Region-first double-view-field glass fiber pultrusion plate defect detection method

Abstract

The invention discloses a region-first double-view-field glass fiber pultrusion plate defect detection method, and belongs to the technical field of plate defect detection.The method comprises the steps that a to-be-detected glass fiber pultrusion plate image is preprocessed, and a preprocessed image is obtained; wherein the to-be-detected image is a glass fiber pultrusion plate image acquired by a cross stroboscopic image acquisition system formed by two lighting fields of reflected light and backlight; based on the preprocessed image, judging whether the glass fiber pultrusion plate image has oscillation or not, and if the image does not have oscillation, carrying out area edge searching on the left and right boundaries of the plate and cutting the area of the plate; and based on the cut image, obtaining a defect position and category result in the glass fiber pultrusion plate image. By adopting the scheme, the defect detection efficiency, the identification effect and the automation degree of the glass fiber pultrusion plate can be improved, and the high-quality development of the glass fiber pultrusion plate industry is promoted.

Description

Technical Field

[0001] The invention relates to the technical field of plate defect detection, and in particular to a method for detecting defects of glass fiber pultruded plates with dual viewing fields and with regional priority. Background Art

[0002] Fiber-reinforced plastic is a common industrial product, which is very different from tempered glass. It includes carbon fiber reinforced composite plastic, glass fiber reinforced composite plastic, boron fiber reinforced composite plastic, etc. It has the characteristics of light weight and hard texture, high mechanical strength and wide application range. Fiberglass pultruded sheet, also known as FRP, is a typical sheet and is widely used in the wind turbine blade industry. The quality of FRP products is related to the length of time they serve on site and the effect of their application. Therefore, the detection of their defects has also been the focus of the industry for a long time. The types of defects of FRP mainly include cracking, color difference, inclusions, fiber contamination, fiber joints, etc. These common defects often exist during the production of FRP sheets, which can easily lead to more serious mechanical damage and secondary problems during transportation and on-site service, which seriously hinders the application, promotion and development of this product in the sheet industry.

[0003] In order to detect defects in FRP sheets, the industry usually uses infrared thermal imaging, laser ultrasound, terahertz non-destructive testing, manual inspection and other methods for detection. Among them, the manual inspection method is usually limited by the working time, comprehensiveness of detection and the working status of personnel. The detection accuracy is poor and it is very easy to miss the plate surface. Infrared thermal imaging, laser ultrasound, terahertz non-destructive testing and other methods often require the installation of precise detection devices, and the subsequent maintenance is difficult and the maintenance cost is high. There are few types of defects that can be effectively detected, and it is difficult to be applied on a large scale in product production sites. Therefore, how to find a method to improve the efficiency of FRP defect detection and ensure the quality of products before leaving the factory has become an urgent problem that the industry needs to solve.

[0004] In recent years, the research on machine vision-based target detection methods has been gaining popularity, and the method of using images for real-time target detection has become an important means of detecting quality defects of various types of plates. In order to meet the needs of internal defect detection of FRP, dual-field cross-strobe imaging has gradually been used, generally including reflection field imaging and backlight field imaging. Compared with conventional lighting methods, this method can reduce the number of cameras required, but the combined image obtained by this dual-field cross-strobe lighting method sometimes has image oscillation problems due to abnormalities such as on-site jitter and electromagnetic interference. Therefore, in order to reduce the problem of false detection caused by oscillation, it is very important to determine whether typical oscillation occurs in the image.

[0005] In addition, the range of the plate in the whole image is limited in reality, and the background area in the image does not belong to the scope of defect detection. The operation of searching the plate area based on the idea of ​​area priority can effectively reduce the complexity of the system and enhance the reliability of the detection system. Unlike other common plates, the dual-field imaging results of FRP generally contain backlight imaging content. Its plate area positioning and defect detection are significantly different from common reflected light images. Therefore, how to distinguish the lighting field attributes of the odd-line image and the even-line image in the dual-field combined image, and find the effective area range of the plate in the image for defect detection, has become an important guarantee for the accuracy and real-time performance of FRP product detection. At present, in the field of FRP defect detection, there are relatively few algorithms based on target detection, and there are few methods that can be applied to actual production on a large scale. However, in actual applications, the detection method is often required to have high detection real-time performance and a certain ability to handle abnormal situations. Therefore, how to design a detection method that can detect most common defects in FRP plates and has the characteristics of low cost, simple installation and maintenance, good detection real-time performance, and a certain ability to handle abnormal situations is of great significance in the production process of products. Summary of the invention

[0006] The present invention provides a region-prioritized dual-field-of-view glass fiber pultruded plate defect detection method to solve the technical problems of the prior art, such as high cost, complex installation and maintenance, and poor real-time detection.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In one aspect, the present invention provides a method for detecting defects of a glass fiber pultruded plate with dual fields of view and region priority, and the method comprises:

[0009] Preprocessing the image of the glass fiber pultruded plate to be detected to obtain a preprocessed image; wherein the image of the glass fiber pultruded plate to be detected is an image of the glass fiber pultruded plate acquired by a cross-stroboscopic image acquisition system formed by two lighting fields of reflected light and backlight;

[0010] Based on the preprocessed image, determine whether the glass fiber pultruded plate image has oscillation. If the image does not have oscillation, perform edge finding and plate area cropping operations on the left and right boundaries of the plate based on the preprocessed image.

[0011] Based on the cropped image, the defect locations and categories in the glass fiber pultruded sheet image are obtained.

[0012] Furthermore, the image of the glass fiber pultruded plate to be inspected is preprocessed, including:

[0013] The image of the glass fiber pultruded plate to be tested is converted into a grayscale image, and then all the odd-numbered rows of data in the converted grayscale image are sequentially spliced ​​into a sub-image as the first lighting field image, and all the even-numbered rows of data in the converted grayscale image are sequentially spliced ​​into another sub-image as the second lighting field image.

[0014] Further, judging whether the glass fiber pultruded plate image has oscillation based on the preprocessed image, and if the image does not have oscillation, performing edge finding and plate area cropping operations on the left and right boundaries of the plate based on the preprocessed image, includes:

[0015] Step 1, set the gradient threshold thre_G, the effective threshold of the lighting field thre_O, the backlight oscillation threshold thre_T, the super bright threshold thre_S, the oscillation number threshold thre_C, the binarization threshold thre_B and the brighter threshold thre_L;

[0016] Step 2, performing gradient edge search on the first lighting field image and the second lighting field image respectively, and then determining the left boundary pixel position le_0 and the right boundary pixel position re_0 of the first lighting field image, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of the second lighting field image according to thre_G;

[0017] Step 3, if |(re_0-le_0)-(re_1-le_1)|>thre_O, it is determined that the image does not oscillate, and the process goes to step 4; otherwise, it is determined that the image oscillates, and the process ends;

[0018] Step 4: if (re_0-le_0)>(re_1-le_1), the first lighting field image is determined as a backlight image, and the second lighting field image is determined as a reflected light image; otherwise, the first lighting field image is determined as a reflected light image, and the second lighting field image is determined as a backlight image;

[0019] Step 5, judging whether the backlight image is oscillating according to thre_T and thre_C, if the backlight image is not oscillating, go to step 6; otherwise, end the program;

[0020] Step 6, setting the grayscale values ​​of pixels in the backlight image whose grayscale values ​​are greater than thre_S to 0, and then performing binarization and expansion operations on the first lighting field image and the second lighting field image according to thre_B to obtain the processed first lighting field image and the processed second lighting field image;

[0021] Step 7, according to thre_L, find the area where the plate is located in the first and second light-emitting field images after processing, and perform a plate area cropping operation according to the determined area where the plate is located.

[0022] Further, gradient edge search is performed on the first lighting field image and the second lighting field image respectively, and then the left boundary pixel position le_0 and the right boundary pixel position re_0 of the first lighting field image, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of the second lighting field image are obtained according to thre_G, including:

[0023] Performing lateral gradient calculation based on the Sobel operator on the first lighting field image and the second lighting field image respectively to obtain a gradient grayscale image corresponding to the first lighting field image and a gradient grayscale image corresponding to the second lighting field image;

[0024] Find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from left to right in the gradient grayscale image corresponding to the first lighting field image, as the left boundary pixel position le_0 of the first lighting field image, and find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from right to left in the gradient grayscale image corresponding to the first lighting field image, as the right boundary pixel position re_0 of the first lighting field image;

[0025] Find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from left to right in the gradient grayscale image corresponding to the second light field image, as the left boundary pixel position le_1 of the second light field image, and find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from right to left in the gradient grayscale image corresponding to the second light field image, as the right boundary pixel position re_1 of the second light field image;

[0026] If there is no pixel gradient greater than thre_G in the image, it is determined that there is no obvious gradient change in the corresponding image, and the corresponding left boundary pixel position and right boundary pixel position are both set to 0.

[0027] Further, judging whether the backlight image is oscillating according to thre_T and thre_C includes:

[0028] Extract two rows of data from the backlight image, and then search the two rows of data from left to right at the same time to find the first pixel with a grayscale value greater than thre_T in the two rows of data, and then count how many pixels in the column where this pixel is located have a grayscale value less than thre_T. If the statistical result is less than thre_C, it is determined that the backlight image does not oscillate; otherwise, it is determined that the backlight image oscillates.

[0029] Further, binarization and expansion operations are performed on the first lighting field map and the second lighting field map according to thre_B to obtain the processed first lighting field map and the processed second lighting field map, including:

[0030] Divide the first lighting field image horizontally into 5 segments, then take the maximum value of the grayscale mean of each of the 5 segments, multiply it by thre_B, and use it as the grayscale corresponding to the first lighting field image. Figure 2 The actual threshold of the value, and then the gray value of the first light field image is greater than or equal to the gray value corresponding to the first light field image Figure 2 The pixel grayscale value of the actual threshold of the quantization is set to 255, which is smaller than the grayscale value corresponding to the first light field image. Figure 2 The pixel grayscale value of the actual threshold of the binarization is set to 0, and the binarization of the first light field image is completed. Then, a dilation operation with a kernel size of 3×3 is performed on the binarized first light field image to obtain a processed first light field image;

[0031] Divide the second lighting field image horizontally into 5 segments, then take the maximum value of the grayscale mean of each of the 5 segments, multiply it by thre_B, and use it as the grayscale corresponding to the second lighting field image. Figure 2 The actual threshold of the value, and then the gray value of the second light field image is greater than or equal to the gray value corresponding to the second light field image Figure 2 The pixel grayscale value of the actual threshold of the quantization is set to 255, which is smaller than the grayscale value corresponding to the second light field image. Figure 2 The pixel grayscale value of the actual threshold of the binarization is set to 0, and the binarization of the second lighting field map is completed. Then, a dilation operation with a kernel size of 3×3 is performed on the binarized second lighting field map to obtain a processed second lighting field map.

[0032] Further, according to thre_L, searching for the area where the plate is located in the processed first light field image and the second light field image includes:

[0033] Take out one line of data from every four lines of the processed first light field image to form a new grayscale image, recorded as the first grayscale image, and then average the grayscale values ​​of the first grayscale image vertically to obtain a line of data representing the grayscale distribution characteristics of the first grayscale image, and filter out the horizontal position range of pixels corresponding to the area with a grayscale value greater than thre_L, and take the area corresponding to the maximum horizontal position range of pixels as the area where the plate is located;

[0034] For the processed second lighting field image, one line of data is taken out every four lines to form a new grayscale image, which is recorded as the second grayscale image. The grayscale values ​​of the second grayscale image are then averaged vertically to obtain a line of data representing the grayscale distribution characteristics of the second grayscale image. The horizontal position range of pixels corresponding to the area with a grayscale value greater than thre_L is screened out, and the area corresponding to the maximum horizontal position range of pixels is taken as the area where the plate is located.

[0035] Furthermore, based on the cropped image, the defect locations and categories in the glass fiber pultruded sheet image are obtained, including:

[0036] The cropped image of the reflected light image is scaled, and then uniformly cropped horizontally to obtain multiple sub-images of the same height as reflected light sub-images;

[0037] The cropped backlight image is scaled, and then uniformly cropped horizontally to obtain multiple sub-images of the same height as backlight sub-images;

[0038] Use the preset reflection field detection model to detect the defect position of the reflected photon image and obtain the detection result;

[0039] Use the preset backlight field detection model to detect the defect position of the backlight sub-image and obtain the detection result;

[0040] The preset classification model is used to classify the local color images corresponding to the detection results output by the reflection field detection model and the backlight field detection model to obtain the defect locations and categories in the glass fiber pultruded sheet images.

[0041] On the other hand, the present invention further provides an electronic device, comprising a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the above method.

[0042] In yet another aspect, the present invention further provides a computer-readable storage medium, wherein at least one instruction is stored in the storage medium, and the instruction is loaded and executed by a processor to implement the above method.

[0043] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0044] The scheme of the present invention first converts the FRP color composite image obtained by the reflection field and backlight field double-field cross-stroboscopic acquisition device into a grayscale image and performs odd-even row splitting, uses the lighting field judgment logic to clarify the attributes of the reflection field image and the backlight field image, and then performs oscillation attribute judgment on the image, and then performs edge search and cropping operations on the left and right boundaries of the plate on these two images, and finally uses the detection algorithm and classification algorithm to obtain the location and category results of the defects in the FRP plate image. The scheme of the present invention can improve the detection efficiency, recognition effect and automation degree of defects in glass fiber pultruded plates, and promote the high-quality development of the glass fiber pultruded plate industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0046] Figure 1 It is a schematic diagram of the execution flow of the area-priority dual-field-of-view glass fiber pultruded plate defect detection method provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of the odd-even splitting effect of the grayscale combination graph of the glass fiber reinforced plastic plate provided by an embodiment of the present invention;

[0048] Figure 3 is a schematic diagram of the gradient edge finding effect provided by an embodiment of the present invention;

[0049] Figure 4 The grayscale provided by the embodiment of the present invention Figure 2 Schematic diagram of the effect after valorization and expansion;

[0050] Figure 5 is a schematic diagram of the effect of the plate area after cutting provided by an embodiment of the present invention;

[0051] Figure 6 is a schematic diagram of the effects of scaling and sub-image clipping of a plate area provided by an embodiment of the present invention;

[0052] Figure 7 is a schematic diagram of the effect of defect detection and classification of glass fiber reinforced plastic sheet provided by an embodiment of the present invention;

[0053] Figure 8 It is a system block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0055] First of all, it should be noted that in the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplarily" is intended to present the concept in a concrete way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0056] First embodiment

[0057] The present embodiment provides a method for defect detection of glass fiber pultruded plates with dual fields of view and region priority. The method is a target detection algorithm based on machine vision, and the object to be processed is a color composite image of FRP acquired by a dual field of view cross-strobe system based on the reflected field and the backlight field. The FRP dual field of view cross-strobe acquisition system can image the surface and interior of FRP. The FRP reflected light image represents the morphology of the plate surface, and has a lateral brightness distribution characteristic of dark-bright-dark, while the backlight image represents the basic morphology of the interior of the plate. Due to its light-transmitting imaging characteristic, the camera will directly capture the area where the light source is directly illuminated in some areas, so it has a lateral brightness distribution characteristic of dark-extremely bright-bright-extremely bright-dark. According to the lateral brightness imaging characteristics of the two illumination fields, the reflection field and the backlight field, this method preferentially designs the plate area edge finding method of these two illumination images, that is, first convert the FRP color combination image of the reflection field and the backlight field into a grayscale image and split the odd and even lines to obtain two grayscale images, and then judge the illumination field attributes of the two images, and then judge the oscillation attributes of the images, and then use the respective regional edge finding methods for the two images to determine the position of the plate in the image. In terms of FRP regional edge finding, this regional edge finding method has the characteristics of unique results and high accuracy compared with the traditional contour finding method. Then, the plate area of ​​the image is cropped, and then the two plate images are detected by using the corresponding illumination field detection model. Finally, the classification model is used to classify the color image corresponding to the detection result, and accurate defect category information is obtained, which reduces the redundant calculation amount, thereby effectively improving the defect detection efficiency of glass fiber pultruded plates, and thus effectively ensuring the quality and quality of products before leaving the factory.

[0058] The execution process of the dual-field glass fiber pultruded plate defect detection method with priority in this area is as follows: Figure 1 The method may be implemented by an electronic device, which may be a terminal or a server. The method includes:

[0059] S1, preprocessing the image of the glass fiber pultruded plate to be detected to obtain a preprocessed image;

[0060] In this embodiment, the image of the glass fiber pultruded plate to be detected is a dual-field color combination image of the FRP reflection field and the backlight field; the above S1 is to first convert the input FRP reflection field and the backlight field dual-field color combination image into a grayscale image, and then split the odd and even rows to obtain the illuminated field image A and the illuminated field image B.

[0061] Among them, the combined image of the reflected field and the backlight field refers to the original FRP color image when input into this algorithm for processing, which is generally acquired by a cross-stroboscopic image acquisition system formed by the two lighting fields of reflected light and backlight. The use of a color line scan camera can allow the acquired color image to better show the overall picture of the FRP sheet. The reflected light image is the data obtained after the light of the reflected light source is reflected by the surface of the sheet and returns to the camera photosensitive element. It is used to present the morphology of the sheet surface, while the backlight image is the data obtained after the light of the backlight light source directly penetrates the sheet and reaches the camera photosensitive element. It is used to present the morphology of the material inside the sheet.

[0062] Even-odd row splitting refers to sequentially splicing all odd-numbered row data in an image into one sub-image, and sequentially splicing all even-numbered row data into another sub-image.

[0063] It should be noted that converting the input fiberglass color composite image into a grayscale image can reduce the amount of calculation when searching for the target boundary and speed up the target detection speed while retaining the original edge contour information of the image. By splitting and recombining the grayscale image into odd and even rows, two grayscale images representing different lighting fields can be obtained. In this embodiment, a normal fiberglass composite image A and a fiberglass composite image B with occasional oscillations are used for illustration. Figure 2 shown.

[0064] S2, based on the preprocessed image, determining whether the glass fiber pultruded plate image has oscillation, if the image does not have oscillation, performing edge search and plate area cropping operations on the left and right boundaries of the plate;

[0065] S21, setting the gradient threshold thre_G, the effective threshold of the lighting field thre_O, the backlight oscillation threshold thre_T, the super-bright threshold thre_S, the oscillation number threshold thre_C, the binarization threshold thre_B and the brighter threshold thre_L;

[0066] Among them, it should be noted that the gradient threshold thre_G is used to determine the approximate position of the left and right boundaries of the plate in the gradient calculation result image, and the effective threshold of the lighting field thre_O is used to preliminarily determine whether the combined image has oscillation. If it is determined that the image has oscillation, the subsequent algorithm logic is terminated to avoid false detection caused by the oscillating image. The backlight oscillation threshold thre_T is used to determine whether the backlight image has oscillation. If it is determined that the image has oscillation, the subsequent algorithm logic is terminated. The super bright threshold thre_S is used to set the extremely high brightness area in the backlight image to completely black. This operation can make this edge finding logic applicable to plates of different widths. The binary threshold thre_B is used to convert the reflected light image and the backlight image into a binary grayscale image, which has the characteristics of grayscale values ​​of 0 or 255, which is convenient for regional edge finding operations. The brighter threshold thre_L is used to find the boundaries of all the brighter areas of the reflected light image and the backlight image, so as to determine the regional position where the plate may appear in the image. In addition, it should be noted that, according to the actual imaging characteristics of FRP on-site, the above thresholds can be dynamically adjusted to achieve better defect detection effects.

[0067] S22, perform gradient edge search on image A and image B respectively, and then determine the left boundary pixel position le_0 and the right boundary pixel position re_0 of image A, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of image B according to thre_G.

[0068] Among them, it should be noted that according to the imaging characteristics of FRP production and acquisition devices, in the process of finding the area where the plate is located in the image, the upper boundary and the lower boundary of the plate area are the upper boundary and the lower boundary of the image respectively, and the left boundary pixel position and the right boundary pixel position that need to be focused on refer to the pixel horizontal coordinate positions corresponding to the left end and the right end of the plate in the image respectively. The regional edge search operation of the plate in the image plays an important role. On the one hand, it can reduce the redundant calculation amount of the subsequent detection algorithm and improve the reliability of the platform where the algorithm is located. On the other hand, it can eliminate unnecessary false detection problems caused by the junction of the non-plate area and the plate area in the original image. Gradient edge search refers to the use of gradient operators for preliminary horizontal edge detection, and then according to thre_G, find the horizontal coordinate of the leftmost boundary point as the horizontal position of the left boundary of the plate in the image, and find the horizontal coordinate of the rightmost boundary point as the horizontal position of the right boundary of the plate in the image.

[0069] Specifically, in this embodiment, the grayscale image A and the grayscale image B of each of the image A and the image B are respectively calculated based on the Sobel operator, and the gradient grayscale images corresponding to the four images can be obtained. Then, according to thre_G, the horizontal coordinates of the pixels where the threshold grayscale appears for the first time from left to right and from right to left in the gradient grayscale image are found, and the left boundary pixel position le_0 and the right boundary pixel position re_0 of the image A, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of the image B can be obtained. If the pixel grayscale greater than the threshold has not appeared in the image, it is determined that the image has no obvious gradient change, and the corresponding left boundary pixel position and right boundary pixel position are set to 0.

[0070] Based on the above, in this embodiment, when thre_G is set to 40, the gradient graph A and the gradient graph B corresponding to the graph A can obtain le_0=2313, re_0=2321, le_1=735 and re_1=3388, and when thre_G is set to 40, the gradient graph A and the gradient graph B corresponding to the graph B can obtain le_0=1695, re_0=2337, le_1=658 and re_1=3412. Figure 3 shown.

[0071] S23, if |(re_0-le_0)-(re_1-le_1)|>thre_O, it is determined that the image does not oscillate, and the process goes to S24; otherwise, it is determined that the image oscillates, and the process ends;

[0072] It should be noted that the basis for determining that the combined image is oscillating in this embodiment is that part of the backlight image content appears in the reflected light image, and part of the reflected light image content also appears in the backlight image. Since the oscillating image is very likely to cause false detection problems in the detection stage, in order not to affect the detection effect, if it is determined that oscillation occurs, the subsequent processing steps of the combined image are terminated.

[0073] Specifically, in this embodiment, if thre_O=100 is set, since the corresponding |(re_0-le_0)-(re_1-le_1)|=|(2321-2313)-(3388-735)|=2645>thre_O of Figure A, it is initially determined that there is no large-scale oscillation in Figure A, and the corresponding |(re_0-le_0)-(re_1-le_1)|=|(2337-1695)-(3412-658)|=2112>thre_O of Figure B, it is initially determined that there is no large-scale oscillation in Figure B, and both can be transferred to S24.

[0074] S24, if (re_0-le_0)>(re_1-le_1), then image A is determined to be a backlit image, and image B is determined to be a reflected light image; otherwise, image A is determined to be a reflected light image, and image B is determined to be a backlit image;

[0075] It should be noted that, based on the content distribution characteristics of the images collected by the reflection field and the backlight field, the image with a wider bright area is determined to be the image collected by the backlight field, and the image with a narrower bright area is determined to be the image collected by the reflection field.

[0076] Specifically, in this embodiment, (re_0-le_0)=8, (re_1-le_1)=2653 corresponding to Figure A, that is, (re_0-le_0)<=(re_1-le_1), so the grayscale image A of Figure A is determined to be a reflected light image, and the grayscale image B of Figure A is a backlit image. Similarly, (re_0-le_0)=642, (re_1-le_1)=2754 corresponding to Figure B, that is, (re_0-le_0)<=(re_1-le_1), so the grayscale image A of Figure B is determined to be a reflected light image, and the grayscale image B is a backlit image.

[0077] S25, judging whether the backlight image is oscillating according to thre_T and thre_C, if the backlight image is not oscillating, go to S26; otherwise, end the program;

[0078] Among them, it should be noted that the logic of judging whether the backlight image oscillates refers to extracting two rows of data from the backlight image, and then finding the super-bright area based on the backlight oscillation threshold thre_T, and judging how many times the overall grayscale value of the column where the leftmost boundary pixel position is located in these super-bright areas has the characteristic of lower grayscale value of the reflected light image. If the number of occurrences is less than thre_C, go to S26; otherwise, it is judged that the image has frequent oscillations. In order not to affect the detection effect, the subsequent processing steps of the combined image will be terminated.

[0079] Specifically, in this embodiment, it is necessary to determine whether the backlight grayscale image B of Figure A oscillates. If thre_T=250 and thre_C=5 are set, the second row and the second to last row of grayscale data are extracted from the backlight grayscale image B of Figure A. Then, the horizontal coordinate position of the first pixel with a grayscale value greater than thre_T is found from left to right for these two rows of data, and the horizontal coordinate is 741. Then, the 741st column of the grayscale image B of Figure A is searched from top to bottom to find out how many pixels have a grayscale value less than thre_T. It can be obtained that there are 0 places that meet the condition, 0<thre_C, indicating that the grayscale distribution characteristics of the reflected light image do not appear in the entire backlight image. Therefore, it is determined that the combined Figure A corresponding to the backlight grayscale image B does not have obvious oscillation, and jump to S26. Similarly, to determine whether the backlight grayscale map B of Figure B oscillates, set thre_T=250 and thre_C=5. Then extract the second and second to last rows of grayscale data from the backlight grayscale map B of Figure B. Then find the first pixel horizontal coordinate position whose grayscale value is greater than thre_T from left to right for these two rows of data. The horizontal coordinate is 733. Then, search from top to bottom for the total number of pixels whose grayscale values ​​are less than thre_T in the 733rd column of the grayscale map B of Figure B. There are 53 places that meet the conditions, 53>thre_C, indicating that the grayscale distribution characteristics of the reflected light image frequently appear in the entire backlight image. Therefore, it is determined that the combined Figure B corresponding to the grayscale map B has obvious oscillations. Figure B no longer participates in the subsequent algorithm logic, and the program ends.

[0080] S26, set the grayscale values ​​in the backlight image that are greater than thre_S to 0, then perform binarization and dilation operations on Figure A and Figure B according to thre_B, then find the boundaries of all brighter areas in each image according to thre_L, determine that the widest brightness area in the image is the area where the plate is located, and perform plate area cropping operations on Figure A and Figure B according to the area edge search results to obtain the plate area image.

[0081] It should be noted that setting the grayscale of pixels in the backlight image whose grayscale value is greater than the threshold thre_S to 0 can enable the subsequent regional edge-finding logic to effectively calculate the horizontal coordinate positions of the left and right boundaries of the plate for plates of different widths. The widest brightness area is determined to be the area where the plate is located in order to reduce the adverse effects of abnormal conditions such as dust and uneven lighting in the image that originate from the acquisition environment. Binarization refers to setting the grayscale of all pixels in the grayscale image to 0 or 255 according to the threshold thre_B, and the dilation operation can locally expand the size of brighter targets in the image and locally reduce the size of darker targets.

[0082] Specifically, in this embodiment, if thre_S=254, thre_B=0.43, then the grayscale values ​​of pixels in the backlight grayscale image B of Figure A whose grayscale is greater than the thre_S value are set to 0. This operation can weaken the effect of the area edge finding effect of this algorithm when the width of the plate changes. Then the reflected light grayscale image A of Figure A is evenly divided into 5 segments horizontally, and then the maximum value of the grayscale mean of each of the 5 segments is taken and multiplied by thre_B as the grayscale value. Figure 2 The actual threshold of binarization is set, and the grayscale values ​​of pixels whose grayscale values ​​are greater than or equal to the actual threshold of binarization are set to 255, and the grayscale values ​​of pixels whose grayscale values ​​are less than the threshold are set to 0, and then a dilation operation with a kernel size of 3×3 is performed to obtain a grayscale image after binary dilation. Similarly, the above-mentioned binarization and dilation operations are also performed on the backlight grayscale image B of the aforementioned image A that has been partially set to 0, and the result of the backlight grayscale image B of the image A after processing can be obtained, as shown in FIG. Figure 4. Then, take out one line every 4 lines of the processed image of the reflected light grayscale image A of Figure A to form a new grayscale image, and then average the grayscale values ​​of the image vertically to obtain a line of data representing the grayscale distribution characteristics of the image, and then use the threshold thre_L to find out which areas in the line of data have grayscale values ​​greater than the brighter threshold thre_L. For example, after the above operation, the reflected light grayscale image A of Figure A in this embodiment can obtain a set of horizontal position ranges of pixels in the brighter area: [[1340,1343],,[1429,1435],[1493,2711]], because the length of the interval [1493,2711] is the longest among the above set elements, it is determined that the actual horizontal pixel position range of the plate area of ​​the reflected light grayscale image A of Figure A is [1493,2711], and the number of pixels occupied by the plate width is 2711-1493+1=1219. Similarly, after performing the above-mentioned operations of extracting rows, averaging and searching for brighter areas on the backlight grayscale image B of Figure A, the set of horizontal position ranges of pixels in the brighter areas that meet the conditions is obtained as [[693,744],[753,817],[1520,2687],]. Since the length of the interval [1520,2687] is the longest among the above set elements, it is determined that the horizontal pixel position range of the plate area of ​​the backlight grayscale image B of Figure A is [1520,2687], and the number of pixels occupied by the plate width is 2687-1520+1=1168. Generally speaking, the thickness of the edge of the FRP plate is thinner than the thickness in the middle of the plate. Therefore, the edge of the plate in the backlight image is easily transmitted by the light source and appears almost completely white. The edge-finding width in the backlight image is often narrower than that in the reflected light image. Therefore, the edge-finding effect of the reflected light image is relatively more accurate than that of the backlight image. You can choose whether to use the regional edge-finding result of the reflected light image as the actual position of the plate in the reflected light image and the backlight image according to actual needs.

[0083] S3, based on the cropped image, obtain the defect location and category in the glass fiber pultruded sheet image;

[0084] Among them, in this embodiment, the above S3 is to scale the cropped plate image to a certain size, and perform a uniform horizontal cropping operation to obtain a number of sub-images of the same height, and use the two detection models, the reflection field detection model and the backlight field detection model, to respectively perform defect detection operations on the grayscale sub-images of the plate areas of the two lighting fields, and finally use a unified classification model to classify the local color images corresponding to the detection results, and obtain the defect position and category results in the FRP image, and the program ends. Among them, it should be noted that the focus of the FRP defect detection algorithm is the two operations of defect position detection and defect category classification. The two detection models, the reflection field detection model and the backlight field detection model, are to adapt to the differences between the FRP reflected light image and the backlight image. On the basis of the defect position results obtained by these two detection models, a unified classification model is used to classify the local color images corresponding to all detection results, and finally the defect position and defect classification results of the FRP reflected light image and the backlight image are obtained.

[0085] Specifically, in this embodiment, the reflected light grayscale image A and the backlight grayscale image B of the image A are cropped according to the edge finding result of the reflected light grayscale image A of the image A. Since most sites need to detect the "edge yarn" defect near the edge of the plate, it is possible to expand a certain width of the area outside the plate based on the left and right boundaries of the plate as needed, for example, expand the width of 100 pixels to the left and right respectively, and then perform cropping, such as Figure 5 The cropped plate image is scaled to 2048×504 and evenly cropped horizontally to obtain 4 sub-images of 512×504. The uniform scaling and sub-image cropping operations can effectively detect defects with smaller areas, as shown in Figure 1. Figure 6 As shown. YOLOv5 is a common anchor frame-based network structure in the current target detection field. The above sub-image can be detected using the reflection field model and backlight field model based on YOLOv5 to obtain the detection results of the defect position. Inception-V3 is a common neural network used for large-scale classification tasks. The unified Inception-V3 classification model can be used to classify the local defect color images corresponding to the above two detection positions, and finally obtain the position and category information of the defects in each image. The classification results of the defect in the sub-images of the reflected light image and the backlight image of the FRP plate are both "fiber joints", as shown in Figure 7 shown.

[0086] In summary, this embodiment provides a region-prioritized dual-field-of-view glass fiber pultruded plate defect detection method, which first converts the glass fiber reinforced plastic color combination image obtained by the reflection field and backlight field dual-field-of-view cross-stroboscopic acquisition device into a grayscale image and performs odd-even row splitting, uses the lighting field judgment logic to clarify the attributes of the reflection field image and the backlight field image, and then performs oscillation attribute judgment on the image, and then performs regional edge search and cropping operations on the left and right boundaries of the plate on these two images, and finally uses the detection algorithm and classification algorithm to obtain the location and category results of the defects in the glass fiber reinforced plastic plate image. The technical solution of the present invention can improve the detection efficiency, recognition effect and degree of automation of defects in glass fiber pultruded plates, and promote the high-quality development of the glass fiber pultruded plate industry.

[0087] Second embodiment

[0088] This embodiment provides an electronic device, such as Figure 8 As shown, the electronic device includes: a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method of the first embodiment. In addition, the electronic device may also include a transceiver, the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0089] Next, combine Figure 8 The following is a detailed introduction to the various components of the electronic device:

[0090] Among them, the processor is the control center of the electronic device, and the electronic device may include multiple processors, each of which may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may be a processor or a general term for multiple processing elements. For example, the processor is one or more central processing units (CPUs), or other general-purpose processors, application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement an embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (field programmable gate arrays, FPGAs), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor may execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0091] In a specific implementation, as an embodiment, the processor may include one or more CPUs, such as Figure 8 The CPU0 and CPU1 shown in the figure are, of course, only exemplary.

[0092] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0093] Optionally, the memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and accessed through the interface circuit ( Figure 8 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0094] The transceiver may include a receiver and a transmitter ( Figure 8 The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function. The transceiver can be integrated with the processor or exist independently and communicate with the electronic device through the interface circuit ( Figure 8 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0095] In addition, it should be noted that Figure 8 The structure of the electronic device shown in the figure does not constitute a limitation on the device, and the actual device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. In addition, the technical effects achieved by the electronic device when executing the method of the first embodiment above can refer to the technical effects described in the first embodiment above, so they are not repeated here.

[0096] Third embodiment

[0097] This embodiment provides a computer-readable storage medium, which stores at least one instruction, and the instruction is loaded and executed by a processor to implement the method of the first embodiment. The computer-readable storage medium may be a ROM, a random access memory, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc. The instructions stored therein may be loaded by a processor in a terminal to execute the method.

[0098] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiment of the present invention can be in the form of a full or partial hardware embodiment, a full or partial software embodiment or an embodiment combining software and hardware. Moreover, when implemented using software, the embodiment of the present invention can be in the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center containing one or more available media sets. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium. The semiconductor medium may be a solid state hard disk.

[0099] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0101] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements. In addition, the term "and / or" is only an association relationship describing the associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone, wherein A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding. "At least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can be represented by: a, b, c, ab, ac, bc or abc, where a, b, c can be single or plural.

[0102] In addition, it can be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0103] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0104] In several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of functional modules / units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0105] If the method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0106] Finally, it should be noted that the above is only a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for ordinary technicians in this technical field, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A method for detecting defects of glass fiber pultruded plates with dual fields of view and with regional priority, characterized in that: The area-prioritized dual-field-of-view glass fiber pultruded plate defect detection method comprises: Preprocessing the image of the glass fiber pultruded plate to be detected to obtain a preprocessed image; wherein the image of the glass fiber pultruded plate to be detected is an image of the glass fiber pultruded plate acquired by a cross-stroboscopic image acquisition system formed by two lighting fields of reflected light and backlight; Based on the preprocessed image, determine whether the glass fiber pultruded plate image has oscillation. If the image does not have oscillation, perform edge finding and plate area cropping operations on the left and right boundaries of the plate based on the preprocessed image. Based on the cropped image, the defect locations and categories in the glass fiber pultruded sheet image are obtained.

2. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 1, characterized in that: Preprocess the image of the glass fiber pultruded sheet to be inspected, including: The image of the glass fiber pultruded plate to be tested is converted into a grayscale image, and then all the odd-numbered rows of data in the converted grayscale image are sequentially spliced ​​into a sub-image as the first lighting field image, and all the even-numbered rows of data in the converted grayscale image are sequentially spliced ​​into another sub-image as the second lighting field image.

3. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 2, characterized in that: The method of judging whether the glass fiber pultruded plate image has oscillation based on the preprocessed image, and if the image does not have oscillation, performing edge finding and plate area cropping operations on the left and right boundaries of the plate based on the preprocessed image, includes: Step 1, set the gradient threshold thre_G, the effective threshold of the lighting field thre_O, the backlight oscillation threshold thre_T, the super bright threshold thre_S, the oscillation number threshold thre_C, the binarization threshold thre_B and the brighter threshold thre_L; Step 2, performing gradient edge search on the first lighting field image and the second lighting field image respectively, and then determining the left boundary pixel position le_0 and the right boundary pixel position re_0 of the first lighting field image, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of the second lighting field image according to thre_G; Step 3, if |(re_0-le_0)-(re_1-le_1)|>thre_O, it is determined that the image does not oscillate, and the process goes to step 4; otherwise, it is determined that the image oscillates, and the process ends; Step 4: if (re_0-le_0)>(re_1-le_1), the first lighting field image is determined as a backlight image, and the second lighting field image is determined as a reflected light image; otherwise, the first lighting field image is determined as a reflected light image, and the second lighting field image is determined as a backlight image; Step 5, judging whether the backlight image is oscillating according to thre_T and thre_C, if the backlight image is not oscillating, go to step 6; otherwise, end the program; Step 6, setting the grayscale values ​​of pixels in the backlight image whose grayscale values ​​are greater than thre_S to 0, and then performing binarization and expansion operations on the first lighting field image and the second lighting field image according to thre_B to obtain the processed first lighting field image and the processed second lighting field image; Step 7, according to thre_L, find the area where the plate is located in the first and second light-emitting field images after processing, and perform a plate area cropping operation according to the determined area where the plate is located.

4. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 3, characterized in that: Performing gradient edge search on the first lighting field image and the second lighting field image respectively, and then determining the left boundary pixel position le_0 and the right boundary pixel position re_0 of the first lighting field image, and the left boundary pixel position le_1 and the right boundary pixel position re_1 of the second lighting field image according to thre_G, including: Performing lateral gradient calculation based on the Sobel operator on the first lighting field image and the second lighting field image respectively to obtain a gradient grayscale image corresponding to the first lighting field image and a gradient grayscale image corresponding to the second lighting field image; Find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from left to right in the gradient grayscale image corresponding to the first lighting field image, as the left boundary pixel position le_0 of the first lighting field image, and find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from right to left in the gradient grayscale image corresponding to the first lighting field image, as the right boundary pixel position re_0 of the first lighting field image; Find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from left to right in the gradient grayscale image corresponding to the second light field image, as the left boundary pixel position le_1 of the second light field image, and find the horizontal coordinate of the pixel with a gradient greater than thre_G that appears for the first time from right to left in the gradient grayscale image corresponding to the second light field image, as the right boundary pixel position re_1 of the second light field image; If there is no pixel gradient greater than thre_G in the image, it is determined that there is no obvious gradient change in the corresponding image, and the corresponding left boundary pixel position and right boundary pixel position are both set to 0.

5. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 3, characterized in that: Determine whether the backlight image is oscillating based on thre_T and thre_C, including: Extract two rows of data from the backlight image, and then search the two rows of data from left to right at the same time to find the first pixel with a grayscale value greater than thre_T in the two rows of data, and then count how many pixels in the column where this pixel is located have a grayscale value less than thre_T. If the statistical result is less than thre_C, it is determined that the backlight image does not oscillate; otherwise, it is determined that the backlight image oscillates.

6. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 3, characterized in that: The first lighting field image and the second lighting field image are binarized and expanded according to thre_B to obtain the processed first lighting field image and the processed second lighting field image, including: The first lighting field image is evenly divided into 5 segments horizontally, and then the maximum value of the grayscale mean values ​​of the respective regions of the 5 segments is taken, and it is multiplied by thre_B as the actual threshold value of the grayscale image binarization corresponding to the first lighting field image, and then the grayscale value of the pixel whose grayscale value in the first lighting field image is greater than or equal to the actual threshold value of the grayscale image binarization corresponding to the first lighting field image is set to 255, and the grayscale value of the pixel whose grayscale value is less than the actual threshold value of the grayscale image binarization corresponding to the first lighting field image is set to 0, and the binarization of the first lighting field image is completed, and then the binary first lighting field image is expanded once with a kernel size of 3×3 to obtain the processed first lighting field image; The second lighting field map is evenly divided into 5 segments horizontally, and then the maximum value of the grayscale means of each area of ​​the 5 segments is taken, and it is multiplied by thre_B as the actual threshold for binarization of the grayscale image corresponding to the second lighting field map, and then the grayscale value of the pixel in the second lighting field map whose grayscale value is greater than or equal to the actual threshold for binarization of the grayscale image corresponding to the second lighting field map is set to 255, and the grayscale value of the pixel less than the actual threshold for binarization of the grayscale image corresponding to the second lighting field map is set to 0, to complete the binarization of the second lighting field map, and then perform an expansion operation with a kernel size of 3×3 on the binarized second lighting field map to obtain the processed second lighting field map.

7. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 3, characterized in that: According to thre_L, find the area where the plate is located in the first and second light field images after processing, including: Take out one line of data from every four lines of the processed first light field image to form a new grayscale image, recorded as the first grayscale image, and then average the grayscale values ​​of the first grayscale image vertically to obtain a line of data representing the grayscale distribution characteristics of the first grayscale image, and filter out the horizontal position range of pixels corresponding to the area with a grayscale value greater than thre_L, and take the area corresponding to the maximum horizontal position range of pixels as the area where the plate is located; For the processed second lighting field image, one line of data is taken out every four lines to form a new grayscale image, which is recorded as the second grayscale image. The grayscale values ​​of the second grayscale image are then averaged vertically to obtain a line of data representing the grayscale distribution characteristics of the second grayscale image. The horizontal position range of pixels corresponding to the area with a grayscale value greater than thre_L is screened out, and the area corresponding to the maximum horizontal position range of pixels is taken as the area where the plate is located.

8. The method for detecting defects of glass fiber pultruded plates with dual fields of view and region priority as claimed in claim 3, characterized in that: Based on the cropped image, the defect locations and categories in the glass fiber pultruded sheet image are obtained, including: The cropped image of the reflected light image is scaled, and then uniformly cropped horizontally to obtain multiple sub-images of the same height as reflected light sub-images; The cropped backlight image is scaled, and then uniformly cropped horizontally to obtain multiple sub-images of the same height as backlight sub-images; Use the preset reflection field detection model to detect the defect position of the reflected photon image and obtain the detection result; Use the preset backlight field detection model to detect the defect position of the backlight sub-image and obtain the detection result; The preset classification model is used to classify the local color images corresponding to the detection results output by the reflection field detection model and the backlight field detection model to obtain the defect locations and categories in the glass fiber pultruded sheet images.

Images