Transparent coiled material surface defect detection method robust to flare

Abstract

The invention discloses a flare-robust transparent coiled material surface defect detection method, which comprises the following steps of: firstly, shooting a polarization mosaic image of a coiled material surface under a flare condition, obtaining polarization images of four angles through interpolation calculation, and calculating an intensity image and a polarization phase angle image; gaussian filtering is carried out on the polarization phase angle image, binarization is carried out on the filtered image, a connected region is detected in a binary image, a bounding box of the maximum connected region is drawn, and a surface flare region is obtained; a surface flare area is subtracted from the intensity image, Gaussian filtering is carried out on the remaining part of the image, connected areas are detected, and a bounding box of the maximum connected area is drawn; and when the bounding box of the maximum connected area is detected in the intensity image, the winding system stops running, and a surface defect detection result is output. The method has the advantages of high flare robustness, high detection stability and high detection speed.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and in particular relates to a method for detecting surface defects of transparent coiled materials which is robust to glare. Background Art

[0002] The coil surface defect detection sensor is a real-time monitoring device for the coil surface status during the winding and unwinding process. The sensor can identify typical defects such as scratches, holes, and cracks on the coil surface, playing a vital role in ensuring product quality and maintaining safe operation of equipment.

[0003] In the winding process, transparent coils are typical non-Lambertian materials, and their surfaces are prone to glare, which poses a great challenge to defect detection based on computer imaging technology. Existing methods have poor stability in identifying and separating glare areas, and often produce false alarms due to glare interference. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a method for detecting surface defects of transparent coils that is robust to glare. First, a polarized mosaic image of the coil surface under glare conditions is captured, and polarized images at four angles are obtained by interpolation calculation, and an intensity image and a polarization phase angle image are calculated; then, the polarization phase angle image is Gaussian filtered, and the filtered image is binarized, and the connected area is detected in the binary image, and the bounding box of the largest connected area is drawn to obtain the area of ​​surface glare; then, the area of ​​surface glare is subtracted from the intensity image, and the remaining part of the image is Gaussian filtered to detect the connected area and draw the bounding box of the largest connected area; when the bounding box of the largest connected area is detected in the intensity image, the winding system stops running and outputs the surface defect detection result. The method of the present invention has the advantages of strong robustness to glare, high detection stability, and fast detection speed.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0006] Step 1: After the winding system is turned on, a split-focal plane polarization camera equipped with a bi-telecentric lens is used to capture the polarization mosaic image of the i-th frame under flare conditions, and the polarization images I0, I1 at four angles of 0°, 45°, 90°, and 135° are obtained through interpolation calculation. 45 , I 90 , I 135 ;

[0007] Step 2: Use the formula S0 = 0.5*(I0 + I 45 +I 90 +I 135), calculate the intensity image S0 of the i-th frame; use the formula AoP = 0.5*arctan(S2 / S1) to calculate the polarization phase angle image AoP of the i-th frame, where S1 = I0-I 90 , S2=I 45 -I 135 ;

[0008] Step 3: Perform Gaussian filtering on the polarization phase angle image AoP, binarize the filtered image, detect connected areas in the binary image, draw the bounding box of the largest connected area, obtain the polarization phase angle image AoP′ containing the bounding box, and obtain the surface glare area D;

[0009] Step 4: Subtract the surface glare area D from the intensity image S0, perform Gaussian filtering on the remaining part of the image, binarize it after filtering, detect the connected area in the binary image, draw the bounding box of the largest connected area, and obtain the intensity image containing the bounding box; when the bounding box of the largest connected area is detected in the intensity image of the i-th frame, the winding system stops running and outputs the surface defect detection results.

[0010] Preferably, the binarization is performed using the Otsu method.

[0011] Preferably, when Gaussian filtering is performed on the polarization phase angle image AoP, the Gaussian kernel is 5*5 and the standard deviation is 2.

[0012] The beneficial effects of the present invention are as follows:

[0013] The present invention forms an imaging system by using a split-focus plane polarization camera and a double telecentric lens, and utilizes the characteristic that the polarization phase angle image is sensitive to the glare on the surface of the transparent coil to obtain the surface glare area. After the glare area is separated, the defect is quickly detected on the intensity image. The method of the present invention has the advantages of strong robustness against glare, high detection stability and fast detection speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the method of the present invention;

[0015] Figure 2 is the polarization image of four angles of the i-th frame in the embodiment of the present invention;

[0016] Figure 3 is the polarization phase angle AoP and intensity S0 image of the i-th frame in the embodiment of the present invention;

[0017] Figure 4 is the detection result of the i-th frame polarization phase angle image in the embodiment of the present invention;

[0018] Figure 5 is the detection result of the i-th frame intensity image in the embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0020] The purpose of the present invention is to provide a method for detecting surface defects of transparent coils that is robust to glare, so as to solve the problem of interference of glare on defect detection. In the winding process, transparent coils, as typical non-Lambertian materials, are often easily affected by glare on their surfaces, which poses a great challenge to defect detection based on computer imaging technology. Existing methods have poor stability when identifying and separating glare areas, and often produce false alarms due to glare interference. The present invention proposes a stable and fast detection method to effectively overcome the adverse effects of glare on the detection of surface defects of transparent coils and improve detection stability and speed.

[0021] A method for detecting surface defects of a transparent coil that is robust to glare comprises the following steps:

[0022] S1. After the winding system is turned on, a split-focal plane polarization camera equipped with a double telecentric lens is used to capture the polarization mosaic image of the i-th frame under glare conditions, and the polarization images I0, I1 at four angles of 0°, 45°, 90°, and 135° are obtained through interpolation calculation. 45 , I 90 , I 135 .

[0023] S2. Use the formula S0=0.5*(I0+I 45 +I 90 +I 135 ), calculate the intensity image S0 of the i-th frame; use the formula AoP = 0.5*arctan(S2*S1) to calculate the polarization phase angle image AoP of the i-th frame, where S1 = I0-I 90 , S2=I 45 -I 135 .

[0024] S3. Perform Gaussian filtering on the polarization phase angle image AoP, binarize the filtered image using the Otsu method, detect connected areas in the binary image, draw the bounding box of the largest connected area, and obtain the polarization phase angle image AoP′ containing the bounding box. Utilize the characteristic that the polarization phase angle image is sensitive to the glare on the surface of the transparent coil to obtain the area D of the surface glare.

[0025] S4, subtract the surface glare area D from the intensity image, perform Gaussian filtering on the remaining part of the image, use Otsu's method to binarize after filtering, detect the connected area in the binary image, draw the bounding box of the largest connected area, and obtain the intensity image containing the bounding box. When the bounding box of the largest connected area is detected in the intensity image of the i-th frame, the winding system stops running and outputs the surface defect detection result.

[0026] Example:

[0027] like Figure 1 , which is a schematic diagram of the overall detection method according to an embodiment of the present invention.

[0028] like Figure 2 As shown in the figure, after the winding system is turned on, a split-focal plane polarization camera equipped with a double telecentric lens is used to capture the polarization mosaic image of the i-th frame under glare conditions, and the polarization images I0, I1 at four angles of 0°, 45°, 90°, and 135° are obtained through interpolation calculation. 45 , I 90 , I 135 .

[0029] like Figure 3 As shown, using the formula S0 = 0.5*(I0+I 45 +I 90 +I 135 ), calculate the intensity image S0 of the i-th frame, and use the formula AoP = 0.5*arctan(S2 / S1) to calculate the polarization phase angle image AoP of the i-th frame, where S1 = I0-I 90 , S2=I 45 -I 135 .

[0030] like Figure 4 As shown, the polarization phase angle image AoP is Gaussian filtered, where the Gaussian kernel is 5*5 and the standard deviation is 2. The filtered image is binarized using the Otsu method, and the connected area is detected in the binary image. The bounding box of the largest connected area is drawn to obtain the polarization phase angle image AoP′ containing the bounding box. The polarization phase angle image is sensitive to the glare on the surface of the transparent coil, and the surface glare area D is obtained.

[0031] like Figure 5 As shown, first, the intensity image is subtracted from the surface glare area D, and the remaining part of the image is Gaussian filtered, where the Gaussian kernel is 5*5 and the standard deviation is 2. After filtering, the Otsu method is used for binarization, and the connected area is detected in the binary image. The bounding box of the largest connected area is drawn to obtain the intensity image containing the bounding box. Since the bounding box of the largest connected area is detected in the intensity image of the i-th frame, the winding system stops running and outputs the surface defect detection result diagram S0′.

Claims

1. A method for detecting surface defects of transparent coils that is robust to glare. It is characterized in that The steps include: Step 1: After the winding system is turned on, use a split-focal plane polarization camera equipped with a double telecentric lens to capture the polarization mosaic image of the i-th frame under flare conditions, and obtain the polarization images I at four angles of 0°, 45°, 90°, and 135° through interpolation calculation. 0 , I 45 , I 90 , I 135 ; Step 2: Using formula S 0 =0.5*(I 0 +I 45 +I 90 +I 135 ), calculate the intensity image S of the i-th frame 0 ; Using the formula AoP = 0.5*arctan(S 2 / S 1 ), calculate the polarization phase angle image AoP of the i-th frame, where S 1 =I 0 -I 90 , S 2 =I 45 -I 135 ; Step 3: Perform Gaussian filtering on the polarization phase angle image AoP, binarize the filtered image, detect connected areas in the binary image, draw the bounding box of the largest connected area, obtain the polarization phase angle image AoP′ containing the bounding box, and obtain the surface glare area D; Step 4: Intensity Image S 0 Subtract the surface glare area D, perform Gaussian filtering on the remaining part of the image, binarize it after filtering, detect the connected area in the binary image, draw the bounding box of the largest connected area, and obtain an intensity image containing the bounding box; when the bounding box of the largest connected area is detected in the intensity image of the i-th frame, the winding system stops running and outputs the surface defect detection result.

2. A method for detecting surface defects of transparent coiled materials that is robust to glare according to claim 1, It is characterized in that The binarization was performed using the Otsu method.

3. A method for detecting surface defects of transparent coiled materials that is robust to glare according to claim 1, It is characterized in that When Gaussian filtering is performed on the polarization phase angle image AoP, the Gaussian kernel is 5*5 and the standard deviation is 2.

Images