction to be inspected is automatically generated. FIG. 4 shows an example of a rectangular region when the weave structure is plain weave.
(A) and (b). Although the length of the rectangular area in the major axis direction is not particularly limited, the accuracy of defect detection tends to be improved as the length is set longer for continuously occurring defects such as thread dropout. In addition, the shorter the defect due to fluff or the like, the higher the accuracy. The length in the major axis direction may be determined according to the characteristics of the woven fabric to be inspected, and the same process may be performed a plurality of times while varying the length during the inspection.

A rectangular area is set by a method of setting a pair of adjacent rectangular areas (area 1 = one) as shown in FIGS.
A × H, region 2 = B × H).
That is, as shown in FIG. 4C, the rectangular areas to be compared are alternately changed (area 1 = A × H + C × H, area 2 = B × H + D
× H). However, it is premised that the size (the number of pixels) of the rectangular area in the short axis direction matches the weaving period of the yarn in the inspection target yarn direction. By setting the rectangular area in this manner, the same inspection can be performed regardless of the structure pattern of the woven fabric. However, if the phase of the long axis direction of the rectangular area is out of phase with the direction of the yarn to be inspected, the accuracy of defect detection is reduced. To avoid this problem, the direction of the yarn to be inspected is automatically calculated as described above, and the longitudinal direction of the rectangular region is made to follow the direction of the target yarn, or the image is rotated by the amount of inclination of the target yarn, and the rectangular region is rotated. Align the long axis of the target with the direction of the yarn to be inspected. As a result, it is possible to avoid a decrease in defect detection accuracy due to, for example, an inclination of the warp that occurs on both sides of the woven fabric during weaving and an error in centering when the sensor is fixed.

In addition, in the case of inspecting woven fabrics having different woven densities, the average density of the yarn to be inspected is calculated, and the long axis of the rectangular area for calculating the statistical value is automatically optimized. This has the advantage that the same inspection can be performed without any adjustment of the optical conditions. Furthermore, even if the light amount of the light source relatively decreases or increases during the inspection, for example, during the process of comparing the statistics between adjacent rectangular areas, these effects are advantageously offset. However, in the case where light causing halation exists nearby, the shielding plate 21 shown in FIG. 1 may be provided.

The combination information integration processing circuit 11 labels a binary image in a rectangular area. For example, FIG.
In the case of the binary image shown in (c), a label is attached to each of the binary images extracted as the warp components on the weft on the interlaced point. Then, the feature amount extraction circuit 12
The feature amount (statistical amount) of the binary image in the rectangular area is extracted.

The statistics in the rectangular area include the total number of black or white pixels, the total number of labels, the area, fillet diameter, shape ratio, and principal axis angle, which are the characteristic amounts of each labeled binary image. ,
The maximum value, the average value, or the minimum value of the perimeter, the distance between the centers of gravity of the adjacent labels, and the like. The defect is extracted by comparing the difference value of these statistics, the image pattern correlation or the statistics with the reference data. . However, the image pattern correlation may be performed not only on the binary image but also on the original grayscale image. Further, the fillet diameter means the length of the projection part in the X-axis direction and the length of the projection part in the Y-axis direction of a group of labeled binary images.

FIGS. 5A and 5B are diagrams showing an example of defect extraction. (A) shows normal woven tissue,
(B) shows an example of two crossover defects. Although it is difficult to extract such a defect by comparing the conventional woven fabric openings, as shown in FIG. 5, when the binary images obtained by the above-described binarization processing are compared in a rectangular area, (b) Are completely different in the position pattern consistency in the left and right rectangular regions. That is, the position patterns of the warp components (represented by black squares in FIG. 5) on the weft on the intertwining points are completely different. Therefore, it can be seen that by calculating the pattern consistency between a pair of rectangular regions, a defect in the woven tissue can be easily extracted. Note that the CPU 17 performs defect extraction based on the statistics described above. The calculation result is input / output device 20
Is output by a display or the like.

When the woven fabric during weaving is inspected in-line, foreign matter such as fly cotton may adhere to the surface of the woven fabric. As described in the related art, when comparing the presence or absence of a defect based on the feature amount of the woven fabric opening, or when trying to extract the defect only by comparing the density data in two pairs of neighboring areas,
These foreign substances are erroneously determined as defects. In order to avoid this problem, the above-described comprehensive comparison of the statistics and comparison between the statistics and the reference value are performed. For example, in a transmission inspection,
In the case where there is a foreign matter, when the warp defect caused by the wrong lead and the defect caused by the foreign matter are compared between the above-described rectangular areas, the average density value, which is the average value of the density data in the rectangular area, is clearly different. Therefore, by adding such a statistical comparison to the determination in addition to the conventional calculation method, it is possible to separate a defect from a disturbance element such as a foreign matter. The statistic serving as a parameter at the time of the determination shown here is not particularly limited, and a statistic indicating a unique feature of the target defect is obtained in advance by an experiment or the like, or a histogram is created in-line. The feature of the target defect may be obtained from the histogram, or may be processed in combination.
In addition, since the present method can simultaneously extract abundant statistics, it is also possible to identify defects by, for example, fuzzy inference or multiple regression analysis.

FIG. 6 is a flowchart showing a processing procedure of the woven cloth inspection device in this embodiment. First, C
The grayscale image data captured by the CD camera 4 is A
After being converted into 8-bit digital image data by the / D converter 5, it is taken into the frame memory 6. At this time, if the CCD camera 4 is a line type, the one-dimensional / two-dimensional converter 30 converts the one-dimensional image data into two-dimensional image data, and then takes in the frame memory 6 (S1).

Next, the pre-processing circuit 7 includes the frame memory 6
Is read out and subjected to pre-processing such as differential enhancement for enhancing the outline of the image (S
2). The CPU 17 extracts the texture period of the inspection target yarn based on the pre-processed gray image data by the method described above. Further, after the density image data is density-projected by the density projection circuit 9, the thread direction is calculated by performing a Fourier transform on the density-projected image data by the FFT circuit 8 (S3).

The binarizing circuit 10 performs a filtering process on the grayscale image data, thereby forming a binary image (first binary image) of only the warp component on the weft on the interlaced point.
Is generated (S4), and a binary image (second binary image) of only the weft components including the upper and lower portions of the warp on the interlaced point is generated (S5).

The image logical operation circuit 14 performs an operation on the first binary image and the second binary image, and outputs a binary image of only the warp component on the weft on the interlaced point or the binary image on the interlaced point. 3rd binary image of only the weft component above the warp
A value image is generated (S6).

Next, the CPU 17 sets a rectangular area in the third binary image based on the tissue cycle in the thread direction to be inspected and the thread direction (S7). The feature amount extraction circuit 12 extracts various statistics from the binary image data in the set rectangular area (S
8). When a pair of rectangular regions is set, the presence or absence of a defect is determined by comparing various statistics of each rectangular region. If a reference value has been set in advance, the presence or absence of a defect is determined by comparing various statistical quantities of the rectangular area with the reference value.

In step S1, it is determined whether or not inspection has been performed on all of the captured image data. If the inspection is not completed (S10, No), the third
The rectangular area set in the binary image is shifted in the direction perpendicular to the direction of the yarn to be inspected (S12). Then, the processing of steps S8 and S9 is repeated. If the inspection has been completed (S10, Yes), it is determined whether or not to inspect the yarn to be inspected in another direction. When performing an inspection (S11,
In Yes), the processes in and after step S7 are repeated. When the inspection is not performed (S11, No), the inspection result is output to a display or the like of the input / output device 20 (S1).
3).

The program for performing the overall control is as follows:
It is stored in the ROM 18. When inspecting the entire width of the woven fabric, the sensors may be traversed in the woven fabric width direction or a plurality of sensors may be arranged at equal intervals in the woven fabric width direction.

In the above description, the processing relating to the in-line inspection of the woven fabric during weaving has been described, but the present invention can also be applied to automatic inspection of a woven fabric that has been woven. Further, the present invention can be applied to a sheet having regular characteristics other than the woven fabric.

FIG. 7 is a diagram showing a case where the inspection apparatus according to the present embodiment is applied to a loom. The light source 1 is irradiated from below onto the woven fabric 2 during weaving, and an image is taken by the CCD camera 4. The CCD camera 4 is mounted so as to be movable along a movement axis 40.

As described above, according to the present invention, it has become possible to reliably detect a defect having a different woven texture, which has been difficult to detect. Further, even if the weaving density changes or the yarn direction of the inspection object changes, the defect can be extracted simultaneously and with high accuracy without changing the conditions of the optical system at all. Further, by extracting abundant statistics, it is possible to separate a disturbance element such as a surface deposit from a defect, thereby enabling more accurate discrimination.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of a woven cloth inspection device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of a raw waveform and a filter waveform for explaining a binarization process.

FIG. 3 is an example of a binary image for describing a binarization process and a logical operation process.

FIG. 4 is a diagram for explaining a rectangular area set for extracting a statistic;

FIG. 5 is a diagram showing an example of a normal woven structure and a defective woven structure.

FIG. 6 is a flowchart showing a processing procedure of the woven cloth inspection device according to the embodiment of the present invention.

FIG. 7 is a diagram illustrating a case where the inspection apparatus according to the present embodiment is applied to a loom;

[Explanation of symbols]

 Reference Signs List 1 light source 2 woven fabric during weaving 3 optical lens 4 CCD camera 5 A / D converter 6 frame memory 7 preprocessing circuit 8 FFT circuit 9 density projection circuit 10 binarization circuit 11 combination information integration circuit 12 feature amount extraction circuit Reference Signs List 13 filter circuit 14 image logic operation circuit 15 image bus 16 CPU bus 17 CPU 18 ROM 19 RAM 20 input / output device 21 shielding plate 30 one-dimensional / two-dimensional converter

Claims (1)

[Claims] An apparatus for inspecting a woven cloth based on image data of a woven cloth taken from an image of the woven cloth, wherein the woven cloth is inspected based on the image data of the woven cloth. A tissue cycle calculating means for calculating a tissue cycle of; a comparison area setting means for setting a comparison area of the image data based on the tissue cycle; and extracting a statistic from the image data in the comparison area. And a defect extracting means for extracting a defect based on the statistic.