A paint defect detection system based on machine vision

Abstract

The present invention discloses a paint defect detection system based on machine vision, which relates to the field of defect detection technology, and solves the problem of simply generating corresponding abnormal signals without conducting deeper analysis, resulting in the need for subsequent maintenance personnel to analyze the corresponding abnormal causes. The present invention is aimed at small-area images, and adopts a method of determining point-line segmentation and determining the difference to determine whether the corresponding small-area image is a bubble image or a corresponding scratch image. If it is a corresponding bubble image, it is analyzed in turn to determine the specific cause of the corresponding abnormal image, and transmitted to an external display terminal in the form of signal display. In this way, not only can it fully detect whether there are defects on the corresponding paint surface, but it can also fully analyze and confirm the specific cause of the defect, and display the analyzed and confirmed specific cause for external personnel to view and take timely countermeasures, thereby improving the practicality of this defect detection system.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and in particular to a paint surface defect detection system based on machine vision. Background Art

[0002] Machine vision is a rapidly developing branch of artificial intelligence. Simply put, machine vision is the use of machines to replace human eyes for measurement and judgment. The machine vision system uses machine vision products (i.e. image acquisition devices, which are divided into CMOS and CCD) to convert the captured target into image signals and transmit them to a dedicated image processing system.

[0003] The application with patent publication number CN109239086A discloses a vehicle paint and appearance defect detection system, including a pair of brackets, each bracket having a crossbeam and two columns, and a detection area is formed by the four columns of the pair of brackets, and a Y-direction linear guide is provided on the crossbeam of the bracket, and the slide of the detection system is provided with an X-direction linear guide, and the X-direction linear guide is mounted on the Y-direction linear guide on the two crossbeams, and under the drive of a first motor, the X-direction linear guide moves the slide back and forth along the Y-axis direction, and the slide moves back and forth along the X-axis direction under the drive of a second motor. The slide has a rotating axis (completing A-axis rotation), and below the slide, the rotating axis is connected to a rotating arm, and the rotating arm is in a "C" shape. A top edge intelligent industrial camera is provided on the top edge of the rotating arm, and the lens of the top edge camera faces downward to the top of the vehicle. A side edge intelligent industrial camera is provided on the side of the rotating arm, and the lens of the side edge camera faces the side of the vehicle. The camera on the rotating arm is connected to the rotating arm through a camera gimbal.

[0004] During the defect detection process of the paint surface, image analysis is generally used to confirm whether the corresponding paint surface is abnormal. If it is abnormal, a corresponding abnormal signal is generated. However, this method simply generates the corresponding abnormal signal without conducting a deeper analysis, resulting in the need for subsequent maintenance personnel to analyze the corresponding cause of the abnormality, which is not conducive to the maintenance efficiency of the maintenance personnel. The same defect will still appear later. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a paint defect detection system based on machine vision, which solves the problem of simply generating corresponding abnormal signals without conducting deeper analysis, resulting in subsequent maintenance personnel having to analyze the corresponding abnormal causes.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A paint defect detection system based on machine vision, comprising:

[0007] The image acquisition end uses a specific scanning and recognition device to acquire the image of a single side of the vehicle body, and transmits the acquired single-side image to the preliminary comparison end. During the acquisition process, the single-side images at different vehicle body positions are marked;

[0008] The database stores standard images corresponding to different positions of the vehicle body;

[0009] The preliminary comparison end compares the single-sided image with the standard image stored in the database, and marks the corresponding standard image as a double-different image or a single-different image according to the comparison result. The specific method is as follows:

[0010] The acquired single-sided image is compared with the standard image at the corresponding position. When the comparison results are consistent, no processing is performed. When the comparison results are inconsistent, the corresponding single-sided image is marked as an abnormal image.

[0011] Then confirm whether the single-sided picture opposite to the abnormal picture is also an abnormal picture. If it is an abnormal picture, combine the two abnormal pictures to determine whether they are the same picture. If they are the same picture, mark the picture as a double-different picture. If they are not the same picture, mark the determined abnormal picture as a single-different picture, and transmit the single-different picture to the picture analysis end;

[0012] The signal generating end receives the dual-different images, generates a device abnormality signal, and transmits it to the external display end;

[0013] The picture analysis end includes a different area confirmation unit, a small area analysis unit and a large area analysis unit;

[0014] The different area confirmation unit confirms the total area of ​​the single different picture, and marks the single different picture as a line point picture or a different surface picture according to the confirmation result. The specific method is as follows:

[0015] Confirm the total area of ​​the single-image screen and mark the confirmed total area as ZM k , where k represents different single-image pictures, and the total area ZM k Compare with the preset parameter Ys, where Ys is the preset value;

[0016] When ZM k When <Ys, the corresponding single-different picture is marked as a line-point picture, otherwise, the corresponding single-different picture is marked as a different-surface picture, and the marked line-point picture is transmitted to the small-area analysis unit, and the marked different-surface picture is transmitted to the large-area analysis unit;

[0017] The small area analysis unit receives the line point image, confirms the edge points of the line point image, establishes an edge segmentation line, divides it into two groups of areas through the edge segmentation line, and then confirms the longest line length of the two groups of areas. According to the specific recognition parameters, it is determined whether the line point image is an abnormal point or a scratch image. The specific method is as follows:

[0018] The line point picture is received, and two points with the farthest distance from each other are selected from the line point picture, and both points are marked as edge points. The two edge points are connected to confirm the edge dividing line, and the line length corresponding to the edge dividing line is recorded and marked as XS. The area located at the edge dividing line in the line point picture is marked as one side area, and the other area located at the edge dividing line is marked as the other side area;

[0019] Establish a vertical line perpendicular to the edge dividing line, then construct several parallel lines parallel to the vertical line in the two areas, mark the line segments of the corresponding parallel lines in the two areas as parallel line segments, then record the lengths of several parallel line segments, confirm the longest line length, mark it as the standard line length, and record the specific value SZ of the standard line length;

[0020] After performing difference processing on XS and SZ and performing absolute value processing, confirm the verification parameter HD, and compare HD with the preset parameter Y1, where Y1 is the preset value;

[0021] When HD<Y1, it means that the corresponding line and dot image is caused by bubbles, and the corresponding paint flow abnormality signal is generated by the signal generating end and transmitted to the external display end;

[0022] The large area analysis unit receives the out-of-plane images and performs color difference analysis on them to determine whether the out-of-plane images are caused by color difference. When HD≥Y1, it means that the corresponding line and dot images are caused by cracks, and a corresponding uneven paint mixing signal is generated through the signal generation end and transmitted to the external display end.

[0023] Preferably, the large area analysis unit performs color difference analysis in the following specific manner:

[0024] Confirm the vehicle surface to which the different-plane image belongs, mark the area outside the different-plane image in the vehicle surface as the standard area, and perform color difference analysis and comparison between the standard area and the different-plane image to determine whether the colors of the two images are consistent. If they are consistent, other abnormal signals are generated through the signal generation end. If they are inconsistent, no processing is performed.

[0025] Beneficial Effects

[0026] The present invention provides a paint defect detection system based on machine vision. Compared with the prior art, it has the following beneficial effects:

[0027] The present invention first confirms the corresponding paint surface with defects and marks it as an abnormal picture. Subsequently, the abnormal picture is analyzed to determine the specific area parameters of the picture. According to the specific results of different area parameters, it is analyzed whether the corresponding picture is a small area picture or a large area picture.

[0028] For small-area images, the method of determining point and line segmentation and determining the difference is used to determine whether the corresponding small-area image is a bubble image or a corresponding scratch image. If it is a corresponding bubble image, it is analyzed in turn to determine the specific cause of the corresponding abnormal image, and transmitted to the external display terminal in the form of a signal display. This method can not only fully detect whether there are defects on the corresponding paint surface, but also fully analyze and confirm the specific causes of the defects, and display the specific causes for external personnel to view and take timely response measures, thereby improving the practicality of this defect detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the principle framework of the present invention;

[0030] Figure 2 This is a principle framework diagram of the picture analysis end of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Embodiment 1

[0033] See also Figure 1 as well as Figure 2 , the present application provides a paint defect detection system based on machine vision, including a picture acquisition end, a signal generation end, a preliminary comparison end, a database and a picture analysis end, the picture acquisition end is electrically connected to the preliminary comparison end input node, the database is electrically connected to the preliminary comparison end input node, the preliminary comparison end is electrically connected to the picture analysis end and the signal generation end input node, and the picture analysis end is electrically connected to the signal generation end input node;

[0034] The picture analysis end includes a different area confirmation unit, a small area analysis unit and a large area analysis unit, wherein the different area confirmation unit is electrically connected to the input nodes of the small area analysis unit and the large area analysis unit respectively;

[0035] The image acquisition end uses a specific scanning and recognition device to acquire the image of a single side of the vehicle body, and transmits the acquired single-side image to the preliminary comparison end. During the acquisition process, the single-side images of different vehicle body positions are marked, and the marked content is the number content of the specific position of the vehicle body;

[0036] The database internally stores standard images corresponding to different positions of the vehicle body, wherein the standard images are all preset images, which are prepared in advance by the operator based on experience;

[0037] The preliminary comparison end compares the single-sided picture with the standard picture stored in the database, and marks the corresponding standard picture as a double-different picture or a single-different picture according to the comparison result, and transmits the single-different picture to the picture analysis end, and transmits the double-different picture to the signal generation end, wherein the specific method of performing the comparison is:

[0038] The acquired single-sided image is compared with the standard image at the corresponding position. When the comparison results are consistent, no processing is performed. When the comparison results are inconsistent, the corresponding single-sided image is marked as an abnormal image.

[0039] Then confirm whether the single-sided picture opposite to the abnormal picture is also an abnormal picture. If it is an abnormal picture, combine the two abnormal pictures to determine whether they are the same picture. If they are the same picture, mark the picture as a double-different picture. If they are not the same picture, mark the determined abnormal picture as a single-different picture, and transmit the single-different picture to the picture analysis end. Specifically, when there is an abnormality in the picture of one group of vehicle faces, if the other group of vehicle faces is also abnormal, it is necessary to compare and analyze the two abnormal pictures to determine whether they are the same picture. If they are the same picture, then the corresponding paint spraying equipment may have certain problems, so it can be directly analyzed and determined to generate and display the corresponding signal;

[0040] The signal generating end receives the dual-different images, generates an equipment abnormality signal, and transmits it to the external display end for external personnel to view and take timely response measures to prevent other problems from occurring. Specifically, when abnormal images appear continuously in the same position, it means that there are certain problems with the corresponding painting equipment during the painting process, resulting in continuous abnormal images.

[0041] Embodiment 2

[0042] In the specific implementation process, this embodiment is different from the first embodiment in that:

[0043] The different area confirmation unit inside the picture analysis end confirms the total area of ​​the single different picture, and marks the single different picture as a line point picture or a different surface picture according to the confirmation result, wherein the specific method of confirmation is:

[0044] Confirm the total area of ​​the single-image screen and mark the confirmed total area as ZM k , where k represents different single-image pictures, and the total area ZM k Compare with the preset parameter Ys, where Ys is a preset value, and its specific value is determined by the operator based on experience;

[0045] When ZM k When <Ys, the corresponding single-different picture is marked as a line-point picture, otherwise, the corresponding single-different picture is marked as a different-surface picture, and the marked line-point picture is transmitted to the small-area analysis unit, and the marked different-surface picture is transmitted to the large-area analysis unit;

[0046] Specifically, when the area of ​​the corresponding different-region screen is smaller as a whole, it means that the area is smaller. Therefore, only point abnormalities or small abnormalities will cause such a situation. On the contrary, when the area of ​​the corresponding different-region screen is larger as a whole, it means that the area is larger. Therefore, only large areas will cause such a situation. Therefore, it is generally a regional abnormality.

[0047] The small area analysis unit receives the line point image, confirms the edge points of the line point image, establishes an edge segmentation line, divides it into two groups of areas by the edge segmentation line, confirms the longest line length of the two groups of areas, and determines whether the line point image is an abnormal point or a scratch image according to specific recognition parameters, wherein the specific method of making the determination is:

[0048] The line point picture is received, and two points with the farthest distance from each other are selected from the line point picture, and both points are marked as edge points. The two edge points are connected to confirm the edge dividing line, and the line length corresponding to the edge dividing line is recorded and marked as XS. The area located at the edge dividing line in the line point picture is marked as one side area, and the other area located at the edge dividing line is marked as the other side area;

[0049] Establish a vertical line perpendicular to the edge dividing line, then construct several parallel lines parallel to the vertical line in the two areas, mark the line segments of the corresponding parallel lines in the two areas as parallel line segments, then record the lengths of several parallel line segments, confirm the longest line length, mark it as the standard line length, and record the specific value SZ of the standard line length;

[0050] After performing difference processing on XS and SZ and performing absolute value processing, confirm the verification parameter HD, and compare HD with the preset parameter Y1, where Y1 is a preset value, and its specific value is determined by the operator based on experience;

[0051] When HD<Y1, it means that the corresponding line and dot image is caused by bubbles, and the corresponding paint flow abnormality signal is generated through the signal generating end and transmitted to the external display end. On the contrary, it means that the corresponding line and dot image is caused by cracks, and the corresponding paint mixing unevenness signal is generated through the signal generating end and transmitted to the external display end.

[0052] Specifically, the preset parameter Y1 is a standard value, and its value is generally not too large, generally taking a value of 2cm. When the difference between the two line lengths is too large, it means that the corresponding area is definitely not circular, but long. Only when it is circular, it will be caused by bubbles. If it is long, it will be caused by the corresponding cracks. Generally, when there are bubbles on the paint surface of the corresponding car surface, it is generally caused by the flow of the paint used. If it is a crack, it is either scratched by an external sharp object, or the paint is unevenly mixed during the manufacturing process. It is directly displayed, and the operator can check the specific cause in time;

[0053] Embodiment 3

[0054] In the specific implementation process, this embodiment is different from the above embodiment in that:

[0055] The large area analysis unit receives the out-of-plane picture and performs color difference analysis on the out-of-plane picture to determine whether the out-of-plane picture is caused by color difference. The specific method of performing the color difference analysis is as follows:

[0056] Confirm the vehicle surface to which the different-face image belongs, mark the area outside the different-face image in the vehicle surface as the standard area, and perform color difference analysis and comparison between the standard area and the different-face image to determine whether the colors of the two images are consistent. If they are consistent, generate other abnormal signals through the signal generation end; if they are inconsistent, do not perform any processing;

[0057] Specifically, when the total area of ​​the corresponding non-planar images is too large, it is either a color difference problem or a problem with the corresponding sheet metal or internal parts. Therefore, it is possible to determine whether there is a corresponding color difference problem by confirming whether there is a color difference, and display such problems for external operators to check in time and take timely countermeasures.

[0058] Embodiment 4

[0059] The specific implementation process of this embodiment includes all the implementation processes of the above embodiments.

[0060] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0061] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A paint defect detection system based on machine vision, characterized in that: include: The image acquisition end uses a specific scanning and recognition device to acquire the image of a single side of the vehicle body, and transmits the acquired single-side image to the preliminary comparison end. During the acquisition process, the single-side images at different vehicle body positions are marked; The database stores standard images corresponding to different positions of the vehicle body; The preliminary comparison end compares the single-sided image with the standard image stored in the database, and marks the corresponding standard image as a double-different image or a single-different image according to the comparison result, and transmits the single-different image to the image analysis end, and transmits the double-different image to the signal generation end. The specific method is as follows: The acquired single-sided image is compared with the standard image at the corresponding position. When the comparison results are consistent, no processing is performed. When the comparison results are inconsistent, the corresponding single-sided image is marked as an abnormal image. Then confirm whether the single-sided picture opposite to the abnormal picture is also an abnormal picture. If it is an abnormal picture, combine the two abnormal pictures to determine whether they are the same picture. If they are the same picture, mark the picture as a double-different picture. If they are not the same picture, mark the determined abnormal picture as a single-different picture, and transmit the single-different picture to the picture analysis end; The signal generating end receives the dual-different images, generates a device abnormality signal, and transmits it to the external display end; The picture analysis end includes a different area confirmation unit, a small area analysis unit and a large area analysis unit; The different area confirmation unit confirms the total area of ​​the single different screen, and marks the single different screen as a line point screen or a different surface screen according to the confirmation result. The specific method is as follows: Confirm the total area of ​​the single-image screen and mark the confirmed total area as ZM k , where k represents different single-image pictures, and the total area ZM k Compare with the preset parameter Ys, where Ys is the preset value; When ZM k When <Ys, the corresponding single-different picture is marked as a line-point picture, otherwise, the corresponding single-different picture is marked as a different-surface picture, and the marked line-point picture is transmitted to the small-area analysis unit, and the marked different-surface picture is transmitted to the large-area analysis unit; The small area analysis unit receives the line point image, confirms the edge points of the line point image, establishes an edge segmentation line, divides it into two groups of areas through the edge segmentation line, confirms the longest line length of the two groups of areas, and determines whether the line point image is an abnormal point or a scratch image based on specific recognition parameters; The large area analysis unit receives the out-of-plane picture and performs color difference analysis on the out-of-plane picture to determine whether the out-of-plane picture is caused by color difference.

2. The machine vision-based paint defect detection system according to claim 1, characterized in that: The specific method of the small area analysis unit to determine whether the line point image is an abnormal point or a scratch image is as follows: The line point picture is received, and two points with the farthest distance from each other are selected from the line point picture, and both points are marked as edge points. The two edge points are connected to confirm the edge dividing line, and the line length corresponding to the edge dividing line is recorded and marked as XS. The area located at the edge dividing line in the line point picture is marked as one side area, and the other area located at the edge dividing line is marked as the other side area; Establish a vertical line perpendicular to the edge dividing line, then construct several parallel lines parallel to the vertical line in the two areas, mark the line segments of the corresponding parallel lines in the two areas as parallel line segments, then record the lengths of several parallel line segments, confirm the longest line length, mark it as the standard line length, and record the specific value SZ of the standard line length; After performing difference processing on XS and SZ and performing absolute value processing, confirm the verification parameter HD, and compare HD with the preset parameter Y1, where Y1 is the preset value; When HD<Y1, it means that the corresponding line and dot image is caused by bubbles, and a corresponding paint flow abnormality signal is generated through the signal generating end and transmitted to the external display end.

3. The machine vision-based paint defect detection system according to claim 2, characterized in that: When HD≥Y1, it means that the corresponding line and dot image is caused by cracks, and the corresponding uneven paint mixing signal is generated through the signal generating end and transmitted to the external display end.

4. The paint defect detection system based on machine vision according to claim 1, characterized in that: The specific method of the large area analysis unit for color difference analysis is: Confirm the vehicle surface to which the different-plane image belongs, mark the area outside the different-plane image in the vehicle surface as the standard area, and perform color difference analysis and comparison between the standard area and the different-plane image to determine whether the colors of the two images are consistent. If they are consistent, other abnormal signals are generated through the signal generation end. If they are inconsistent, no processing is performed.

Images