Thermal insulation layer defect detection method based on machine vision

Abstract

The invention discloses a thermal insulation layer defect detection method based on machine vision, relates to the technical field of defect detection, and solves the technical problem that single data matching cannot comprehensively judge defect reasons and causes errors to defect judgment. The method can automatically adjust the threshold value according to the gray feature distribution of different segmented areas in the whole image, is more suitable for the local difference and overall change conditions of the thermal insulation layer image, carries out the detailed analysis of the edge features through building a rectangular coordinate system for a target area with no gray feature matching result, and improves the accuracy of gray feature matching. According to the method, more clues about the defects can be mined from the edge information, and the distribution and formation mechanism of the defects in the thermal insulation layer can be more comprehensively understood by judging the integrity of the abnormal shape and combining and analyzing the associated region of the incomplete shape, so that the defect reasons can be more accurately determined.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and in particular to a method for detecting defects in a thermal insulation layer based on machine vision. Background Art

[0002] In the production and application process of the insulation layer, it is crucial to ensure its quality and freedom from defects. In this context, insulation layer defect detection technology based on machine vision has come into being. It can use advanced image acquisition equipment and intelligent image processing algorithms to quickly, accurately and automatically detect and evaluate the quality status of the insulation layer.

[0003] According to publication number CN114113110B, a method for detecting defects in the insulation layer of a natural gas pipeline based on machine vision is disclosed, comprising: arranging a liquefied natural gas pipeline shooting environment, taking a surface image of the insulation layer of the liquefied natural gas pipeline to be inspected; uploading the obtained image to a computer and counting the grayscale value of the obtained image; performing noise reduction processing on the image; using a neural network to train a grayscale threshold detection model; identifying defects in the insulation layer of the pipeline to be inspected, repairing the defects and re-inspecting.

[0004] However, when some existing detection methods are used in actual detection scenarios, photographic environmental factors such as lighting conditions, shooting angles and distances will have a significant impact on the quality of the acquired insulation layer image, thereby interfering with the accuracy of subsequent defect detection. Secondly, simply relying on the simple matching of grayscale features with the defect grayscale value range may not be able to fully and accurately determine the cause of the defect. For some abnormal areas whose grayscale features are not within the known defect range, more in-depth analysis methods are needed to explore the root causes of the defects. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for detecting defects in an insulation layer based on machine vision, which solves the problem that a single data matching cannot comprehensively determine the cause of the defect, causing errors in the determination of the defect.

[0006] To achieve the above objectives, the present invention is implemented by the following technical solutions: a method for detecting defects in an insulation layer based on machine vision, the method specifically comprising the following steps:

[0007] Step S1, obtaining image information of the insulation layer under standard conditions through an industrial camera;

[0008] Step S2, performing a preprocessing operation on the obtained thermal insulation layer image to obtain a preprocessed thermal insulation layer image, and extracting grayscale features of the preprocessed thermal insulation layer image;

[0009] Step S3, classifying the segmented image according to the obtained grayscale features, obtaining the target area and the background area by comparing the grayscale features with the comparison threshold, and performing defect recognition analysis on the obtained target area to obtain the analysis result;

[0010] Step S4, analyzing the matching existence results in the analysis results, generating specific defect causes in combination with the corresponding defect gray value intervals, and displaying the specific defect causes to the corresponding inspection personnel;

[0011] Step S5, analyzing the non-existent matching results in the analysis results, analyzing the edge features of the corresponding target area, and analyzing the specific defect causes according to the coordinate positions of the edge features, and generating specific defect cause information.

[0012] As a further solution of the present invention, the specific method of extracting the grayscale features of the pre-processed insulation layer image in step S2 is:

[0013] Acquire the thermal insulation layer image, and grayscale the thermal insulation layer image to obtain a grayscale image. Then, filter and de-noise the obtained grayscale image and perform image enhancement processing to obtain a pre-processed thermal insulation layer image. The obtained pre-processed thermal insulation layer image is divided into nine equal parts to obtain segmented images and labeled as i, where i=1, 2, ..., 9, where j represents the number of segmented images i, and any group of segmented images i is obtained as the analysis target;

[0014] Then obtain the number of rows and columns in the analysis target, record them as M and N respectively, and obtain the gray value of the pixel (a, b) as i(a, b), and substitute the obtained parameters into the formula The grayscale mean Qi corresponding to the analysis target is calculated, and the calculated grayscale mean Qi corresponds to the grayscale features of the segmented image.

[0015] As a further solution of the present invention, the specific method of classifying the segmented image according to the grayscale feature in step S3 is:

[0016] Obtain the grayscale feature Qi corresponding to the segmented image i, determine the comparison threshold at the same time, obtain all the segmented image grayscale features Qi, calculate the grayscale feature mean Qp of all the grayscale features Qi, and use the grayscale feature mean Qp as the comparison threshold;

[0017] Then, the grayscale feature Qi is compared with the comparison threshold Qp, and the segmented image corresponding to the grayscale feature Qi being greater than the comparison threshold Qp is marked as the target area, and the segmented image corresponding to the grayscale feature Qi being less than the comparison threshold Qp is marked as the background area.

[0018] As a further solution of the present invention, the specific method of performing defect recognition analysis on the target area in step S3 to obtain the analysis result is:

[0019] All target areas are obtained and labeled as o, where o=1, 2, ..., g, where g represents the number of target areas. Then historical data is obtained, and the corresponding defect causes in the historical data are sorted out to obtain the defect grayscale value interval. At the same time, the grayscale features corresponding to the target area o are matched with the defect grayscale value interval;

[0020] If the grayscale feature exists in the defect grayscale value interval, a matching result is generated. Otherwise, if the grayscale feature does not exist in the defect grayscale value interval, a matching result is generated.

[0021] As a further solution of the present invention, the specific method of analyzing the matching results in step S4 to generate specific defect causes is:

[0022] The target area corresponding to the matching result is obtained, and the defect gray value interval corresponding to the target area is obtained, and the defect cause corresponding to the defect gray value interval is obtained, and then the specific defect cause is generated based on the defect cause.

[0023] As a further solution of the present invention, the specific method of analyzing the non-existence of matching results in step S5 is:

[0024] The target area corresponding to the matching result is obtained and recorded as the area to be analyzed, and the label is recorded as c, and c = 1, 2, ..., r, where r represents the number of the area to be analyzed. Then, any group of areas to be analyzed is obtained as the target object, and a corresponding rectangular coordinate system is established with the center point of the target area as the origin. At the same time, a selected area with a radius of D is obtained, and then the edge features of the selected area and the target object are matched. The maximum edge feature corresponding to the selected area cannot be selected as a feature point;

[0025] In this way, all feature points are acquired, and then the selected feature points are connected in a clockwise order to obtain a feature shape, and then a secondary analysis is performed on the obtained feature shape.

[0026] As a further solution of the present invention, the specific method of performing secondary analysis on the characteristic shape in step S5 is:

[0027] Obtain the characteristic shape, and at the same time obtain all the pixels in the characteristic shape, and classify the pixels into normal pixels and abnormal pixels, then obtain all the abnormal pixels, and at the same time connect all the abnormal pixels to obtain the abnormal shape;

[0028] Similarly to the above analysis, all the areas to be analyzed are processed to obtain abnormal shapes, and then the integrity of the abnormal shapes corresponding to the areas to be analyzed is identified and judged.

[0029] As a further solution of the present invention, the specific method of identifying and judging the integrity of the abnormal shape corresponding to the area to be analyzed in step S5 is:

[0030] Determine whether the abnormal shape intersects with the boundary of the area to be analyzed. If so, the corresponding abnormal shape is classified as an incomplete shape. Otherwise, if not, the corresponding abnormal shape is classified as a complete shape. At the same time, the specific defect cause is generated based on the abnormal shape as a standard.

[0031] Then, the areas to be analyzed corresponding to all incomplete shapes are obtained, and the corresponding labels of the areas to be analyzed are obtained, and the areas to be analyzed with adjacent labels are recorded as associated areas. Then, the incomplete shapes corresponding to the associated areas are combined to obtain the complete shapes, and the specific defect causes are generated based on the complete shapes.

[0032] The present invention provides a method for detecting insulation layer defects based on machine vision. Compared with the prior art, it has the following beneficial effects:

[0033] The present invention can effectively reduce the adverse effects of environmental factors on image quality and improve the stability and analyzability of images by precisely controlling photographic environmental factors such as lighting conditions, shooting angles and distances, and adopting appropriate grayscale processing, filtering denoising and image enhancement techniques in the image preprocessing stage.

[0034] The present invention adopts a method for determining the comparison threshold based on the mean value of grayscale features, which can automatically adjust the threshold according to the grayscale feature distribution of different segmented areas in the whole image, and is more adaptable to the local differences and overall changes of the thermal insulation layer image. For the target area where there is no result of grayscale feature matching, the edge features are analyzed in detail by establishing a rectangular coordinate system, including a series of operations such as matching the selected area with the edge features, extracting abnormal pixels and connecting them to form abnormal shapes, which can mine more clues about defects from the edge information. In addition, by judging the integrity of the abnormal shape and combining the analysis of the associated areas of the incomplete shape, the distribution and formation mechanism of defects in the thermal insulation layer can be more comprehensively understood, so as to more accurately determine the cause of the defect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a step method diagram of the present invention;

[0036] Figure 2 A schematic diagram of the region selected for the present invention;

[0037] Figure 3 A schematic diagram of the defect shape of the present invention. DETAILED DESCRIPTION

[0038] 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.

[0039] For example, see Figure 1 , Figure 2 and Figure 3 The present application provides a method for detecting thermal insulation layer defects based on machine vision, which specifically includes the following steps:

[0040] Step S1, obtaining image information of the insulation layer under standard conditions through an industrial camera, and the standard conditions here are represented by the corresponding photographic environment, such as controlling lighting conditions and shooting angles and distances, for example, using uniform front lighting, side lighting or back lighting.

[0041] Step S2, performing a preprocessing operation on the obtained thermal insulation layer image to obtain a preprocessed thermal insulation layer image, and extracting the grayscale features of the preprocessed thermal insulation layer image.

[0042] The thermal insulation layer image is obtained, and the thermal insulation layer image is grayed to obtain a gray image. Then, the obtained gray image is filtered and de-noised and image enhanced to obtain a pre-processed thermal insulation layer image. The filtering and de-noising median filtering method is used for processing. At the same time, the image enhancement is performed by processing the contrast and brightness of the gray image. The obtained pre-processed thermal insulation layer image is divided into nine equal parts to obtain segmented images and labeled as i, and i=1, 2, ..., 9, where j represents the number of segmented images i. At the same time, any group of segmented images i is obtained as the analysis target, and then the number of rows and columns in the analysis target is obtained, which are recorded as M and N respectively, and the gray value of the pixel point (a, b) is obtained as i(a, b), and the obtained parameters are substituted into the formula The grayscale mean Qi corresponding to the analysis target is calculated, and the calculated grayscale mean Qi corresponds to the grayscale features of the segmented image.

[0043] Suppose we select the segmented image i=3 as the analysis target. The size of this segmented image is M=200 rows (because the original image is 600 pixels high, and each of the nine equal parts is 600÷3=200 pixels high), N=800÷3≈267 columns (the original image is 800 pixels wide, rounded for convenience of subsequent illustration), for a certain pixel point in this segmented image (a=50, b=100) (the coordinate value here is just an example of relative coordinates in this segmented image), its grayscale value is recorded as i(50, 100) and its value is assumed to be 140 (actually according to the numerical value after graying in the previous step), suppose we simply calculate the sum of the grayscale values ​​of all pixels in this segmented image i=3 to be 70000 (here it is assumed to be a convenient sum of 70000), then (The calculation results here are just examples, the actual situation is calculated based on the true grayscale value of the image).

[0044] Step S3, classifying the segmented image according to the obtained grayscale features, obtaining the target area and the background area by comparing the grayscale features with the comparison threshold, and performing defect recognition analysis on the obtained target area to obtain the analysis result.

[0045] Obtain the grayscale feature Qi corresponding to the segmented image i, and determine the comparison threshold at the same time, and the value of the comparison threshold is determined by the grayscale feature of the segmented image. The comparison threshold is determined in the following manner: obtain all the grayscale features Qi of the segmented images, calculate the grayscale feature mean Qp of all the grayscale features Qi, and use the grayscale feature mean Qp as the comparison threshold, then compare the grayscale feature Qi with the comparison threshold Qp, mark the segmented image corresponding to the grayscale feature Qi greater than the comparison threshold Qp as the target area, and mark the segmented image corresponding to the grayscale feature Qi less than the comparison threshold Qp as the background area, and the background area here represents the normal area of ​​the segmented image, and the target area represents the abnormal area of ​​the segmented image;

[0046] For example, the grayscale feature Q1 of segmented image 1 is 100, the grayscale feature Q2 of segmented image 2 is 110, the grayscale feature Q3 of segmented image 3 is 90, the grayscale feature Q4 of segmented image 4 is 120, the grayscale feature Q5 of segmented image 5 is 105, the grayscale feature Q6 of segmented image 6 is 115, the grayscale feature Q7 of segmented image 7 is 95, the grayscale feature Q8 of segmented image 8 is 108, and the grayscale feature Q9 of segmented image 9 is 112. Then calculate the mean Qp of all grayscale features, and compare each Qi with Qp≈106.11 as follows:

[0047] Since Q1=100<106.11, the segmented image 1 is marked as the background area (normal area).

[0048] Since Q2=110>106.11, the segmented image 2 is marked as a target region (abnormal region).

[0049] All target areas are obtained and labeled as o, where o=1, 2, ..., g, where g represents the number of target areas. Then historical data is obtained, and the corresponding defect causes in the historical data are sorted out to obtain the defect grayscale value interval. At the same time, the grayscale features corresponding to the target area o are matched with the defect grayscale value interval. If the grayscale feature exists in the defect grayscale value interval, a matching result is generated. Otherwise, if the grayscale feature does not exist in the defect grayscale value interval, a matching result is generated.

[0050] For example, the defect grayscale value intervals obtained from the historical data are as follows: the defect grayscale value interval caused by material moisture is 120-130, the defect grayscale value interval due to cracks caused by external force impact is 135-145, and the defect grayscale value interval due to material aging and discoloration is 150-160. For the target area o=1, its grayscale feature Qo1=125. Since 120<125<130, this grayscale feature is within the defect grayscale value interval caused by material moisture. Therefore, a matching existence result is generated for the target area 1, indicating that the area may have defects caused by material moisture. For the target area 2, its grayscale feature Qo2=105 does not exist in the desired defect grayscale value interval, so a matching non-existence result is generated.

[0051] Step S4, analyzing the matching existence results in the analysis results, generating specific defect causes in combination with corresponding defect gray value intervals, and displaying the specific defect causes to corresponding inspection personnel.

[0052] The target area corresponding to the matching result is obtained, and the defect gray value interval corresponding to the target area is obtained, and the defect cause corresponding to the defect gray value interval is obtained, and then the specific defect cause is generated based on the defect cause.

[0053] Step S5, analyzing the non-existent matching results in the analysis results, analyzing the edge features of the corresponding target area, and analyzing the specific defect causes according to the coordinate positions of the edge features, and generating specific defect cause information.

[0054] The target area corresponding to the matching result is obtained and recorded as the area to be analyzed, and the label is recorded as c, and c = 1, 2, ..., r, where r represents the number of the area to be analyzed. Then, any group of areas to be analyzed is obtained as the target object, and the corresponding rectangular coordinate system is established with the center point of the target area as the origin. The rectangular coordinate system here is as shown in the attached figure. Figure 2As shown, a selected area with a radius of D is obtained, and then the selected area is matched with the edge feature of the target object, and the matching here means that the edge feature is included in the selected area, and the edge feature is represented by the corresponding pixel points and obtained by the edge detection algorithm, and the edge detection algorithm used in this application is the Sobel algorithm. For example, the selected area includes multiple pixel points corresponding to the edge feature, and the maximum edge feature corresponding to the selected area cannot be selected as a feature point;

[0055] All feature points are obtained in this way, and then the selected feature points are connected in a clockwise order to obtain a feature shape, and the connecting line here is a straight line. Then the obtained feature shape is analyzed again, and the specific analysis method is as follows:

[0056] Acquire the characteristic shape, and at the same time acquire all the pixels in the characteristic shape, and classify the pixels into normal pixels and abnormal pixels, and the classification standard here is to compare the gray value of the pixel with the normal gray value, and the normal gray value is obtained according to the normal image, and the pixels whose gray value is within the range of the normal gray value are classified as normal pixels, and the pixels that are not within the range of the normal gray value are classified as abnormal pixels, and then acquire all the abnormal pixels, and at the same time connect all the abnormal pixels to obtain the abnormal shape;

[0057] Similarly to the above analysis, all the areas to be analyzed are processed to obtain abnormal shapes, and then the integrity of the abnormal shapes corresponding to the areas to be analyzed is identified and judged, and the specific judgment method is: judge whether the abnormal shape intersects with the boundary of the area to be analyzed, if there is an intersection, the corresponding abnormal shape is classified as an incomplete shape, otherwise if there is no intersection, the corresponding abnormal shape is classified as a complete shape, and the specific defect cause is generated based on the abnormal shape as the standard; for example, one end of the curve figure extends to the left boundary of the rectangular area and is connected to it. According to the judgment method, as shown in the attached Figure 2 As shown in the figure, this abnormal shape is classified as an incomplete shape. On the contrary, if this irregular curve figure is completely inside the area to be analyzed, there is no contact point with the four boundaries and no intersection, as shown in the figure. Figure 3 As shown, the abnormal shape is classified as a complete shape.

[0058] Then, the areas to be analyzed corresponding to all incomplete shapes are further obtained, and the corresponding labels of the areas to be analyzed are obtained, and the areas to be analyzed with adjacent labels are obtained as associated areas. Then, the incomplete shapes corresponding to the associated areas are combined to obtain the complete shapes, and the specific defect causes are generated based on the complete shapes.

[0059] In a specific example, it is found that there is a suspected defect of an incomplete shape in the area to be analyzed 3, which is similar to a semicircle, and the semicircle intersects with the right boundary of the area 3, so it is determined to be an incomplete shape. At the same time, it is found that there is also an incomplete shape in the area to be analyzed 6 with an adjacent number, which is a triangle-like shape, and the triangle intersects with the left boundary of the area 6.

[0060] Area 3 and area 6 are used as associated areas, and the incomplete shapes in these two areas are combined. The semicircle and the triangle can be combined to form a complete fan-like shape.

[0061] Embodiment 2, as Embodiment 2 of the present invention, is implemented on the basis of Embodiment 1, and differs from Embodiment 1 in that:

[0062] The comparison threshold in step S3 is determined in different ways. In this embodiment, the comparison threshold is a fixed threshold, and the specific value is set by the operator. For example, assuming that the gray value range of the normal insulation layer is between 100-150, when the gray value of a pixel is less than 100 or greater than 150, it is determined as a defective area.

[0063] Embodiment 3, as the embodiment 3 of the present invention, focuses on combining the implementation processes of embodiment 1 and embodiment 2 for implementation.

[0064] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0065] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for detecting defects in thermal insulation layer based on machine vision, characterized in that: The method specifically comprises the following steps: Step S1, obtaining image information of the insulation layer under standard conditions through an industrial camera; Step S2, performing a preprocessing operation on the obtained thermal insulation layer image to obtain a preprocessed thermal insulation layer image, and extracting grayscale features of the preprocessed thermal insulation layer image; Step S3, classifying the segmented image according to the obtained grayscale features, obtaining the target area and the background area by comparing the grayscale features with the comparison threshold, and performing defect recognition analysis on the obtained target area to obtain the analysis result; Step S4, analyzing the matching existence results in the analysis results, generating specific defect causes in combination with the corresponding defect gray value intervals, and displaying the specific defect causes to the corresponding inspection personnel; Step S5, analyzing the non-existent matching results in the analysis results, analyzing the edge features of the corresponding target area, and analyzing the specific defect causes according to the coordinate positions of the edge features, and generating specific defect cause information.

2. The method for detecting thermal insulation layer defects based on machine vision according to claim 1, characterized in that: The specific method of extracting the grayscale features of the pre-processed insulation layer image in step S2 is: Acquire the thermal insulation layer image, and grayscale the thermal insulation layer image to obtain a grayscale image. Then, filter and de-noise the obtained grayscale image and perform image enhancement processing to obtain a pre-processed thermal insulation layer image. The obtained pre-processed thermal insulation layer image is divided into nine equal parts to obtain segmented images and labeled as i, where i=1, 2, ..., 9, where j represents the number of segmented images i, and any group of segmented images i is obtained as the analysis target; Then obtain the number of rows and columns in the analysis target, record them as M and N respectively, and obtain the gray value of the pixel (a, b) as i(a, b), and substitute the obtained parameters into the formula The grayscale mean Qi corresponding to the analysis target is calculated, and the calculated grayscale mean Qi corresponds to the grayscale features of the segmented image.

3. The method for detecting thermal insulation layer defects based on machine vision according to claim 1, characterized in that: The specific method of classifying the segmented image according to the grayscale feature in step S3 is: Obtain the grayscale feature Qi corresponding to the segmented image i, determine the comparison threshold at the same time, obtain all the segmented image grayscale features Qi, calculate the grayscale feature mean Qp of all the grayscale features Qi, and use the grayscale feature mean Qp as the comparison threshold; Then, the grayscale feature Qi is compared with the comparison threshold Qp, and the segmented image corresponding to the grayscale feature Qi being greater than the comparison threshold Qp is marked as the target area, and the segmented image corresponding to the grayscale feature Qi being less than the comparison threshold Qp is marked as the background area.

4. The method for detecting thermal insulation layer defects based on machine vision according to claim 1, characterized in that: The specific method of performing defect recognition analysis on the target area in step S3 to obtain the analysis result is: All target areas are obtained and labeled as o, where o=1, 2, ..., g, where g represents the number of target areas. Then historical data is obtained, and the corresponding defect causes in the historical data are sorted out to obtain the defect grayscale value interval. At the same time, the grayscale features corresponding to the target area o are matched with the defect grayscale value interval; If the grayscale feature exists in the defect grayscale value interval, a matching result is generated. Otherwise, if the grayscale feature does not exist in the defect grayscale value interval, a matching result is generated.

5. The method for detecting thermal insulation layer defects based on machine vision according to claim 1, characterized in that: The specific method of analyzing the matching results in step S4 to generate specific defect causes is: The target area corresponding to the matching result is obtained, and the defect gray value interval corresponding to the target area is obtained, and the defect cause corresponding to the defect gray value interval is obtained, and then the specific defect cause is generated based on the defect cause.

6. The method for detecting thermal insulation layer defects based on machine vision according to claim 1, characterized in that: The specific method of analyzing the match non-existence result in step S5 is: The target area corresponding to the matching result is obtained and recorded as the area to be analyzed, and the label is recorded as c, and c = 1, 2, ..., r, where r represents the number of the area to be analyzed. Then, any group of areas to be analyzed is obtained as the target object, and a corresponding rectangular coordinate system is established with the center point of the target area as the origin. At the same time, a selected area with a radius of D is obtained, and then the edge features of the selected area and the target object are matched. The maximum edge feature corresponding to the selected area cannot be selected as a feature point; In this way, all feature points are acquired, and then the selected feature points are connected in a clockwise order to obtain a feature shape, and then a secondary analysis is performed on the obtained feature shape.

7. The method for detecting thermal insulation layer defects based on machine vision according to claim 6, characterized in that: The specific method of performing secondary analysis on the characteristic shape in step S5 is: Obtain the characteristic shape, and at the same time obtain all the pixels in the characteristic shape, and classify the pixels into normal pixels and abnormal pixels, then obtain all the abnormal pixels, and at the same time connect all the abnormal pixels to obtain the abnormal shape; Similarly to the above analysis, all the areas to be analyzed are processed to obtain abnormal shapes, and then the integrity of the abnormal shapes corresponding to the areas to be analyzed is identified and judged.

8. The method for detecting thermal insulation layer defects based on machine vision according to claim 7, characterized in that: The specific method of identifying and judging the integrity of the abnormal shape corresponding to the area to be analyzed in step S5 is: Determine whether the abnormal shape intersects with the boundary of the area to be analyzed. If so, the corresponding abnormal shape is classified as an incomplete shape. Otherwise, if not, the corresponding abnormal shape is classified as a complete shape. At the same time, the specific defect cause is generated based on the abnormal shape as a standard. Then, the areas to be analyzed corresponding to all incomplete shapes are obtained, and the corresponding labels of the areas to be analyzed are obtained, and the areas to be analyzed with adjacent labels are recorded as associated areas. Then, the incomplete shapes corresponding to the associated areas are combined to obtain the complete shapes, and the specific defect causes are generated based on the complete shapes.

Images