CN104680509A - Real-time circular printing image defect detecting method - Google Patents

Abstract

The invention relates to the technical field of processing of images, in particular to a real-time circular printing image defect detecting method. The method comprises the following steps: a standard template graph and an on-line real-time graph of a target to be detected, obtaining a circle point coordinate through an exponential changing and searching step-length method, calculating a circle center and semi-diameter of a circular area of each of the standard template graph and the on-line real-time graph according to a method for positioning a circle through three points, dividing each circular area into a large quantity of annular areas, counting characteristics of grey level histograms of the annular areas, and comparing the characteristics of the grey level histograms of the annular areas of the standard template graph and the on-line real-time graph to complete image defect detection. Through the adoption of the real-time circular printing image defect detecting method, the circular printing image defection speed is improved from two aspects; a practicable method is provided for on-line defect detection based on machine vision.

Description

A Real-time Circular Printing Image Defect Detection Method

technical field

The invention relates to the technical field of image processing, in particular to a real-time circular printing image defect detection method.

Background technique

With the development of industrial technology, the production of products has basically realized assembly line operation. The production speed of circular printed matter such as bottle caps, badges, etc. is generally fast, but due to random errors in equipment, parts, etc., some products may be defective. In order to improve the qualified rate of products and the degree of automation of industrial production, there is an urgent need for a printing defect detection method that can be realized in real time. From the perspective of image matching, the present invention judges whether there is a printing defect in a real-time photographed product according to the similarity between the online real-time image and the standard template image.

There is generally any rotation between the real-time image and the template image on the pipeline, which brings certain difficulties to the application of traditional image matching methods. The existing circular printing image defect detection method is to transform the real-time image and the template image to the same direction for similarity comparison after determining the center, radius and rotation angle of the real-time image. Among them, the circle center and radius are calculated according to the principle of three-point positioning circle after the point on the circle is determined, so the positioning of the point on the circle is the key. The general method uses the gradient change of a row or a column in the binary image to determine the point on the circle on the row or column. This method requires the calculation of the gradient value of the entire row or column of data, and then compares the position coordinates of the points on the circle, which requires a large amount of calculation. The calculation accuracy and speed of the rotation angle and the rotation operation of the image seriously restrict the improvement of the algorithm efficiency.

Aiming at the above two problems of the existing detection methods, the present invention proposes a fast positioning method of points on a circle and an image matching method based on the histogram feature of the circular area. The fast positioning method of the point on the circle changes the search step exponentially to find the point on the circle according to the reference point in the circle, which greatly reduces the number of points involved in the calculation and improves the positioning speed. The rotation invariance of the histogram feature of the annular area can avoid the cumbersome calculation process of the rotation angle, and improve the performance and speed of the algorithm.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention proposes a fast location method for points on a circle and an image matching method based on the histogram feature of the ring area. The fast positioning method of the point on the circle changes the search step exponentially to find the point on the circle according to the reference point in the circle, which greatly reduces the number of points involved in the calculation and improves the positioning speed. The rotation invariance of the histogram feature of the annular area can avoid the cumbersome calculation process of the rotation angle, and improve the performance and speed of the algorithm.

The technical scheme that the present invention adopts for realizing the above object is:

A real-time circular printing image defect detection method, which obtains the standard template map and online real-time map of the target to be detected; obtains the coordinates of points on the circle by exponentially changing the search step method; calculates the standard template map and the online real-time map according to the three-point positioning circle method The center and radius of the circular area of the real-time graph; the circular area is divided into multiple circular areas, and the gray histogram characteristics of the circular area are counted; the gray histogram characteristics of the circular area of the standard template map and the online real-time graph are compared, and the completion Image defect detection.

The exponentially changing search step method includes the following steps:

Perform adaptive threshold binarization on standard template images and online real-time images;

Select any point in the circle as a reference point to determine the 360-degree search direction;

Determine the initial value of the search step and change it exponentially to obtain the coordinates of the point on the circle.

The search directions are preferably four search directions of up, down, left and right.

The exponential change process is:

When the exponential increase condition is met, the search step increases exponentially;

When the exponential increase condition is not satisfied, the search step size decreases exponentially.

The conditions for increasing the index are:

First, the current point is within the image range;

Secondly, the gray value of the current point satisfies the condition of the point inside the circle;

Finally, the proportion of points inside the circle among all points between the current point and the reference point is greater than the credible threshold.

The annular area is concentric with the center of the circular area.

The present invention has the following beneficial effects and advantages:

1. When searching for a point on the circle according to the reference point, the search step changes exponentially, which greatly reduces the amount of calculation and improves the positioning speed of the point on the circle;

2. The histogram feature of the annular region proposed by the present invention has rotation invariance, which overcomes the disadvantage that the traditional image matching method must calculate the rotation angle between the template image and the real-time image, and simplifies the detection process.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Fig. 2 is the program flow chart of point fast location on the circle of the present invention;

Fig. 3 is the template diagram of the beer bottle cap of the embodiment of the present invention;

Fig. 4 is the sample diagram of the defective beer bottle cap of the embodiment of the present invention;

Fig. 5 is the binarization result figure of the beer bottle cap template figure of the embodiment of the present invention;

Fig. 6 is a diagram of the result of concentric ring segmentation of a circular area according to an embodiment of the present invention;

Fig. 7 is a partial online real-time diagram and a template diagram of the embodiment of the present invention, wherein a, b, c, and e are online real-time diagrams of defective beer bottle caps; d is a template diagram of beer bottle caps; f, g, and h are non-defective beer Online real-time graph of bottle caps.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings, taking beer bottle cap defect detection as an example. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

Fig. 1 shows a flow chart of the realization of the circular printing image defect detection method provided by the embodiment of the present invention.

The concrete steps of the detection method that the embodiment of the present invention provides are as follows:

1. Obtain the standard circular printing template image (hereinafter referred to as the template image) and the online circular printing real-time image (hereinafter referred to as the real-time image);

The image of the beer bottle cap template is shown in Figure 3, and the real-time image of defects is shown in Figure 4, represented by tempI and testI respectively, and the image size is size=m*n, where m=240, n=320.

2. Perform adaptive threshold binarization on the template image and the real-time image respectively, and segment the circular image area and the background;

In the actual operation process, the conditions for obtaining real-time images online can generally be manually intervened. The specific method is: when the gray scale of the circular printing image itself is bright, the background can be set to a dark background, otherwise the background can be set to a bright background . In this example, the top gray of the beer bottle cap is brighter, and the sides are darker, forming an automatic separation of the circular target area from the background.

The adaptive threshold is calculated through the statistical histogram, and its criterion is: the mean gray values of the two categories of the histogram divided by the threshold are equal. Then use this threshold to binarize the image, the circular area is 1, and the background area is 0; the binarization result of this example is shown in Figure 5, which is represented by tempIBinary.

3. Roughly determine any point in the circle as a reference point to search, and change the search step size exponentially to realize the rapid positioning of the point on the circle, so as to realize the real-time determination of the center and radius of the circle;

Usually, in order to save costs and improve the detection speed of the algorithm, the circular area in the real-time image accounts for more than 60% of the entire image, so the center of the image is generally considered to be within the circle.

The flow chart of the fast positioning program for the point on the circle is shown in Figure 2. When searching to the right of the reference point, the steps are as follows:

1) First, determine the center point of the image [120,160] as the reference point referP in the circle, and then initialize the position oldcurrentP and newcurrentP of the current point based on this point, and initialize the search step to 2 (the initial search step can also be 3 or 1.5, etc. For other initial values, the general practice is to define the initial search step size as 2);

2) Restore newcurrentP.x to oldcurrentP.x, update newcurrentP.x=newcurrentP.x+step;

3) Judging whether newcurrentP.x is in the image, that is, judging whether 0<newcurrentP.x<320 is true, if true, go to the next step, otherwise go to 7);

4) Judging whether newcurrentP is in the circle, that is, whether the binarization result here is 1, it is to go to the next step, otherwise go to 7);

5) Statistics of the data on the referP.y line in the binarized image, the number of points numall whose abscissa is between newcurrentP.x and the reference point referP.x, and the number of points num1 which is 1, judge num1/numall> Whether thresh is established, if established, go to the next step, otherwise go to 7);

6) Update oldcurrentP.x=newcurrentP.x, while step grows exponentially, step=step*2; then go to 2);

7) The index decreases step, so that step=step/2; go to the next step;

8) Determine whether the step is reduced to a number less than 1, if step>=1, then go to 2), otherwise, end.

In this example, the radius of the circular area is about 90. When searching for points on the circle according to the above method, it only takes 15 iterations at most to complete the search for a point; if the normal method is used, at least 320 calculations are required for the entire row search Compare with.

When searching to the left of the reference point, the principle is the same, except that when updating newcurrentP.x and oldcurrentP.x, the position where the step was originally added is changed to a subtraction operation.

Select k reference points in the circle and search in the left and right directions to get 2*k points on the circle.

The center and radius of the circle can be calculated by the formula for determining a circle with three-point coordinates.

Since the circular edge in the binarization result is not absolutely smooth, we take the mean value of multiple calculation results as the final result of the center and radius. In this example, a total of 10 points on the circle are taken, and they are divided into two groups on average. Three points are selected for every 5 points to calculate the center and radius of the circle. Finally, the average value of the 20 groups of results is used as the final result of the center and radius of the circle.

4. According to the determined center and radius, the circular image is divided into multiple non-overlapping concentric ring regions, and the gray histogram is counted;

In this example, the circular area is divided into 6 circular and annular areas, as shown in Figure 6. The radius of the inner and outer boundary circles of each circle and annular area is shown in the following table:

circle, number of rings Inner radius Outer circle radius center circle 0 10 second ring 11 20 third circle twenty one 30 fourth circle 31 50 fifth circle 51 60 sixth circle 61 80

Table 1 shows the radius of the inner and outer boundary circles of each circle and ring (unit, pixel)

The histogram statistical characteristics of each region were normalized by the number of points that the corresponding region participated in the histogram statistics.

Assume that the statistical results of the normalized histogram of the template graph and the real-time graph are represented by tempHistgram and testHistgram respectively, and tempHistgram[i,:],i=1,2,...,6 represent the statistical histogram characteristics of the first ring of the template graph; Similarly testHistgram[i,:], i=1,2,...,6 represents the statistical histogram features of the first ring of the real-time graph.

5. Calculate the similarity of the histogram features of each corresponding annular area between the template map and the real-time map;

The formula for calculating the similarity between the template image and the i-th ring histogram features of the real-time image is as follows:

$\mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 266</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\sqrt{\mathrm{<span\; class="notranslate">\; tempHistgram</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; testHistgram</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ]</span>}$

6. Judging whether there is a defect in the real-time printed image through the similarity.

Comparing the similarity with the threshold vector, if the condition is satisfied, the real-time image is considered to have no defects, otherwise the real-time image is judged to be a defective image. The judgment conditions in this example are as follows:

$\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.83</span>}\\ \mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.94</span>}\\ \mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.95</span>}\\ \mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.96</span>}\\ \mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.96</span>}\\ \mathrm{<span\; class="notranslate">\; similarity</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0.97</span>}\end{array}\right]$

The following table is the matching result of each sub-image in Figure 7 and Figure 3, and the images and corresponding numbers are listed in Figure 7.

Table 2 shows the matching results of each subgraph in Figure 7 and the template graph

7. The program runs on the 32-bit Windows 7 operating system and Microsoft Visual C++ 6.0 platform, and the detection time is 3-5ms, which not only meets the real-time requirements, but also lays a foundation for further improving productivity.

Claims (6)

1. A real-time circular printing image defect detection method, characterized in that, Obtain the standard template map and online real-time map of the target to be detected; Obtain the coordinates of points on the circle by exponentially changing the search step method; Calculate the center and radius of the circular area of the standard template map and the online real-time map according to the three-point positioning circle method; Divide the circular area into multiple annular areas, and count the gray histogram characteristics of the annular areas; Compare the gray histogram features of the circular area of the standard template image and the online real-time image to complete the image defect detection. 2. a kind of real-time circular printing image defect detection method according to claim 1, is characterized in that: exponentially changing search step size method comprises the following steps: Perform adaptive threshold binarization on standard template images and online real-time images; Select any point in the circle as a reference point to determine the 360-degree search direction; Determine the initial value of the search step and change it exponentially to obtain the coordinates of the points on the circle. 3. A real-time circular printing image defect detection method according to claim 2, characterized in that: said search directions are preferably four search directions of up, down, left and right. 4. A kind of real-time circular printing image defect detection method according to claim 2, is characterized in that: described exponential change process is: When the exponential increase condition is met, the search step increases exponentially; When the exponential increase condition is not met, the search step size decreases exponentially. 5. A kind of real-time circular printing image defect detection method according to claim 4, is characterized in that: described index increase condition is: First, the current point is within the image range; Secondly, the gray value of the current point satisfies the condition of the point inside the circle; Finally, the proportion of points inside the circle among all points between the current point and the reference point is greater than the credible threshold. 6 . The real-time circular printing image defect detection method according to claim 1 , wherein the annular area is concentric with the center of the circular area. 7 .