CN105092597B - A kind of crack detecting method on hard plastic material surface - Google Patents

Abstract

The invention discloses a kind of crack detecting methods on hard plastic material surface, comprise the following steps：S1 obtains hard plastic material surface gray level image；S2 carries out binary conversion treatment to the gray level image acquired and obtains bianry image, extracts the slit region ROI image in the gray level image, the ROI is made to be separated with the background in gray level image；S3 obtains the single pixel framework information in ROI image；S4 carries out pixel point analysis to the single pixel framework information of ROI image, the product for meeting crack is marked.The present invention can quickly, precisely detect hard plastic material face crack, the on-line checking available for hard plastic material face crack.

Description

A Crack Detection Method on the Surface of Hard Plastic Material

technical field

The invention relates to a method for detecting cracks on the surface of hard plastic materials, in particular to a method for detecting cracks on the surface of devices using hard plastic materials as product packaging materials.

Background technique

As a commonly used component of objects, hard plastic materials, such as packaging of electronic devices, protective cases of medical equipment, daily necessities, etc., can even be used in the field of construction, and its application range is very wide. With the excessive consumption of traditional materials , A large number of synthetic hard plastic materials have been developed to make up for the shortage. The quality of hard plastic material products is not critical, and surface cracks are an important factor in the quality of hard plastic materials. During use, cracks may lead to electric leakage, water leakage, etc., reducing the overall life of the device, and even leading to accidents. Since the amount of hard plastic materials brings considerable economic benefits and continues to increase, in order to ensure the quality of hard plastic materials, it is extremely important to find a crack detection method to adapt to advanced production methods.

With the continuous enhancement of computer performance, digital image processing technology is also developing rapidly. In recent years, a number of crack detection methods based on digital image processing have emerged, mainly removing noise and detecting cracks according to the different characteristics of cracks and noise; based on BP neural network The algorithm realizes the detection of cracks. The crack feature is used as the input factor of the neural network, and the output is used as the criterion for judging the crack through learning. However, the structure of the neural network is too complex and the calculation complexity is large, which is not conducive to online detection; according to the linear characteristics of cracks, Some algorithms decompose the crack into different line segments, and judge the crack by judging the complexity of the slope, but in the cracks on the surface of hard plastic materials, many cracks are caused by gravity extrusion, and the cracks are networked, and the linearity is very complicated, which is not conducive to linearity. The crack detection algorithm that combines fuzzy logic and artificial neural network uses the crack features as the input of the model, and judges the crack by the output result. In addition, there are crack detection algorithms based on wavelet transform, detection methods based on histogram statistical analysis, and detection methods based on the combination of histogram projection and morphological operators. The above crack detection methods are aimed at cracks on the surface of concrete and roads. Such cracks are caused by corrosion and heavy rolling, and the crack edges have obvious burrs. Cracks on the surface of hard plastic materials are mainly caused by physical factors such as extrusion and impact. Due to the material characteristics of hard plastic materials, there are fewer burrs on the crack edges. In addition, the commonly used circularity discrimination method cannot exclude linear non-crack defects such as scratches, so the above method is not suitable for the detection of cracks on the surface of hard plastic materials.

With the development of productivity, advanced production equipment can produce a large amount of finished hard plastic materials in a short period of time. In order to ensure production efficiency and improve product inspection quality, it is very important to find a detection method that can adapt to the production environment. The detection algorithm is required to have Rapidity, stability and efficiency.

Contents of the invention

The technical problem to be solved by the present invention is to provide a detection method capable of quickly and accurately detecting cracks on the surface of hard plastic materials.

In order to solve the above technical problems, the present invention takes the following technical solutions:

A method for detecting cracks on the surface of a hard plastic material, comprising the following steps:

S1, acquiring the grayscale image of the surface of the hard plastic material;

S2, performing binarization processing on the obtained grayscale image to obtain a binary image, extracting the ROI image of the crack area in the grayscale image, and separating the ROI from the background in the grayscale image;

S3, acquiring single-pixel skeleton information in the ROI image;

S4, performing pixel point analysis on the single-pixel skeleton information of the ROI image, and marking products conforming to crack features.

When the grayscale image is binarized, the infiltration algorithm is used to process it, which specifically includes the following steps:

S2.1, set the maximum window to M×M, the current window size to N×N, set the pixel _in the center of the current window as the seed pixel, and the seed pixel is a point in the penetration area Dp , from the seed The pixel starts to penetrate, and the brightness I(ps) of the seed pixel is set as the initial threshold T;

S2.2, the 8-neighborhood of the penetration area D _p is D _c , judge the size relationship between the brightness of the pixel in the penetration area D _c and the initial threshold T, if the brightness value of a pixel in the 8-neighborhood D _c If it is smaller than the initial threshold T, the pixel belongs to the penetration area D _p , otherwise the pixel belongs to the background D _b ;

S2.3. Determine whether the penetration area D _p has reached the current window boundary. If the penetration area D _p has not reached the current window boundary, return to step S2.2 to continue the current penetration process; if the penetration area D _p reaches the current window boundary, expand the current window. The size of the window, so that N=N+2;

S2.4, take the expanded window as the current window, start a new round of pixel infiltration, and judge the relationship between the pixel brightness in the 8-neighborhood D _c of the current infiltration area D _p and the initial threshold T; if If there is a pixel in D _c whose luminance is less than the initial threshold T, the pixel is assigned to the infiltration area _Dp ; if there is no pixel in D _c whose luminance is less than the initial threshold T, the infiltration process ends;

S2.5, judge whether the current window size is larger than the maximum window size, that is, whether N is larger than M, if it is larger, the infiltration process is ended, if it is smaller, otherwise return to step S2.3, and perform a new round of infiltration process until the infiltration ends, and obtain Binarize the image and extract the ROI in the image.

The luminance value of the pixel in the window is updated and calculated by the following formula, wherein I ( p ) represents the luminance of the pixel:

.

The acquisition of single-pixel skeleton information in the ROI image specifically includes:

Each pixel in the ROI in the binarized image is matched with 8 preset elimination templates. If there is no identical elimination template, the current pixel is retained; if there is the same elimination template, the The current pixel is matched with 9 reserved templates. If there is the same reserved template, the current pixel is retained. If there is no identical reserved template, the current pixel is deleted until all the pixels in the ROI are matched one by one. Get a single-pixel skeleton of the ROI.

In the step S4, the 8-neighborhood determination method is specifically used for crack determination. If the single-pixel skeleton contains pixels with a number of neighbor points greater than 2, the current hard plastic material surface has cracks.

The invention can quickly and accurately detect cracks on the surface of hard plastic materials, and is especially suitable for the detection method of cracks on the surface of devices (such as current transformers) used as product packaging materials, thereby improving the quality of products.

Description of drawings

Accompanying drawing 1 is the schematic flow chart of the present invention;

Accompanying drawing 2 is a schematic diagram of the processing flow of the penetration algorithm of the present invention;

Accompanying drawing 3 is the schematic diagram of the grayscale image of the crack on the surface of the hard plastic material;

Accompanying drawing 4 is the schematic diagram of the binarization image after binarization processing among the present invention;

Accompanying drawing 5 is a schematic diagram of the field of extraction of current pixels in the present invention;

Accompanying drawing 6 is the schematic diagram of the elimination template predefined in the present invention;

Accompanying drawing 7 is the schematic diagram of the pre-defined retention template of the present invention;

Accompanying drawing 8 is a schematic diagram of a single-pixel skeleton obtained in the present invention;

Accompanying drawing 9 is a schematic diagram of 8 neighborhoods of pixels in the present invention;

Figure 10 is a schematic diagram of the determination result of the neighborhood of the single pixel skeleton 8 having cracks in the present invention.

Detailed ways

In order to facilitate the understanding of those skilled in the art, the present invention will be further described below in conjunction with the accompanying drawings.

As shown in accompanying drawing 1～5, a kind of crack detection method on the surface of hard plastic material comprises the following steps:

S1, acquiring the grayscale image of the surface of the hard plastic material. The target product to be detected is photographed to obtain its grayscale image, as shown in Figure 3.

S2, perform binarization processing on the obtained grayscale image to obtain a binary image, extract the ROI image of the crack region in the grayscale image, and separate the ROI from the background in the grayscale image, as shown in Figure 4 . The ROI image is the target crack area in the grayscale image.

S3. Obtain the single-pixel skeleton information in the ROI image, as shown in FIG. 8 , which is the pixel skeleton information with the background removed.

S4. Perform pixel point analysis on the single-pixel skeleton information of the ROI image, and mark products that meet the crack characteristics. Specifically, the 8-neighborhood judgment method is used for crack judgment. If the single-pixel skeleton contains more than 2 neighborhood points Pixels, the current hard plastic material has cracks on the surface. As shown in Figure 10, it is a product with cracks.

In addition, when the grayscale image is binarized, the permeation algorithm is used for processing, as shown in Figure 2, which specifically includes the following steps:

S2.1, set the maximum window to M×M, the current window size to N×N, set the pixel _in the center of the current window as the seed pixel, and the seed pixel is a point in the penetration area Dp , from the seed The pixel starts to penetrate, and the brightness I(ps) of the sub pixel is set as the initial threshold T.

S2.2, the 8-neighborhood of the penetration area D _p is D _c , the size relationship between the brightness of the pixels in the penetration area D _c and the initial threshold T, if the brightness value of a pixel in the 8-neighborhood D _c If it is smaller than the initial threshold T, the pixel belongs to the penetration area D _p , otherwise the pixel belongs to the background D _b . With the seed pixel point as the center point, it starts to infiltrate to the surrounding continuously, and compares the brightness value of the pixel point with the initial threshold T every time a pixel point is infiltrated.

S2.3. Determine whether the infiltration area Dp has reached the boundary of the current window. If the infiltration area Dp has not _reached the boundary of the current window, that is, all the pixels in the current window have not been infiltrated, return to _step S2.2 to continue the current infiltration process ; If the infiltration area D _p reaches the boundary of the current window, that is, all the pixels in the current window have been infiltrated, then expand the size of the current window to make N=N+2, that is to say, the size of the original window is N×N Expand to (N+2)×(N+2), taking +2 as the base.

During the penetration process, the brightness values of different pixels are calculated by downloading the column formula, where I ( p ) represents the brightness of the pixel, that is, the brightness value is calculated every time a pixel is passed, and then compared with the initial threshold T.

.

S2.4, take the expanded window as the current window, start a new round of pixel infiltration, and judge the relationship between the pixel brightness in the 8-neighborhood D _c of the current infiltration area D _p and the initial threshold T; if If there is a pixel in the 8-neighborhood D _c whose brightness is less than the initial threshold T, the pixel is assigned to the infiltration area _Dp ; if there is no pixel in the 8-neighborhood D _c whose luminance is less than the initial threshold T, the infiltration ends The process shows that there are no pixels belonging to the penetration area D _p after expanding the window.

S2.5, judge whether the current window size is larger than the maximum window size, that is, whether N is larger than M, if it is larger, the infiltration process is ended, if it is smaller, otherwise return to step S2.3, and perform a new round of infiltration process until the infiltration ends, and obtain Binarize the image and extract the ROI in the image. N here is an enlarged numerical value. That is to say, if the size of the expanded window after the first round of step S2.3 and step S2.4 is already greater than the maximum window size, then the infiltration process ends. If the window size after the first round of expansion through steps S2.3 and S2.4 is smaller than the maximum window size, then expand the current window again, so that the size of the current window is increased by 2. Then return to steps S2.3 and S2.4 for a new round of infiltration. In this way, after multiple rounds of infiltration of pixels, the grayscale image is converted into a binary image, the ROI is separated from the background, and the contrast between the ROI and the background is enhanced, as shown in Figures 3 and 4.

For obtaining the single-pixel skeleton information in the ROI image, the improved OPTA algorithm is used to obtain it. The specific process of the improved OPTA algorithm is as follows, in which the current detection pixel is set as P , and the adjacent pixel is Q , and the extraction is shown in Figure 5 Field, in the deletion template and the retention template, 1 represents the pixel point of ROI, and 0 represents the background point.

Each pixel in the ROI in the binarized image is matched with 8 preset elimination templates. If there is no identical elimination template, the current pixel is retained; if there is the same elimination template, the The current pixel is matched with 9 reserved templates. If there is the same reserved template, the current pixel is retained. If there is no identical reserved template, the current pixel is deleted until all the pixels in the ROI are matched one by one. Get a single-pixel skeleton of the ROI.

That is to say, first compare all the pixels of the ROI in the current binary image with the 8 predefined elimination templates shown in Fig. 9 reserved templates are used for matching, otherwise the pixel is reserved. Then continue to match with the 9 pre-set reserved templates shown in Figure 7. If there is a pixel that matches a certain template, then keep the pixel, otherwise set the pixel as the background point, from the pixel skeleton delete. Finally, it is determined whether all pixels have been matched with the elimination template and the retention template, if yes, then end, if not, continue to match until all pixel points are matched. In this way, through the above-mentioned multiple rounds of iterations, a single-pixel skeleton as shown in FIG. 8 is obtained.

Finally, for the judgment of cracks, the surface of hard plastic materials contains defects such as blocky spots, oil stains, and cracks. The following judgment steps are given according to the geometric characteristics of different defects.

According to the difference of the geometric shape of the defect, the circularity is used to remove the blocky defect from the ROI extracted by the penetration algorithm, and finally obtain the ROI containing only the linear region. The circularity is shown in the following formula. Where P represents the perimeter of the ROI, S is the area of the ROI, and C is the circularity. The circularity is used to judge the linear complexity. The circularity of a point or sheet area is close to 1. The larger the circularity, the more complex the linearity of the area. Set the circularity threshold as T _circle , and delete the ROI whose C is smaller than T _circle .

Skeleton extraction is performed on the ROI. Due to the particularity of hard plastic materials and the mechanism of crack formation, the cracks on the surface of hard plastic materials are slender, complex, and have bifurcation points. The improved OPTA is used to obtain the single-pixel skeleton information in the crack area, and the crack skeleton is analyzed and judged by the 8-neighborhood point discrimination method. N2, ..., N8 are 8 neighborhood points. The selected crack image is shown in Figure 8 and the scratch image is shown in Figure 10. According to the characteristics of bifurcation points in cracks, if there are pixels with more than 2 neighbor points in the skeleton image, it is judged as a crack, that is, as shown in Figure 10, there are pixels with more than 2 neighbor points in Figure 8 There are two bifurcations, and the pixel points at the center of the two bifurcations have two neighbor points, indicating that the surface image of the hard plastic material contains cracks. Of course, in practice, there may be 3, 4 or more neighboring points, as long as the number of neighboring points is greater than 2, it is determined to have cracks.

It should be noted that the above description is not a limitation to the technical solution of the present invention, and any obvious replacements are within the protection scope of the present invention without departing from the inventive concept of the present invention.

Claims (3)

1. A crack detection method on the surface of a hard plastic material, comprising the following steps: S1, acquiring the grayscale image of the surface of the hard plastic material; S2, performing binarization processing on the acquired grayscale image to obtain a binary image, and processing with a penetration algorithm, extracting the ROI image of the crack area in the grayscale image, and separating the ROI from the background in the grayscale image, Calculate the circularity of the ROI, Where P represents the perimeter of the ROI, S is the area of the ROI, and C is the circularity. The circularity threshold is set to Tcircle, and the ROI with C smaller than Tcircle is deleted; S3, obtain the single-pixel skeleton information in the ROI image, and use the improved OPTA algorithm to obtain, specifically including: Each pixel in the ROI in the binarized image is matched with 8 preset elimination templates. If there is no identical elimination template, the current pixel is retained; if there is the same elimination template, the The current pixel is matched with 9 reserved templates. If there is the same reserved template, the current pixel is retained. If there is no identical reserved template, the current pixel is deleted until all the pixels in the ROI are matched one by one. Get the single-pixel skeleton of ROI; S4, performing pixel point analysis on the single-pixel skeleton information of the ROI image, and marking products that meet the crack characteristics; When the grayscale image is binarized, the infiltration algorithm is used to process it, which specifically includes the following steps: S2.1, set the maximum window to M×M, the current window size to N×N, set the pixel in the center of the current window as the seed pixel, and the seed pixel is a point in the penetration area _Dp , from the seed The pixel starts to penetrate, and the brightness I(ps) of the seed pixel is set as the initial threshold T; S2.2, the 8-neighborhood of the penetration area D _p is D _c , judge the size relationship between the brightness of the pixel in the penetration area D _c and the initial threshold T, if the brightness value of a pixel in the 8-neighborhood D _c If it is smaller than the initial threshold T, the pixel belongs to the penetration area D _p , otherwise the pixel belongs to the background D _b ; S2.3. Determine whether the penetration area D _p has reached the current window boundary. If the penetration area D _p has not reached the current window boundary, return to step S2.2 to continue the current penetration process; if the penetration area D _p reaches the current window boundary, expand the current window. The size of the window such that N=N+2; S2.4, take the expanded window as the current window, start a new round of pixel infiltration, and judge the relationship between the pixel brightness in the 8-neighborhood D _c of the current infiltration area D _p and the initial threshold T; if If there is a pixel in D _c whose luminance is less than the initial threshold T, the pixel is assigned to the infiltration area _Dp ; if there is no pixel in D _c whose luminance is less than the initial threshold T, the infiltration process ends; S2.5, judge whether the current window size is larger than the maximum window size, that is, whether N is larger than M, if it is larger, the infiltration process is ended, if it is smaller, otherwise return to step S2.3, and perform a new round of infiltration process until the infiltration ends, and obtain Binarize the image and extract the ROI in the image. 2. the crack detection method on the hard plastic material surface according to claim 1, is characterized in that, the luminance value of the pixel point in the described window is updated and calculated by following formula, and wherein I (p) represents the pixel point luminance: <mrow><mi>T</mi><mo>=</mo><mi>m</mi><mi>a</mi><mi>x</mi><mrow><mo>(</mo><munder><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow><mrow><mi>p</mi><mo>&amp;Element;</mo><msub><mi>D</mi><mi>p</mi></msub></mrow></munder><mo>(</mo><mrow><mi>I</mi><mrow><mo>(</mo><mi>p</mi><mo>)</mo></mrow></mrow><mo>)</mo><mo>,</mo><mi>T</mi><mo>)</mo></mrow><mo>+</mo><mi>w</mi><mo>.</mo></mrow> 3. The crack detection method on the surface of a hard plastic material according to claim 2, characterized in that, in the step S4, specifically adopting the 8-neighborhood determination method to determine the crack, if the single-pixel skeleton contains a number of neighborhood points If the number of pixels is greater than 2, the surface of the current hard plastic material has cracks.