Intelligent control method for plate dipping and drying machine

Abstract

The present invention relates to the field of graphic recognition technology, and proposes an intelligent control method for a plate dipping and drying machine, comprising: capturing grayscale images of the plate before and after dipping; performing edge detection on the two grayscale images to obtain two edge images, performing difference on the edge images to obtain an edge difference map, extracting a high gradient threshold and a low gradient threshold in the edge detection process of the grayscale image after dipping; extracting the edge point distribution degree and the edge point aggregation degree according to the edge point distribution in the edge difference map, and extracting the edge point target degree, and combining the low gradient threshold to obtain the neighboring gradient threshold of each edge point; obtaining the final defect edge detection result according to the neighboring gradient threshold of each edge point in the edge difference map, identifying the uneven dipping area, and completing the intelligent control of dipping. The present invention aims to solve the problem that the conventional edge detection double threshold is fixed, resulting in the loss of some weak edges and affecting the dipping recognition result.

Description

Technical Field

[0001] The invention relates to the technical field of pattern recognition, and in particular to an intelligent control method for a board dipping and drying machine. Background Art

[0002] The board dipping and drying machine needs to judge whether the dipping condition is intact and enter the drying stage according to the dipping condition. If there is an uneven dipping area, it will greatly affect the final dipping quality of the board and the dipping needs to be continued. Therefore, it is necessary to identify the dipping defects of the board during the dipping process and judge whether to continue dipping based on the identification results. Since there are obvious grayscale mutations and gradient changes in the uneven dipping area, the existing technology extracts the edges of the uneven dipping area through edge detection, thereby identifying the uneven dipping area and completing the judgment.

[0003] In edge detection, the canny operator determines the edge points through the image gradient matrix and double thresholds. However, the fixed global double threshold will cause the loss of weak edge parts with small gradients in some areas with uneven glue dipping, resulting in edge breakage, which in turn affects the identification of glue dipping defects. Therefore, it is necessary to quantify the target degree of edge points according to the distribution of edge points, so as to achieve adaptive adjustment of the low threshold of adjacent gradients for edge points, and re-extract edge points, thereby improving the accuracy of edge detection results, ensuring that uneven glue dipping areas can be identified, and better controlling the glue dipping of panels. Summary of the invention

[0004] The present invention provides an intelligent control method for a plate glue dipping and drying machine to solve the problem that the existing double thresholds for edge detection are fixed, resulting in the failure to completely extract weak edges and affecting the dipping recognition result. The technical solution adopted is as follows:

[0005] An embodiment of the present invention provides an intelligent control method for a plate dipping and drying machine, the method comprising the following steps:

[0006] Capture the grayscale image of the board before and after dipping;

[0007] According to the edge detection results of the grayscale image before and after dipping, the edge difference map and the gradient low threshold corresponding to the grayscale image after dipping are extracted;

[0008] According to the distribution of edge points in the edge difference map, the edge point distribution degree and edge point aggregation degree of each edge point are extracted; through the two indicators of the edge point distribution degree and edge point aggregation degree of each edge point in the edge difference map, the edge point target degree and the adjacent gradient threshold of each edge point are extracted;

[0009] The final defect edge detection result is obtained according to the neighboring gradient threshold of each edge point in the edge difference map, the uneven glue dipping area is identified, and the intelligent glue dipping control is completed.

[0010] Furthermore, the specific method of extracting the edge difference map and the gradient low threshold corresponding to the grayscale image after dipping includes:

[0011] Perform edge detection on the grayscale image before and after dipping in glue, obtain the edge image before and after dipping in glue, and record the gradient low threshold value corresponding to the grayscale image after dipping in glue;

[0012] The edge image after dipping in glue is differentiated from the edge image before dipping in glue. For the pixel points at the same position, the edge image after dipping in glue is subtracted from the edge image before dipping in glue. The edge points that are edges in both edge images in the difference result are set to 0, the edge points that exist alone in the edge image after dipping in glue are retained, and the edge points that exist alone in the edge image before dipping in glue are set to 0, and the retained edge points are set to 1 to obtain the difference result of the two edge images, which is recorded as the edge difference map.

[0013] Furthermore, the method of obtaining the edge image before and after dipping in glue and recording the gradient low threshold corresponding to the grayscale image after dipping in glue includes:

[0014] The edge detection of the grayscale image before and after dipping is completed according to the canny operator, and the two edge images obtained are respectively recorded as the edge image before and after dipping;

[0015] During the edge detection process of the grayscale image after dipping in glue, the low threshold in the double threshold determination is recorded to obtain the gradient low threshold corresponding to the grayscale image after dipping in glue.

[0016] Furthermore, the method of extracting the edge point distribution degree and the edge point aggregation degree of each edge point according to the distribution of edge points in the edge difference map includes the following specific methods:

[0017] Extract the connected domains in the edge difference map, record the number of edge points in each connected domain, use the number of edge points in the connected domain where each edge point is located as the edge distribution parameter of each edge point, normalize the edge distribution parameters of all edge points in the edge difference map, and record the result as the edge point distribution degree of each edge point;

[0018] Extract the edge point clustering degree of each edge point.

[0019] Furthermore, the edge point aggregation degree of each edge point is obtained by:

[0020] Take any edge point as the target edge point, take the target edge point as the center, and obtain the circle with a radius of r corresponding to the target edge point, where , represents the area of ​​the edge difference map, Indicates the number of edge points;

[0021] The number of pixels contained in the target edge point circle is , the number of edge points contained in the circle is , get the maximum number of edge points, expressed as , the calculation method is ;

[0022] The ratio of the number of edge points in the target edge point circle to the maximum number of edge points in the circle is recorded as the edge point aggregation degree of the target edge point, and the edge point aggregation degree of each edge point is obtained.

[0023] Furthermore, the edge point target degree and the adjacent gradient threshold of each edge point are specifically obtained by:

[0024] The product of the edge point distribution degree and the edge point aggregation degree of each edge point is used as the edge point target degree of each edge point; the product of the edge point target degree of each edge point and the gradient low threshold corresponding to the grayscale image after dipping is used as the neighboring gradient threshold of each edge point.

[0025] Furthermore, the method of obtaining the final defect edge detection result according to the neighboring gradient threshold of each edge point in the edge difference map includes the following specific methods:

[0026] Compare the gradient of the neighborhood pixel point with the neighborhood gradient threshold of the corresponding edge point. If it is greater than the neighborhood gradient threshold, it is considered a weak edge point and connected to the strong edge point; if it is less than or equal to the neighborhood gradient threshold, it does not need to be connected;

[0027] Weak edge point determination is performed on the neighboring pixel points of each edge point in the grayscale image after dipping in the edge difference map. When a new weak edge point appears, the neighboring gradient threshold of the new weak edge point is calculated and the determination is continued until no new weak edge points are added, thereby obtaining the final defect edge detection result.

[0028] The beneficial effects of the present invention are as follows: the present invention performs edge detection on the captured grayscale images before and after dipping in glue, and performs differentiation on the edge detection results to obtain an edge difference map, in which only the edge part newly added after dipping in glue is retained, and the newly added edge part is more likely to correspond to the uneven dipping area; the target degree of the edge point is quantified according to the distribution of the edge points in the edge difference map, and then the neighboring gradient threshold is obtained, wherein the target degree of the edge point is extracted by two indicators, the distribution degree of the edge point and the aggregation degree of the edge point, the distribution degree of the edge point is quantified by the size of the connected domain where the edge point is located, and the aggregation degree of the edge point is quantified by the concentrated distribution of the edge points, and the neighboring gradient threshold of each edge point is adjusted by the edge point target degree, so that the double threshold in the edge detection can select the low threshold adaptively according to the distribution of the edge points, thereby improving the defect edge recognition result, making the final obtained uneven dipping area more accurate, thereby improving the working efficiency of the intelligent control of dipping in glue. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0030] Figure 1 A schematic flow chart of an intelligent control method for a plate dipping and drying machine provided in accordance with an embodiment 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] See also Figure 1 , which shows a flow chart of an intelligent control method for a plate dipping and drying machine provided by an embodiment of the present invention, the method comprising the following steps:

[0033] Step S001, capturing grayscale images of the board before and after dipping in glue.

[0034] The purpose of this embodiment is to identify defects in the image of the plate after dipping in glue, determine whether there is an area of ​​uneven dipping in glue, and then determine whether to continue dipping in glue. Therefore, it is necessary to first obtain images of the plate before and after dipping in glue; in this embodiment, the plate dipping desiccant is composed of dipping equipment, an industrial camera, a processing unit and a drying equipment. The images of the plate before and after dipping in glue are captured by the industrial camera, and the two captured images are preprocessed. In this embodiment, the image preprocessing is grayscale processing to obtain a grayscale image of the plate before dipping in glue and a grayscale image of the plate after dipping in glue. It should be noted that the shooting angles before and after dipping in glue are the same, and the shots are taken from the feed port and the discharge port of the dipping process respectively to ensure that there will be no image differences due to environmental factors.

[0035] At this point, the grayscale images of the board before and after dipping in glue are obtained.

[0036] Step S002, performing edge detection on the two grayscale images to obtain two edge images, performing difference on the edge images to obtain an edge difference map, and extracting a high gradient threshold and a low gradient threshold in the process of edge detection of the grayscale images after dipping.

[0037] It should be noted that after the board is dipped in glue, the dip will make the board surface more glossy, and the grayscale value of the grayscale image after dipping will be larger than that of the grayscale image before dipping. However, normal dipping will only affect the overall grayscale of the board surface and will not cause gradient changes. Only when uneven areas of dipping appear, due to the difference in dipping thickness, the grayscale image on the board surface will show a sudden change in grayscale and then a gradient change. Therefore, the grayscale images before and after dipping can be used for edge detection, and the texture and edges of the board itself can be removed by differentiating the edge detection results, such as the tree texture on the board surface, to obtain an edge difference map. At the same time, in the edge detection process of the grayscale image after dipping, the double thresholds obtained by the canny operator are extracted to provide a basis for the subsequent improvement of the defect edge detection results.

[0038] Specifically, firstly, edge detection is performed on the two grayscale images before and after dipping in glue by using the Canny operator. During the edge detection process, the Canny operator needs to select a double threshold, wherein the double threshold extraction method is as follows: taking the grayscale image after dipping in glue as an example, after obtaining the gradient of each pixel in the image, a gradient statistical histogram is obtained, and the frequency of the gradient is accumulated in order from small to large. When the accumulated frequency reaches a preset first threshold, the corresponding gradient is used as the high gradient threshold of the grayscale image after dipping in glue. In this embodiment, the preset first threshold is described as 0.8, and the product of the high gradient threshold and the preset second parameter is used as the low gradient threshold. In this embodiment, the preset second parameter is described as 0.4, and the high gradient threshold and the low gradient threshold of the grayscale image after dipping in glue are obtained; the high gradient threshold and the low gradient threshold of the grayscale image before dipping in glue are extracted in the same way, and the edge detection of the grayscale image before dipping in glue and the grayscale image after dipping in glue is completed according to the subsequent algorithm of the Canny operator, and the two edge images obtained are respectively recorded as the edge image before dipping in glue and the edge image after dipping in glue.

[0039] Furthermore, the edge image after immersion in glue and the edge image before immersion in glue are differentiated, and the edge image before immersion in glue is subtracted from the edge image after immersion in glue for the pixel points at the same position. The edge points that are edges in both edge images in the difference result are set to 0, and the edge points that exist alone in the edge image after immersion in glue are retained, and the edge points that exist alone in the edge image before immersion in glue also have the results set to 0. The retained edge points are set to 1, and the difference result of the two edge images can be obtained, which is recorded as the edge difference map.

[0040] At this point, the gradient high threshold and gradient low threshold in the edge detection process of the grayscale image after dipping are extracted, and the edge difference map is obtained. The edge difference map only retains the edge points that exist alone in the edge image after dipping, which may be some edge points in the uneven dipping area.

[0041] Step S003: extract the edge point distribution degree and edge point aggregation degree according to the edge point distribution in the edge difference map, extract the edge point target degree, and combine the gradient low threshold to obtain the neighboring gradient threshold of each edge point.

[0042] It should be noted that after obtaining the edge difference map, it is necessary to perform edge recognition on other pixel points in the neighborhood of each edge point according to the distribution of edge points in the edge difference map, and adjust the gradient low threshold corresponding to the grayscale image after dipping by analyzing the indicator of edge point target degree; the edge point target degree is extracted by two indicators: edge point distribution degree and edge point aggregation degree. The edge point distribution degree is quantified by the size of the connected domain where the edge point is located. The fewer the number of edge points in the connected domain where the edge point is located, the smaller the range of the connected domain is, and it is necessary to further verify whether there is an unrecognized edge that causes the connected domain to become smaller, then the possibility of containing unrecognized edges is greater; at the same time, the edge point aggregation degree is quantified by the concentrated distribution of edge points. The more concentrated the edge point distribution is, the more drastic the regional grayscale mutation is, and the smaller the possibility of the existence of unrecognized edges is; the smaller the possibility of the existence of unrecognized edges in the neighborhood, the greater the edge point target degree, that is, there is no need to reduce the gradient low threshold too much to perform edge judgment on the pixels in the neighborhood.

[0043] Specifically, firstly, connected domains are extracted from the edge difference map, the number of edge points in each connected domain is recorded, and the number of edge points in the connected domain where each edge point is located is used as the edge distribution parameter of each edge point. If the edge point is distributed on the dividing line of two connected domains, the mean number of edge points in the two connected domains is used as the edge distribution parameter of the edge point. The edge distribution parameters of all edge points in the edge difference map are linearly normalized, and the result is recorded as the edge point distribution degree of each edge point.

[0044] Furthermore, a circle with a radius of r and an arbitrary edge point as the center is selected as the evaluation range of the edge point aggregation degree of the edge point, and the number of other edge points in the circle is counted to quantify the edge point aggregation degree. It should be noted that for the radius r, its size selection is more important for the measurement of the aggregation degree. If the radius is too large, it is easy to cause all circles to contain more edge points, and it is impossible to judge the aggregation degree of a certain area. If the radius is too small, it may cause all circles to contain fewer edge points, which will also make the judgment of the density wrong. Therefore, in this embodiment, the area of ​​the edge difference map is evenly distributed to each edge point, and the radius ,in Represents the area of ​​the edge difference map, expressed as the number of pixels in the edge difference map. represents the number of edge points. For any edge point, the number of pixels contained in the corresponding circle is , the number of edge points contained in the circle is ; For the circle corresponding to the edge point, the number of edge points in the edge difference map is , the number of pixels in the circle is , then theoretically the maximum number of edge points contained in the circle is and The minimum value of , then the maximum number of edge points in the circle is the number of pixels in the circle, that is, all pixels in the circle are edge points; if the total number of edge points If it is less than the number of pixels in the circle, the maximum number of edge points in the circle is the total number of edge points, that is, all edge points are in the circle. The maximum number of edge points is obtained, which is expressed as , the calculation method is .

[0045] Furthermore, taking any edge point as an example, the number of edge points within the edge point circle is , the maximum number of edge points within the circle is , the ratio of the number of edge points in the edge point circle to the maximum number of edge points in the circle is recorded as the edge point aggregation degree of the edge point, and the edge point aggregation degree of each edge point is extracted according to the above method.

[0046] Furthermore, taking any edge point as an example, the product of the edge point distribution degree and the edge point aggregation degree of the edge point is taken as the edge point target degree of the edge point, and the edge point target degree of each edge point is extracted according to the above method; the product of the edge point target degree of each edge point and the gradient low threshold corresponding to the grayscale image after immersion is taken as the adjacent gradient threshold of each edge point, which is used for subsequent defect edge improvement.

[0047] At this point, by analyzing the concentrated distribution of edge points and the range of connected domains, the target degree of edge points is obtained, and then the neighboring gradient threshold of each edge point in the edge difference map is obtained.

[0048] Step S004, obtaining the final defect edge detection result according to the neighboring gradient threshold of each edge point in the edge difference map, identifying the non-uniform glue dipping area, and completing the intelligent glue dipping control.

[0049] After obtaining the neighboring gradient threshold, according to the correspondence between the edge point in the edge difference map and the corresponding pixel point in the grayscale image after immersion, the eight neighboring pixel points of each edge point in the grayscale image after immersion can be obtained, and then the gradient of each neighboring pixel point can be obtained; the gradient of the neighboring pixel point is compared with the neighboring gradient threshold of the corresponding edge point. If it is greater than the neighboring gradient threshold, it can be regarded as a weak edge point and connected with the strong edge point to improve the defect edge; if it is less than or equal to the neighboring gradient threshold, it does not need to be connected. It should be noted that after the neighboring gradient threshold is corrected around an edge point, if multiple pixels with gradients greater than the neighboring gradient threshold appear, the pixel with the largest gradient value is selected as the weak edge point for connection; according to the above method, the weak edge point judgment is performed on the neighboring pixel points of each edge point in the grayscale image after immersion in the edge difference map. When a new weak edge point appears, the neighboring gradient threshold of the new weak edge point is calculated according to the above method and the judgment is continued until no new weak edge points are added, and the final defect edge detection result is obtained.

[0050] Furthermore, after obtaining the defect edge detection result, the uneven glue dipping area can be identified according to the defect edge detection result, and whether the glue dipping needs to be continued is judged by the proportion of the uneven glue dipping area to the area of ​​the plate, wherein the area of ​​the uneven glue dipping area and the area of ​​the plate are directly obtained, and this embodiment sets a preset third threshold for judgment. The preset third threshold in this embodiment is described as 0.1, and if the area proportion is greater than the preset third threshold, it is necessary to continue the glue dipping to remove the uneven glue dipping area, and if it is less than or equal to the third threshold, the glue dipping can be completed, and the glue dipping desiccant can be controlled to dry the glue dipping surface of the plate to complete the intelligent control of the glue dipping.

[0051] At this point, the intelligent control of the board impregnation and the subsequent processing of the impregnated board by the desiccant have been completed.

[0052] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent control method for a plate impregnation dryer, characterized in that: The method comprises the following steps: Capture the grayscale image of the board before and after dipping; According to the edge detection results of the grayscale image before and after dipping, the edge difference map and the gradient low threshold value corresponding to the grayscale image after dipping are obtained; According to the distribution of edge points in the edge difference map, the edge point distribution degree and the edge point aggregation degree of each edge point are obtained; through the two indicators of the edge point distribution degree and the edge point aggregation degree of each edge point in the edge difference map, the edge point target degree and the adjacent gradient threshold of each edge point are obtained; The final defect edge detection result is obtained according to the adjacent gradient threshold of each edge point in the edge difference map, the uneven glue dipping area is identified, and the intelligent glue dipping control is completed; The edge point target degree and the adjacent gradient threshold of each edge point are specifically obtained by: The product of the edge point distribution degree and the edge point aggregation degree of each edge point is used as the edge point target degree of each edge point; the product of the edge point target degree of each edge point and the gradient low threshold corresponding to the grayscale image after dipping is used as the neighboring gradient threshold of each edge point; Among them, the final defect edge detection result is obtained according to the neighboring gradient threshold of each edge point in the edge difference map, and the specific method includes: comparing the gradient of the neighborhood pixel point with the neighborhood gradient threshold of the corresponding edge point, and considering the point greater than the neighborhood gradient threshold as a weak edge point and connecting it with the strong edge point; if it is less than or equal to the neighborhood gradient threshold, it does not need to be connected; for each edge point in the edge difference map, the neighborhood pixel points in the grayscale image after dipping are determined as weak edge points, and when a new weak edge point appears, the neighborhood gradient threshold of the new weak edge point is calculated and the determination is continued until no new weak edge points are added, and the final defect edge detection result is obtained; The method for obtaining the degree of edge point distribution is as follows: extracting the connected domains in the edge difference map, recording the number of edge points in each connected domain, taking the number of edge points in the connected domain where each edge point is located as the edge distribution parameter of each edge point, normalizing the edge distribution parameters of all edge points in the edge difference map, and recording the obtained result as the degree of edge point distribution of each edge point; The method for obtaining the edge point aggregation degree of each edge point is as follows: taking any edge point as the target edge point, taking the target edge point as the center of the circle, and obtaining a circle with a radius r corresponding to the target edge point, where , represents the area of ​​the edge difference map, Indicates the number of edge points; The number of pixels contained in the target edge point circle is , the number of edge points contained in the circle is , get the maximum number of edge points, expressed as , the calculation method is ; The ratio of the number of edge points in the target edge point circle to the maximum number of edge points in the circle is recorded as the edge point aggregation degree of the target edge point, and the edge point aggregation degree of each edge point is obtained.

2. The intelligent control method of a plate dipping and drying machine according to claim 1 is characterized in that: The specific method of obtaining the edge difference map and the gradient low threshold corresponding to the grayscale image after dipping includes: Perform edge detection on the grayscale image before and after dipping in glue, obtain the edge image before and after dipping in glue, and record the gradient low threshold value corresponding to the grayscale image after dipping in glue; The edge image after dipping in glue is differentiated from the edge image before dipping in glue. For the pixel points at the same position, the edge image after dipping in glue is subtracted from the edge image before dipping in glue. The edge points that are edges in both edge images in the difference result are set to 0, the edge points that exist alone in the edge image after dipping in glue are retained, and the edge points that exist alone in the edge image before dipping in glue are set to 0, and the retained edge points are set to 1 to obtain the difference result of the two edge images, which is recorded as the edge difference map.

3. The intelligent control method of a plate dipping and drying machine according to claim 2 is characterized in that: The specific method of obtaining the edge image before and after dipping and recording the gradient low threshold corresponding to the grayscale image after dipping is as follows: The edge detection of the grayscale image before and after dipping is completed according to the canny operator, and the two edge images obtained are respectively recorded as the edge image before and after dipping; During the edge detection process of the grayscale image after dipping in glue, the low threshold in the double threshold determination is recorded to obtain the gradient low threshold corresponding to the grayscale image after dipping in glue.

Images