DE102010032241A1 - Method for detecting surface defects e.g. cracks at plastic surface of corrugated pipe, involves testing differential image between two partial region images to determine whether coherent regions are provided with grey values - Google Patents

Abstract

The method involves receiving an image from a surface of an object i.e. pipe section (2), by a camera. The image is divided into number of partial region images (4), and an inverted partial region image (12) is produced from a partial region image (6). The inverted partial region image is superimposed with another partial region image (8). A differential image (14) between the partial region images is tested to determine whether coherent regions are provided with grey values that differ from a tolerance value. The two partial region images are partially overlapped. An independent claim is also included for a device for detecting surface defects.

Description

The invention relates to a method and a device for detecting surface defects.

Plane and curved, especially cylindrical surfaces often have defects, such as voids, cracks, holes, notches, scratches, bumps, discoloration and the like. If such defects are not recognized, they cause a depreciation of the respective item up to a uselessness.

Often, therefore, such surfaces are visually inspected for defects, which may result in such optical methods, the problems that some defects are not recognized, but on the other hand, influences such as brightness variations or reflections lead to surfaces classified as flawed, the none at all Have defects.

From the DE 198 20 536 C1 For example, a device for checking surfaces of a body is known in which line patterns are projected onto painted surfaces and irregularities are made visible by movement of the surface. Here, the intervention by operating personnel is mandatory, which makes this process consuming.

The EP 1 581 803 B1 describes the generation of variable grid patterns. Here it is disadvantageous that the evaluation of recorded camera images takes place via reference images, which can easily lead to misinterpretations and thus to incorrect results in the event of positional deviations of the surfaces.

It is therefore an object of the present invention to provide a method and apparatus for detecting surface imperfections that reliably detects defects in surfaces regardless of external influences and without user intervention. Such a method and such a device for detecting surface defects should be suitable for all surfaces, ie not only flat, but also for curved and in particular cylindrical surfaces, and also be able to detect surface defects which lie in edge regions of the considered surface.

These objects are achieved by the subject matter of the independent claims. Advantageous developments emerge from the dependent claims.

In a method according to the invention for detecting surfaces, a picture is taken of a surface with a camera, and this picture is subdivided into a number of partial areas. An inverted sub-area image is generated from a first sub-area, and this inverted sub-area image is superimposed with a second sub-area image. This superimposed image is called difference image in the following. This difference image is then examined to see if there are contiguous regions of gray values that differ from a tolerance value and, in particular, are darker than a given tolerance gray value.

With the method according to the invention, coherent regions with gray values which are darker than a predefined gray value, have sufficient contrast to the immediate and thus to an environment in the difference image, reliably recognized. Such contiguous areas, if not intended features of the surface, are surface imperfections, in particular voids, cracks, indentations, scratches, bumps, discolorations, holes and the like.

By superimposing an inverted sub-area image of a first sub-area with a second non-inverted sub-area image, the gray values largely cancel each other out, with the exception of defects that have no counterpart in the respective other image. These defects are thus visible in the difference image.

Lighting-related or surface-course-related differences in brightness between subareas are ignored in that in the difference image when superimposing the superimposed partial area image with a second subarea image, a gray value is obtained that is less than the predetermined tolerance value.

A brightness gradient, for example along a curvature of the surface of the object, in particular a tube, a tube or the like or a general brightness curve in the image does not affect the detection of the defects. The method runs so stable that a change in the lighting situation, for example due to brightness fluctuations or due to positional shifts of the object, does not disturb the error detection.

The extension of the determined contiguous regions with gray values that are darker than a predefined gray value represents a measure of the size of the surface defect. For example, a very small extent can only be a soiling particle, in particular a dust grain, while a larger contiguous one Area darker than a given gray scale, a blowhole, a crack, a hole, a notch, a scratch, an elevation, a discoloration or the like may be.

With the method according to the invention, any surfaces can be examined, both flat, as well as curved and in particular cylindrical surfaces. The material of the surface is not important, it can be any surface, especially ceramic, plastic, metal, paper, rubber, glass and wood surfaces are examined. The surfaces to be examined can have a diffuse or a reflective reflection behavior and be chrome-plated.

Likewise, the shape of the surface does not matter. It is possible to examine any desired surface shapes, for example surfaces with a flat extension, cylindrical or conical surfaces or also other more complex surfaces.

The surfaces of any objects can be examined, in particular the surfaces of hoses, for example plastic and rubber hoses, the surfaces of pipes, the surfaces of cylindrical and conical pins, the surfaces of cigarettes or unpainted or painted surfaces, in particular body surfaces of any shape.

Likewise, web material can be examined by the method according to the invention. Area cameras or line scan cameras can be used as cameras.

In particular, in the case of objects whose surface has a non-homogeneous grayscale gradient or in which the gray values of the surface overlap with the gray values of the defects, and in cylindrical surfaces with their characteristic dark-light-dark curve along the curvature, in which previous methods for detecting of surface defects have great weaknesses, defects can be reliably and precisely identified with the method according to the invention.

In addition, the disadvantages of previous surface defect detection methods that flawless surfaces due to brightness variations or due to the reflection behavior of the surface have been classified as scrap, reliably avoided with the present inventive method, only surfaces are identified as defective, which actually have defects.

With the method according to the invention, it is possible to study surfaces which have a uniform surface structure, that is to say either a uniform, uniform surface or a uniform pattern with recurring elements. Less or not suitable for this method according to the invention for detecting surface defects are surfaces with marbled structures without recurring elements.

In dividing the image of the surface taken by the camera into sub-area images, the image may be divided into a plurality of equally sized or similarly sized sub-area images of substantially coincident shape. Also, a surface having a regular pattern may be subdivided so that recurring elements of the regular pattern form the subregions.

For more complex surfaces or surfaces with a brightness gradient, the sub-area images into which the captured image is subdivided may be selected to be correspondingly small.

The size of the sub-area images into which the image taken by the camera is subdivided from the surface of the object depends on the overall object surface to be inspected in each case. For surfaces of hoses, pipes, cylindrical and conical pins, cigarettes or the like, it is advantageous to select the sub-area images to include a cross-sectional area from edge to edge. For example, in a corrugated pipe, the sub-area image may cover a cross-sectional area of the corrugated pipe and extend in the longitudinal extent of the corrugated pipe from a narrower cross-section area in one notch to the next adjacent narrower-cross-section area in a notch, or a wide cross-section area to a neighboring area with a broad cross-section.

With the method according to the invention for detecting surface defects, the entire surface is examined, which also includes the edge. It inspects the entire visible area. Accordingly, defects located directly on the edge of the surface of the object can be reliably detected, which was not possible with the previous methods. Likewise, the change in the lighting situation does not affect the outcome of the process in many areas.

With the method according to the invention for detecting surface defects, all existing defects are found without having to use fringe patterns, which avoids the problems associated with such fringe patterns, namely the difficulty of using blob analyzes when defects at the transition from Bright to dark, or creating search areas for the light and dark stripes.

The method according to the invention is outstandingly suitable for the examination of cylindrical surfaces which have a brightness course along the surface curvature from the camera view. In such cylinder surfaces, such a brightness gradient prevents conventional blob analysis, since the imperfections, in particular scratches, voids and protrusions differ from the gray values of the actual surface only in a narrow area on the surface or often at the transition from the light to the darker region of the curved surface Area lie.

If it is determined during the image recording that a brightness gradient, that is to say in particular a brightness gradient, is present along the surface of the object to be examined, it is expedient, when subdividing the recorded image into subregion images, to select these subregion images correspondingly, specifically such that, when the first inverted subregion image is superimposed with a second noninverted subregion image, the gray-scale values that result in a defect-independent color have a low gray value which is below the tolerance value and thus are brighter than it, and are therefore not identified as a surface defect become.

According to a first embodiment of the invention, the sub-area images in the captured image are each arranged next to one another. This is particularly advantageous for objects with a substantially cylindrical or conical surface, because in this way the entire recorded image of the surface of the object can be completely imaged with a series of sub-area images arranged next to one another.

According to a further embodiment of the invention, two sub-area images are respectively arranged on opposite edges of the recorded image or mirror-symmetrically to an axis of symmetry. This subdivision is particularly suitable for objects with a broader extent or in the case of mirror-symmetrical objects.

According to a further embodiment of the invention, adjacent partial area images may partially overlap each other. As a result, errors that are located at the edge of sub-area images can be detected reliably and completely.

According to another embodiment of the invention, in images of objects having a regular surface pattern, the sub-area images are selected to have a substantially coincident shape. This ensures that the first inverted subregion image can be superimposed with a second non-inverted subregion image, and the difference image can be examined for contiguous regions present therein with dark gray values.

According to a further embodiment of the invention, the subdivision of the surfaces of the recorded image into subregion images is performed by a pattern segmentation. In doing so, the pattern of the surface of the captured image is examined for recurring elements, and the recurring elements form the individual sub-area images. In the process, especially the light-dark transitions are considered.

Accordingly, for patterns with regular patterns, these patterns are used to create the subregions, and in the difference image, these patterns become largely invisible, but the flaws are clear.

According to a further embodiment of the invention, the surface of at least one illumination unit is illuminated in the image recording. It is important to ensure that the resulting gradient of brightness over the surface of the image to be recorded is as low as possible.

Alternatively or additionally, the surface of the object to be examined may be self-luminous / electroluminescent.

According to a further embodiment of the invention, the method steps of generating an inverted subregion image from a first subregion image, superimposing this subregional inverted image with a second subregion image and examining the difference image to determine whether there are contiguous gray value regions are different from a tolerance value, repeated for further pairs of sub-area images. As a result, all partial areas into which the recorded image of the surface has been subdivided can be checked, and thus defects in the entire surface of the object under consideration can be detected.

In one embodiment of the invention, all partial area images can be checked in pairs in succession, thus checking the entire surface for a minimum amount of computation. To achieve a higher recognition accuracy, different combinations of sub-area images can be checked against each other.

According to a further embodiment of the invention is on the difference images for Classification of the found defects used a blob analysis and / or a histogram analysis. Blob analysis analyzes areas of the image in which adjacent pixels have the same or similar gray levels to identify contiguous areas with the same or similar gray levels. In a histogram analysis, the frequencies of the individual gray values in an image or an image area are recorded and usually displayed as a bar chart.

According to a further embodiment of the invention, a plurality of image captures are performed while moving the object being viewed and keeping the camera stationary or while the camera is moving and the object is held stationary. This results in views of the object from different perspectives. Defects that only stand out in certain perspectives can be reliably detected.

In a first variant, only the sub-area images generated are each superimposed on one perspective. In a second variant, sub-area images from different perspective images are superimposed with each other. This results in a more comprehensive view of the surface of the object.

According to a further embodiment of the invention, a plurality of image recordings are carried out with different illumination sources or with different exposure times. Defects that are not recognizable in some lighting situations, but highlight in a different lighting situation can be reliably detected.

In a first variant, only the sub-area images of an image acquisition, each with an illumination source and each with an exposure time are superposed against each other. In a further variant, sub-area images which originate from different image recordings or have been illuminated with different illumination sources or have been exposed with different exposure times are compared with one another.

When it is mentioned above that sub-area images are compared with one another, it is to be understood that in accordance with claim 1, an inverted sub-area image is generated from a first sub-area image and this with a second non-inverted sub-area image. Image is overlaid and the difference image thus generated is examined to see whether there are contiguous areas with gray values that differ from a tolerance value.

The invention also relates to a device for detecting surface defects with a camera, which is adapted to take an image of a surface of an object, and to a data processing unit, which is set up in such a way that it integrates the image taken by the camera Dividing a number of sub-area images, creating an inverted sub-area image from a sub-area image, superimposing this sub-area inverted image with a second sub-area image, and examining this difference image to see if there are contiguous areas of gray levels that are different from one another Differ tolerance value.

The advantages of this device according to the invention for detecting surface defects correspond to the above advantages stated with reference to the method for detecting surface defects, which are hereby also disclosed with respect to the device, without repeating them in detail.

According to a first embodiment of the device according to the invention, a lighting unit is further provided for illuminating the surface of the object. Thereby, the surface of the object to be examined can be illuminated, with a particularly uniform illumination with little or at best no brightness gradient is desirable. For this purpose, several lighting units can be used.

According to a further embodiment of the invention, the camera is designed to be movable or displaceable, so that it can make image recordings of the surface of the object from different perspectives. Alternatively or additionally, a slide can be provided on which the object is positioned for image acquisition, and this slide can also be designed to be movable or movable to allow image capturing the object from different perspectives.

The invention is explained in more detail by means of embodiments with reference to the accompanying figures.

1 shows a schematic diagram of the generation of a difference image from an image recording of a pipe section; and

2 shows an image of a corrugated pipe section with sub-area images generated therein.

1 shows a schematic diagram of the generation of a difference image 14 from a picture of a pipe section 2 ,

In 1 At the top is a perspective view of a pipe section 2 shown. The axis of the pipe section 2 runs in the drawing plane from left to right, with the axis is inclined slightly obliquely to the bottom right.

The pipe section 2 is from one in 1 Illuminated lighting unit not shown. Due to the cylindrical shape of the pipe section 2 arise in the lower and upper visible end of the shell of the pipe section 2 shaded areas in 1 sketched by crossed lines. By a in 1 also not shown imaging unit, in particular camera, is now an image of the pipe section 2 made. This image is then replaced by an in 1 also not shown data processing unit in sub-area images 4 divided, of which in 1 exemplarily four partial area images 4 are shown.

The four sub-area pictures 4 each have a rectangular shape and are arranged side by side. The subareas 4 are chosen so large that they each have a complete cross-sectional area of the pipe section 2 from the upper visible edge to the lower visible edge, and that their upper and lower end sections still have the light background behind the tube section 2 demonstrate. The subarea pictures 4 are with respect to the pipe section 2 positioned so that its vertical orientation is not in a plane with the cross-sectional plane of the pipe section 2 but an angle with the cross-sectional plane of the pipe section 2 includes. The juxtaposed subareas 4 are arranged, viewed from left to right, arranged in descending order, so that the cross-sectional area of the pipe section 2 in all sub-area images 4 is arranged in each case in the same height position.

The second sub-area image from the left becomes the first sub-area image below 6 and the third sub-area image from the left will subsequently become a second sub-area image 8th designated. On the outer jacket of the pipe section 2 is an in 1 exemplary jagged defect 10 For example, a spot, a material defect or a hole of the mantle surface of the pipe section 2 available. This defect 10 is in the pipe section 2 arranged to be in the second sub-area image 8th lies.

Now by the data processing unit from the first sub-area image 6 an inverted first partial image 12 generated. The bright areas of the background above the upper visible edge of the pipe section 2 and below the lower visible edge of the pipe section 2 and the bright center area of the mantle surface of the tube section, which on the scale of 0 (black), ..., 254 (white) has a gray value of, for example, 230, inverted into dark areas in the inverted first sub-area image 12 are shown in dashed lines and there accordingly have a gray value of 24. The shady border areas in the first section image 6 have a gray value of, for example, 54. In the first inverted subrange image 12 Accordingly, these border areas have a gray value of 200.

This first inverted subrange image 12 will now be with the second sub-area image 8th superimposed and this difference image is in 1 shown at the bottom. The inverted pipe background areas and the inverted pipe center area, which have a gray value of 24, overlap with the non-inverted light pipe background areas and the non-inverted pipe center area with a gray value of 230 to give a gray value of 254 corresponding to the color white. Accordingly, these areas are shown in white in the difference image.

Likewise, the bright pipe edge areas overlap in the inverted first partial area image 12 that have a gray value of 54 with the dark non-inverted pipe edge areas in the second sub-area image 8th , which have a gray value of 200, so that also gives a gray value of 254 for the pipe edge areas and these pipe edge areas in the difference image 14 are white.

Only the fault 10 which has, for example, a gray value of 180 is superimposed on the gray value of the inverted central region in the first inverted subregion image 12 of 24, so that in the difference image 14 for this defect 10 a gray value of 204 results. By comparing the gray value of 204 of the identified contiguous area 16 in the difference image 14 with a tolerance value of, for example, 230, that results in the identified contiguous area 16 darker than this tolerance value. Accordingly, this contiguous gray area becomes a defect 16 identified.

To classify this defect, a blob analysis or a histogram analysis can now be used. It can be determined that this defect 16 has an expansion indicating that this is not a negligible impurity in the form of a speck of dust or a particle, but rather a more serious defect, such as a surface defect or a hole.

2 shows an image of a corrugated pipe section 18 with sub-area images generated therein 20 ,

This in 2 corrugated tube shown 18 is also illuminated by a lighting unit and recorded with a camera.

In the recorded image of the corrugated tube 18 Now a pre-segmentation is performed. The light-dark transitions are considered. In the present corrugated pipe 18 whose axis lies in the plane of the drawing in the right-left direction, there are four notches each with a reduced diameter and interposed corrugated pipe sections with an enlarged diameter whose shape and course correspond to each other.

As in 1 so is the corrugated pipe 18 of the 2 illuminated so that arise on the upper and lower pipe edge areas shading. The background is bright.

During pre-segmentation, the entire picture of the corrugated tube is now displayed 18 subdivided into sub-area images, of which in 2 exemplarily three partial area pictures 20 are shown side by side. These sub-area images each extend in the direction of corrugation from a reduced diameter notch to the adjacent reduced diameter notches and radially such that the perimeter areas are completely therein as well as a piece of pipe background therein. The subarea pictures 20 are each rectangular and arranged side by side.

In the right-most sub-area image 20 is a defect 22 to recognize with an elongated, curved course.

The further method steps of generating a first inverted subregion image, superposing the inverted first subregion image with a second subregion image, examining this difference image as to whether there are contiguous regions with gray values that differ from a tolerance value, and darker than a given tolerance gray value, are as above with respect to the claims and with reference to the 1 described process steps and are not repeated again in detail.

The defect 22 is represented by a difference image that is from the right-most of the three sub-area images shown 20 and another sub-area image 20 is formed, wherein one of the two sub-area images is previously inverted, visible, because in this difference image has only the contiguous area 22 a gray value that is darker than a predetermined tolerance value.

In alternative embodiments, not shown here, the sub-area images 4 and 22 also overlap each other, so that any defects lying in edge regions of these sub-area images can be identified coherently and reliably in the generated difference image.

Likewise, the sub-area images can also be arranged differently, for example, two superimposed differential images can be provided per cross-sectional area of the object under consideration.

With the method according to the invention, defects can be easily recognized and identified.

LIST OF REFERENCE NUMBERS

2

pipe section

4

subregions

6

first subarea

8th

second subarea

10

void

12

inverted subrange image

14

difference image

16

identified defect

18

Corrugated pipe section

20

subregions

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 19820536 C1 [0004]
• EP 1581803 B1 [0005]

Claims (14)

Method for detecting surface defects, in which a surface of an object ( 2 ) taken a picture with a camera and this picture in a number of sub-area pictures ( 4 ), whereby from a partial area image ( 6 ) an inverted subregion image ( 12 ), this inverted subrange image ( 12 ) with a second partial area image ( 8th ) is superimposed and this difference image ( 14 ) is examined to see if there are contiguous regions of gray values that differ from a tolerance value. Method according to claim 1, wherein the subregion images ( 4 ) are each arranged side by side in the recorded image. The method of claim 1 or 2, wherein each two sub-area images are arranged on opposite edges of the image or mirror-symmetrical to an axis of symmetry. Method according to one of the preceding claims, wherein adjacent sub-area images partially overlap each other. Method according to one of the preceding claims, wherein when dividing the image recorded by the camera with a regular pattern ( 18 ) into a number of sub-area images ( 20 ) are subdivided into a largely consistent form. Method according to one of the preceding claims, wherein the subdivision of the surface of the recorded image ( 18 ) in sub-area images ( 20 ) is done by a pattern segmentation. Method according to one of the preceding claims, wherein in the image recording the surface is illuminated by a lighting unit. Method according to one of the preceding claims, wherein the surface is self-luminous / electroluminescent. Method according to one of the preceding claims, wherein the method steps of generating an inverted subregion image ( 12 ) from a first sub-area image ( 6 ), overlaying this inverted subrange image ( 12 ) with a second partial area image ( 8th ) and examining the difference image ( 14 ) for whether there are contiguous regions with gray values that differ from one tolerance value, for further pairs of subregions ( 4 ) be repeated. Method according to one of the preceding claims, wherein a blob analysis and / or a histogram analysis is used on the difference images. Method according to one of the preceding claims, wherein a plurality of image recordings are carried out while the object ( 2 ) is moved. Method according to one of the preceding claims, wherein a plurality of image recordings are carried out with different illumination sources or or with different exposure times. A device for detecting surface defects, comprising a camera arranged to take an image of a surface of an object; and a data processing unit configured to copy the image captured by the camera into a number of sub-area images ( 4 ), from a sub-area image ( 6 ) an inverted subregion image ( 12 ), this inverted subrange image ( 12 ) with a second partial area image ( 8th superimposed, and this difference image ( 14 ) examines whether there are contiguous regions of gray values that differ from a tolerance value. Apparatus according to claim 13, further comprising at least one lighting unit for illuminating the surface of the object.

Images