Concrete crack identification method based on image processing

Abstract

The invention relates to the technical field of image processing, in particular to a concrete crack recognition method based on image processing, which comprises the following steps: acquiring an image to be analyzed, determining an image state of the image to be analyzed according to an estimated comparison deviation value and a noise pollution degree, and determining an image processing mode according to the image state; in hierarchical segmentation processing, the number of segmentation pixel values is determined according to a crack complex coefficient, a segmentation pixel value selection mode is determined according to a pixel value correlation coefficient, and a selection optimization mode is determined according to an identification difficulty coefficient and a segmentation interaction value under the condition of initial selection completion; in the apparent segmentation processing, a segmentation region is determined according to the association anomaly combination, and the area of the segmentation region is adjusted according to an edge influence coefficient; determining an image refinement mode according to the type of the segmented region, and determining a crack region according to an abnormal region in the refined image under the condition that image refinement is completed; the crack identification precision can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a concrete crack recognition method based on image processing. Background Art

[0002] The presence of cracks in concrete walls poses a huge safety hazard. These cracks will affect the strength and stiffness of concrete, thereby leading to a decrease in bearing capacity. Therefore, accurate detection of concrete cracks is of vital importance to ensuring the safety and stability of concrete structures. However, in the actual crack detection process, due to the large number of image texture features, some non-crack textures are often mistakenly identified as cracks, resulting in the accuracy of crack identification being difficult to meet actual needs. Therefore, how to process images to improve crack identification accuracy is a technical problem that needs to be urgently solved by technical personnel in this field.

[0003] Chinese patent publication number CN113744185A discloses a method for segmenting concrete surface cracks based on deep learning and image processing, including: obtaining a digital image of concrete with cracks on the surface, marking the cracks, and obtaining a crack mask image after processing the marking results; dividing the masked concrete surface digital image into a training set and a verification set, and then cutting it into sub-images; constructing an apparent crack coarse segmentation model, and training the model using the training set; cutting the concrete surface digital image to be detected into sub-images and inputting them into the coarse segmentation model; interpolating the original image to be detected and the crack coarse segmentation result image to obtain a refined segmentation result image; and obtaining the contour of the refined segmentation result image. It can be seen that the above technical solution has the following problems: the image processing method in the crack recognition process is single, and it is impossible to adaptively process the image according to the actual state of the image, resulting in poor crack recognition accuracy. Summary of the invention

[0004] To this end, the present invention provides a concrete crack recognition method based on image processing to overcome the problem that the image processing method in the crack recognition process in the prior art is single and cannot adaptively process the image according to the actual state of the image, resulting in poor crack recognition accuracy.

[0005] To achieve the above object, the present invention provides a concrete crack identification method based on image processing, comprising: Acquire an image to be analyzed, determine an image state of the image to be analyzed according to an estimated comparison deviation value and a noise pollution degree, and determine an image processing method according to the image state, wherein the image processing method includes a hierarchical segmentation process and an apparent segmentation process; In the hierarchical segmentation process, the number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient. Under the condition that the initial selection is completed, the selection optimization method is determined according to the recognition difficulty coefficient and the segmentation interaction value. The segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value or the associated threshold value, and the selection optimization method is to delete the segmentation pixel value or increase the segmentation pixel value; In the apparent segmentation process, the segmentation region is determined according to the associated anomaly combination, and the area of ​​the segmentation region is adjusted according to the edge influence coefficient; Determine an image refinement method according to the segmentation region type, the image refinement method is to determine whether to perform zero-value processing on the pixel point according to the neighborhood state of the pixel point, and determine a pixel optimization method according to the differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization or mean pixel optimization; When the image is refined, the crack area is determined based on the abnormal area in the refined image.

[0006] Further, determining the image processing method according to the image state includes: If the image status is that the estimated comparison deviation value is greater than or equal to the preset estimated comparison deviation value or the noise pollution degree is greater than or equal to the preset noise pollution degree, the image processing method is hierarchical segmentation processing; If the image status is that the estimated comparison deviation value is less than the preset estimated comparison deviation value and the noise pollution level is less than the preset noise pollution level, the image processing method is the appearance segmentation processing.

[0007] Furthermore, the method for confirming the estimated comparison deviation value includes: If the pixel fluctuation value is greater than or equal to the preset pixel fluctuation value, an estimated comparison deviation value is determined according to the position outlier degree and the pixel fluctuation value; If the pixel fluctuation value is less than the preset pixel fluctuation value, the estimated comparison deviation value is determined according to the recognition deviation coefficient.

[0008] Furthermore, the number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient; If the pixel value correlation coefficient is less than the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value; If the pixel value correlation coefficient is greater than or equal to the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the correlation threshold; The number of segmented pixel values ​​is positively correlated with the crack complexity coefficient.

[0009] Furthermore, the optimization method is selected according to the recognition difficulty coefficient and the segmentation interaction value, including: If the recognition difficulty coefficient is greater than or equal to the preset recognition difficulty coefficient or the segmentation interaction value is less than the preset segmentation interaction value, the optimization method is selected to increase the segmentation pixel value; If the recognition difficulty coefficient is less than the preset recognition difficulty coefficient and the segmentation interaction value is greater than or equal to the preset segmentation interaction value, then the selected optimization method is to delete the segmentation pixel value.

[0010] Furthermore, determine the segmentation region according to the associated abnormal combination, and perform an increase adjustment on the area of the segmentation region according to the edge influence coefficient; The increase value of the area of a single segmentation region is positively correlated with the edge influence coefficient corresponding to the segmentation region.

[0011] Furthermore, determine the image thinning method according to the segmentation region type, including: For a type of segmentation region, the image thinning method is to determine the pixel optimization method according to the differential correlation coefficient; For a second type of segmentation region, the image thinning method is to judge whether to process the pixel point as 0 according to the neighborhood state of the pixel point.

[0012] Furthermore, judge whether to process the pixel point as 0 according to the neighborhood state of the pixel point, including: For a single pixel point, If the neighborhood state is that the boundary correlation degree is greater than or equal to the preset boundary correlation degree or the feature coefficient is greater than or equal to the preset feature coefficient, then process the pixel point as 0; If the neighborhood state is that the boundary correlation degree is less than the preset boundary correlation degree and the feature coefficient is less than the preset feature coefficient, then do not process the pixel point as 0.

[0013] Furthermore, determine the pixel optimization method according to the differential correlation coefficient, including: For a single pixel point, If the differential correlation coefficient is less than the preset differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization; If the differential correlation coefficient is greater than or equal to the preset differential correlation coefficient, the pixel optimization method is mean pixel optimization.

[0014] Furthermore, the confirmation method of the differential correlation coefficient includes: If the mutation coefficient is greater than or equal to the preset mutation coefficient, then determine the differential correlation coefficient according to the angle disorder value and the mutation degree; If the mutation coefficient is less than the preset mutation coefficient, then determine the differential correlation coefficient according to the adjacent difference degree.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows. In the technical solution of the present invention, the image state of the image to be analyzed is determined according to the estimated comparison deviation value and the noise pollution degree. The estimated comparison deviation value and the noise pollution degree can effectively reflect the noise situation and the recognition difficulty of the image to be analyzed. Furthermore, different image processing methods are adaptively selected according to the image state, avoiding the problem of poor crack recognition accuracy caused by a single image processing method in the prior art. Through multi-level processing, the noise pollution can be reduced to improve the detail recognition ability of the image to be analyzed, and thus the recognition accuracy of cracks in the image to be analyzed can be improved.

[0016] Furthermore, in the present invention, the correlation degree of the pixel points corresponding to each pixel value in the image to be analyzed is effectively reflected by the pixel value correlation coefficient. Then, the selection method of the segmentation pixel value is determined according to the pixel value correlation coefficient, which can better retain the detail information of the cracks. And the complexity of the selected segmentation pixel value can be effectively reflected by the recognition difficulty coefficient and the segmentation interaction value, and the number of segmentation pixel values is dynamically adjusted to ensure that there is enough segmentation detail in the complex area and the simple area is not over-segmented, thereby improving the recognition accuracy of the cracks.

[0017] Furthermore, in the present invention, the segmentation area is determined according to the associated abnormal combination, which can ensure the integrity of the segmentation area and avoid the appearance of broken or discontinuous areas in the segmentation result. The edge characteristics of the segmentation area are effectively reflected by the edge influence coefficient, and then the segmentation area is dynamically adjusted, which can ensure the complete retention of the crack detail information, thereby improving the crack recognition accuracy.

[0018] Furthermore, in the present invention, the image thinning method is determined according to the segmentation area type, which can meet the thinning requirements of different types of segmentation areas and improve the stability of crack recognition. Through the neighborhood state of the pixel points, it can accurately judge whether to process the pixel points with a value of 0, avoiding over-thinning or under-thinning. According to the differential correlation coefficient, a suitable optimization method is selected, which can adapt to the thinning requirements of different areas, and thus can effectively handle noise interference and improve the accuracy of concrete crack recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic diagram of the concrete crack recognition method based on image processing of the present invention; Figure 2 is a flowchart of determining the image processing method according to the image state of the present invention; Figure 3 is a flowchart of determining the selection method of the segmentation pixel value according to the pixel value correlation coefficient of the present invention; Figure 4 is a flowchart of determining the image thinning method according to the segmentation area type of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0022] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0023] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] See also Figures 1 to 4 As shown, the present invention provides a concrete crack identification method based on image processing, comprising: Acquire an image to be analyzed, determine an image state of the image to be analyzed according to an estimated comparison deviation value and a noise pollution degree, and determine an image processing method according to the image state, wherein the image processing method includes a hierarchical segmentation process and an apparent segmentation process; In the hierarchical segmentation process, the number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient. Under the condition that the initial selection is completed, the selection optimization method is determined according to the recognition difficulty coefficient and the segmentation interaction value. The segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value or the associated threshold value, and the selection optimization method is to delete the segmentation pixel value or increase the segmentation pixel value; In the apparent segmentation process, the segmentation region is determined according to the associated anomaly combination, and the area of ​​the segmentation region is adjusted according to the edge influence coefficient; Determine an image refinement method according to the segmentation region type, the image refinement method is to determine whether to perform zero-value processing on the pixel point according to the neighborhood state of the pixel point, and determine a pixel optimization method according to the differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization or mean pixel optimization; When the image is refined, the crack area is determined based on the abnormal area in the refined image.

[0025] The application scenario of the present invention is the recognition of cracks in the image to be analyzed. In the present invention, the image to be analyzed is an image of a partial area of ​​the concrete wall surface photographed by a camera perpendicular to the concrete wall surface and then subjected to grayscale processing. The specific model of the camera is not limited, and the user can select it according to needs. The grayscale processing converts the color image of the partial area of ​​the concrete wall surface photographed by the camera perpendicular to the concrete wall surface into a grayscale image through OpenCV. This is content that is easy to understand for technicians in this field and will not be described in detail. The image to be analyzed contains a number of pixels, and the pixel value corresponding to a single pixel is greater than or equal to 0 and less than or equal to 255; In the present invention, several historical records are correspondingly provided, and any historical record records the estimated comparison deviation value, noise pollution degree, pixel fluctuation value, pixel value correlation coefficient and recognition difficulty coefficient in at least one historical process of crack recognition in the image to be analyzed, and each historical record corresponds to a qualified mark, which records whether the crack recognition process meets the user's requirements. The qualified mark can be recorded manually. It can be understood that the user can determine whether the crack recognition process meets the requirements according to the self-set indicators. The self-set indicators can be but not limited to the error recognition index, which will not be repeated here, wherein the error recognition index is the number of times the cracks in the image to be analyzed are incorrectly recognized; The image refinement completion condition is that the segmented regions are processed by the image refinement method and the reorganization of each refined segmented region is completed; the refined segmented region is the segmented region processed by the image refinement method; the refined image is the image generated after the refined segmented region is reorganized; The reorganization process for the first type of segmented area processed by the image thinning method and the second type of segmented area processed by the image thinning method is different; When the first-class segmented regions processed by the image thinning method are reorganized, each first-class segmented region is aligned to the same coordinate system by the image registration technology, and the pixel value of each pixel point in the merged image is in [0,250]. This is easy to understand for those skilled in the art and will not be described in detail. When reorganizing the two-category segmented regions processed by image thinning, the segmented regions are spliced ​​according to their positions in the image to be analyzed. If a single pixel in the image to be analyzed corresponds to multiple two-category segmented regions, the average value of the pixel values ​​corresponding to the pixel in each of the two-category segmented regions corresponding to the pixel is used as the pixel value of the pixel; if a single pixel in the image to be analyzed corresponds to a single two-category segmented region, the pixel value corresponding to the pixel in the two-category segmented region corresponding to the pixel is used as the pixel value of the pixel; When determining the crack region according to the abnormal region in the refined image, each abnormal region in the refined image is regarded as the crack region; The initial selection completion condition is that the segmentation pixel values ​​selected by the segmentation pixel value selection method reach the number of segmentation pixel values; The image to be analyzed and the refined image contain several abnormal areas. A single abnormal area is an area containing several pixels. The pixel values ​​of each pixel in the abnormal area are greater than 0 and the pixel values ​​of the adjacent pixels corresponding to the abnormal area are all 0. The adjacent pixels are confirmed in the following way: for a single abnormal area in a single image, each pixel in the abnormal area is recorded as a target pixel, other pixels except the target pixel are recorded as reference pixels, and each reference pixel adjacent to the target pixel is recorded as the adjacent pixel corresponding to the abnormal area.

[0026] Specifically, the image processing method is determined according to the image state, including: If the image status is that the estimated comparison deviation value is greater than or equal to the preset estimated comparison deviation value or the noise pollution degree is greater than or equal to the preset noise pollution degree, the image processing method is hierarchical segmentation processing; If the image status is that the estimated comparison deviation value is less than the preset estimated comparison deviation value and the noise pollution level is less than the preset noise pollution level, the image processing method is the appearance segmentation processing.

[0027] The image state includes a first image state and a second image state. The first image state is that the estimated comparison deviation value is greater than or equal to the preset estimated comparison deviation value or the noise pollution degree is greater than or equal to the preset noise pollution degree. The second image state is that the estimated comparison deviation value is less than the preset estimated comparison deviation value and the noise pollution degree is less than the preset noise pollution. Noise pollution degree = (the number of abnormal areas in the image to be analyzed whose number reference value is less than the preset number reference value) / (the total number of abnormal areas in the image to be analyzed), where the number reference value is the total number of pixels contained in a single abnormal area; The user can determine the values ​​of the preset estimated comparison deviation value, the preset noise pollution level and the preset number reference value according to the actual application scenario. The smaller the values ​​of the preset estimated comparison deviation value and the preset noise pollution level, the greater the user's need for hierarchical segmentation processing. A preset estimated comparison deviation value and a preset noise pollution level are provided, and the historical records of users using apparent segmentation processing are detected. The average value of the estimated comparison deviation values ​​corresponding to the historical records that can meet the user's needs is recorded as the preset estimated comparison deviation value, and the average value of the noise pollution levels corresponding to the historical records that can meet the user's needs is recorded as the preset noise pollution level. The larger the value of the preset number reference value, the greater the user's need to determine the degree of noise influence in the image to be analyzed. A preset number reference value is provided, and the preset number reference value is 15.

[0028] Specifically, the methods for confirming the estimated comparison deviation value include: If the pixel fluctuation value is greater than or equal to the preset pixel fluctuation value, an estimated comparison deviation value is determined according to the position outlier degree and the pixel fluctuation value; If the pixel fluctuation value is less than the preset pixel fluctuation value, the estimated comparison deviation value is determined according to the recognition deviation coefficient.

[0029] The pixel fluctuation value is the standard deviation of the pixel quantity value corresponding to each pixel value, and the pixel value is 0, 1, 2, 3, ..., 250. The pixel quantity value is confirmed by recording the pixel value as the target value for a single pixel value, and the pixel value corresponding to the target value is the total number of pixel points in the image to be analyzed whose pixel value is the target value; The value of the preset pixel fluctuation value can be determined by the user according to the actual application scenario. The larger the value of the preset pixel fluctuation value is, the greater the user's need to determine the estimated comparison deviation value according to the recognition deviation coefficient is. A value of the preset pixel fluctuation value is provided, and the historical records of the user determining the estimated comparison deviation value according to the recognition deviation coefficient are detected, and the average value of the pixel fluctuation values ​​corresponding to the historical records that can meet the user's needs is recorded as the preset pixel fluctuation value; If the pixel fluctuation value is greater than or equal to the preset pixel fluctuation value, the estimated comparison deviation value = position outlier + pixel fluctuation value; If the pixel fluctuation value is less than the preset pixel fluctuation value, the estimated comparison deviation value and the recognition deviation coefficient are positively correlated; The position outlier is the average value of the outlier coefficients corresponding to each pixel value. For a single pixel value, the pixel value is recorded as the first target value. The outlier coefficient corresponding to the first target value is the number of pixel values ​​corresponding to each pixel point in the reference rectangle corresponding to the first target value. The reference rectangle is the minimum rectangle in the image to be analyzed that can contain all the pixel points corresponding to the single pixel value. The method for confirming the identification deviation coefficient is to record the historical records in the historical records that have the same pixel fluctuation value as that corresponding to the image to be analyzed and are processed using apparent segmentation as the first record. The identification deviation coefficient = (the total number of first records - the number of first records that can meet user needs) / (the total number of first records).

[0030] Specifically, the number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient; If the pixel value correlation coefficient is less than the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value; If the pixel value correlation coefficient is greater than or equal to the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the correlation threshold; The number of segmented pixel values ​​is positively correlated with the crack complexity coefficient.

[0031] The pixel values ​​in the image to be analyzed are 0, 1, 2, ..., 250, and the other pixel values ​​except 0 and 250 are recorded as the pixel values ​​to be selected, and the number of segmented pixel values ​​is the number of the selected pixel values ​​to be selected; Crack complexity coefficient = 1 / (average value of distance coefficients corresponding to each abnormal area in the image to be analyzed). The distance coefficient corresponding to a single abnormal area is the average value of the shortest distance from the center point corresponding to the abnormal area to the center points corresponding to other abnormal areas except the abnormal area. The center point corresponding to a single abnormal area is the center of the circumscribed circle corresponding to the abnormal area. The pixel value correlation coefficient is the average value of the correlation thresholds corresponding to the pixel values ​​to be selected. The correlation threshold corresponding to a single pixel value to be selected = (the number of pixel values ​​to be selected corresponding to each pixel point in the reference rectangle corresponding to the pixel value to be selected) / (the total number of pixel values ​​to be selected appearing in the image to be analyzed); The value of the preset pixel value association coefficient can be determined by the user according to the actual application scenario. The larger the value of the preset pixel value association coefficient is, the greater the user's need to select the segmentation pixel value according to the feature reference value is. A value of the preset pixel value association coefficient is provided, and the historical records of selecting the segmentation pixel value according to the association threshold are detected, and the average value of the pixel value association coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset pixel value association coefficient; The characteristic reference value is confirmed in the following manner: for a single pixel value to be selected, the pixel value to be selected is recorded as the second target value, and the pixel value adjacent to the second target value is recorded as the adjacent pixel value, and the characteristic reference value corresponding to the second target value = | (the sum of the number of pixels corresponding to each adjacent pixel value / 2) - (the number of pixels corresponding to the second target value) |; The segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value, and select the pixel value to be selected as the segmentation pixel value in the order of the feature reference value from large to small until the number of segmentation pixel values ​​is reached; The segmentation pixel value selection method is to select the segmentation pixel value according to the associated threshold, and select the pixel value to be selected as the segmentation pixel value in descending order of the associated threshold until the number of segmentation pixel values ​​is reached.

[0032] Specifically, the optimization method is selected based on the recognition difficulty coefficient and the segmentation interaction value, including: If the recognition difficulty coefficient is greater than or equal to the preset recognition difficulty coefficient or the segmentation interaction value is less than the preset segmentation interaction value, the optimization method is selected to increase the segmentation pixel value; If the recognition difficulty coefficient is less than the preset recognition difficulty coefficient and the segmentation interaction value is greater than or equal to the preset segmentation interaction value, the optimization method selected is to delete the segmentation pixel value.

[0033] The selected segmentation pixel values ​​and 0 and 250 are recorded as segmentation values. For a single segmentation value, the segmentation value is recorded as the target segmentation value. The segmentation value adjacent to the target segmentation value and larger than the pixel value corresponding to the target segmentation value is recorded as the adjacent segmentation value. (target segmentation value, adjacent segmentation value) is recorded as an interval segment. The recognition difficulty coefficient is the maximum value of the sub-difficulty coefficients corresponding to each interval paragraph. The sub-difficulty coefficient confirmation method is that for a single interval paragraph, the interval paragraph is recorded as a target paragraph, and each pixel value contained in the target paragraph is recorded as a paragraph value. The sub-difficulty coefficient corresponding to the target paragraph is the absolute value of the difference between the maximum value and the minimum value of the uniformity coefficient corresponding to each paragraph value. The uniformity coefficient confirmation method is that for a single paragraph value, each pixel point corresponding to the paragraph value is detected and recorded as an analysis point. The uniformity coefficient corresponding to the paragraph value is the average value of the point reference values ​​corresponding to each analysis point. The point reference value corresponding to a single analysis point is the average value of the shortest distance from the analysis point to other analysis points except the analysis point. The segmentation interaction value is the maximum value of the sub-interaction coefficients corresponding to each interval paragraph. The method for confirming the sub-interaction coefficient is as follows: for a single interval paragraph, the interval paragraph is recorded as the first paragraph, and the interval paragraph adjacent to the first paragraph is recorded as the second paragraph. The sub-interaction coefficient corresponding to the first paragraph = similarity coefficient - interval reference value, the interval reference value is the number of pixel values ​​contained in the first paragraph, and the similarity coefficient = 1 / |the larger value of the sub-difficulty coefficients corresponding to each second paragraph - the sub-difficulty coefficient corresponding to the first paragraph|; The values of the preset recognition difficulty coefficient and the preset segmentation interaction value can be determined by the user according to the actual application scenario. The smaller the value of the preset recognition difficulty coefficient and the larger the value of the preset segmentation interaction value, the greater the user's need to increase the segmentation pixel value. Provide a method for determining the values of the preset recognition difficulty coefficient and the preset segmentation interaction value. Detect the historical records of deleting segmentation pixel values, and record the average value of the recognition difficulty coefficients corresponding to the historical records that can meet the user's needs as the preset recognition difficulty coefficient, and record the average value of the segmentation interaction values corresponding to the historical records that can meet the user's needs as the preset segmentation interaction value; When the selected optimization method is to increase the segmentation pixel value, use the feature value of the segmentation area whose comprehensive evaluation value is greater than the preset comprehensive evaluation value as the increased segmentation pixel value; the method for determining the feature value is as follows: for a single segmentation area, record this segmentation area as the target segmentation area, and the reference value corresponding to the target segmentation area = (the sum of the two segmentation pixel values corresponding to the target segmentation area) / 2, and the feature value corresponding to the target segmentation area is the smallest pixel value greater than or equal to the reference value corresponding to the target segmentation area; When the selected optimization method is to delete the segmentation pixel value, delete the segmentation pixel values whose radiation coefficient is greater than the preset radiation coefficient; The comprehensive evaluation value = recognition difficulty coefficient + segmentation interaction value. The method for determining the radiation coefficient is as follows: for a single segmentation pixel value, the radiation coefficient = 1 / (the absolute value of the difference between the sub-difficulty coefficients of the two adjacent paragraphs corresponding to this segmentation pixel value). The two adjacent paragraphs corresponding to a single segmentation pixel value are the interval paragraph where this segmentation pixel value is located and the interval paragraph adjacent to this segmentation pixel value; The values of the preset comprehensive evaluation value and the preset radiation coefficient can be determined by the user according to the actual application scenario. The smaller the value of the preset comprehensive evaluation value, the greater the user's need to increase the segmentation pixel value. Provide a method for determining the value of the preset comprehensive evaluation value. Detect the historical records of increasing the segmentation pixel value, and record the average value of the comprehensive evaluation values corresponding to the historical records that can meet the user's needs as the preset comprehensive evaluation value. The smaller the value of the preset radiation coefficient, the greater the user's need to delete the segmentation pixel value. Provide a method for determining the value of the preset radiation coefficient. Detect the historical records of deleting the segmentation pixel value, and record the average value of the radiation coefficients corresponding to the historical records that can meet the user's needs as the preset radiation coefficient; It can be understood that by selecting the optimization method, several divided pixel values can be obtained. For a single divided pixel value, record this divided pixel value as the target divided pixel value, record the divided pixel value that is adjacent to the target divided pixel value and is larger than the pixel value corresponding to the target divided pixel value as the adjacent divided pixel value, and record (the target divided pixel value, the adjacent divided pixel value] as a divided paragraph. Each divided paragraph corresponds to a segmentation area, and a single segmentation area only contains the pixel points corresponding to the pixel values in the interval (the target divided pixel value, the adjacent divided pixel value] in the image to be analyzed; The method for confirming the divided pixel value includes: When the optimization method is selected to increase the segmentation pixel value, the pixel value increased by increasing the segmentation pixel value and each segmentation value are recorded as the division pixel value; When the optimization method is selected to delete the segmentation pixel value, the segmentation value remaining after deleting the segmentation pixel value is recorded as the division pixel value.

[0034] Specifically, the segmentation region is determined according to the associated abnormal combination, and the area of ​​the segmentation region is increased and adjusted according to the edge influence coefficient; The increase in the area of ​​a single segmented region is positively correlated with the edge influence coefficient corresponding to the segmented region.

[0035] Among them, in determining the segmented area according to the associated abnormal combination, an associated analysis is performed on each abnormal area in the image to be analyzed. When an associated analysis is performed on a single abnormal area, the abnormal area is recorded as a target abnormal area, and other abnormal areas except the target abnormal area are recorded as reference abnormal areas. A reference abnormal area whose distance reference value from the target abnormal area is less than a preset distance reference value and a set of the target abnormal area are recorded as an associated abnormal combination, and the associated analysis is continued for each abnormal area that has not been associated until each abnormal area is recorded in the associated abnormal combination, and the minimum rectangle that can contain each abnormal area corresponding to the single associated abnormal combination is recorded as a segmented area; The distance reference value is the length of the line connecting the center points corresponding to the two abnormal areas. The preset distance reference value can be determined by the user according to the actual application scenario. The greater the user's demand for improving the crack identification accuracy, the smaller the preset distance reference value is. A preset distance reference value is provided, and the preset distance reference value is 10 cm. Edge influence coefficient = (number of pixels with pixel values ​​greater than 0 in the edge area) / (total number of pixels in the edge area); the edge area is the area between the inner rectangle boundary and the segmentation area boundary, the inner rectangle is located inside the segmentation area and is similar to the segmentation area, and the shortest distances from each side of the inner rectangle to the corresponding reference side are equal. The method for confirming the reference side is that for a single side of the inner rectangle, the side is recorded as the target side, and the reference side corresponding to the target side is the side of the four sides corresponding to the segmentation area that is parallel to the target side and closest to the target side; the shortest distance from each side of the inner rectangle to the corresponding reference side is the first reference distance, and the value of the first reference distance can be determined by the user according to actual needs. A value of the first reference distance is provided, and the value of the first reference distance is one-fifth of the length of the shorter side of the segmentation area; the segmentation area boundary is the four sides of a single segmentation area; the inner rectangle boundary and the outer rectangle boundary are the four sides of the inner rectangle and the outer rectangle respectively; When the area of ​​the segmented area is increased and adjusted according to the edge influence coefficient, the increased area of ​​the single segmented area is the area between the boundary of the outer rectangle and the boundary of the segmented area, the outer rectangle is located outside the segmented area and is similar to the segmented area, and the shortest distances from each side of the outer rectangle to the corresponding first reference side are all equal. The method for confirming the first reference side is that for a single side of the outer rectangle, the side is recorded as the first target side, and the first reference side corresponding to the first target side is an side of the four sides corresponding to the segmented area that is parallel to the first target side and the closest to the first target side; The area after the segmentation area is enlarged and adjusted according to the edge influence coefficient is used as the adjusted segmentation area; it should be noted that if the outer rectangle is inside the edge of the image to be analyzed, the outer rectangle is used as the segmentation area; if part of the outer rectangle is outside the edge of the image to be analyzed, the area of ​​the outer rectangle that exceeds the image to be analyzed is deleted, and the remaining area of ​​the outer rectangle is used as the segmentation area.

[0036] Specifically, the image refinement method is determined according to the segmentation region type, including: For a type of segmented area, the image refinement method is to determine the pixel optimization method based on the differential correlation coefficient; For the second type of segmented area, the image refinement method is to determine whether to perform 0 value processing on the pixel point according to the neighborhood status of the pixel point.

[0037] The first type of segmented regions are segmented regions obtained through hierarchical segmentation processing, and the second type of segmented regions are adjusted segmented regions obtained through apparent segmentation processing.

[0038] Specifically, judging whether to perform zero-value processing on a pixel point according to the neighborhood state of the pixel point includes: For a single pixel, If the neighborhood state is that the boundary correlation degree is greater than or equal to the preset boundary correlation degree or the characteristic coefficient is greater than or equal to the preset characteristic coefficient, the pixel point is processed as 0; If the neighborhood state is that the boundary correlation degree is less than the preset boundary correlation degree and the characteristic coefficient is less than the preset characteristic coefficient, no 0 value processing is performed on the pixel point.

[0039] The neighborhood state of the pixel point is determined according to the boundary correlation degree and the characteristic coefficient. The neighborhood state of the pixel point includes a first neighborhood state and a second neighborhood state. The first neighborhood state is that the boundary correlation degree is greater than or equal to the preset boundary correlation degree or the characteristic coefficient is greater than or equal to the preset characteristic coefficient. The second neighborhood state is that the boundary correlation degree is less than the preset boundary correlation degree and the characteristic coefficient is less than the preset characteristic coefficient. The method for confirming the boundary association degree is as follows: for a single pixel point, the pixel point is recorded as the target point, and the number of associated points with the same pixel value as the target point is recorded as the boundary association degree. The single associated point corresponding to the target point is the pixel point whose line connecting the target point does not pass through other pixel points except the associated point and the target point; the characteristic coefficient is the average value of the shortest distance from each associated point with the same pixel value as the target point to the target point; The values ​​of the preset boundary correlation degree and the preset characteristic coefficient can be determined by the user according to the actual application scenario. The larger the values ​​of the preset boundary correlation degree and the preset characteristic coefficient are, the greater the demand for 0-value processing for the pixel points is. A value of the preset boundary correlation degree and the preset characteristic coefficient is provided, and the historical records of 0-value processing for the pixel points are detected. The average value of the boundary correlation degree corresponding to the historical records that can meet the user's needs is recorded as the preset boundary correlation degree, and the average value of the characteristic coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset characteristic coefficient; When the pixel point is processed as 0 value, the pixel value of the pixel point is changed to 0, and when the pixel point is not processed as 0 value, the pixel value of the pixel point remains unchanged.

[0040] Specifically, the pixel optimization method is determined according to the differential correlation coefficient, including: For a single pixel, If the differential correlation coefficient is less than the preset differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization; If the differential correlation coefficient is greater than or equal to the preset differential correlation coefficient, the pixel optimization method is mean pixel optimization.

[0041] The value of the preset differential correlation coefficient can be determined by the user according to the actual application scenario. The larger the value of the preset differential correlation coefficient is, the greater the user's need for eigenvalue pixel optimization is. A value of the preset differential correlation coefficient is provided, and the historical records of eigenvalue pixel optimization are detected. The average value of the differential correlation coefficients corresponding to the historical records that can meet the user's needs is recorded as the preset differential correlation coefficient. When the pixel optimization method is eigenvalue pixel optimization, for a single pixel point, the pixel value of the pixel point is changed to a eigenvalue, where the eigenvalue is the average value of the pixel values ​​corresponding to the associated points contained in the reference lines corresponding to the pixel point whose gradient mean value is less than the preset gradient mean value; When the pixel optimization method is mean pixel optimization, for a single pixel point, the pixel value of the pixel point is changed to the average value of the pixel values ​​corresponding to the associated points corresponding to the pixel point; The gradient mean corresponding to a single reference line corresponding to a single pixel point is the standard deviation of the pixel values ​​corresponding to each pixel point contained in the reference line. The value of the preset gradient mean can be determined by the user according to the actual application scenario. The greater the user's demand for image processing accuracy, the smaller the value of the preset gradient mean. A value of the preset gradient mean is provided, and the historical records of eigenvalue pixel optimization are detected, and the average value of the gradient means corresponding to the historical records that can meet the user's needs is recorded as the preset gradient mean.

[0042] Specifically, the confirmation methods of the differential correlation coefficient include: If the mutation coefficient is greater than or equal to the preset mutation coefficient, the differential correlation coefficient is determined according to the angle disorder value and the mutation degree; If the mutation coefficient is less than the preset mutation coefficient, the differential association coefficient is determined according to the proximity difference.

[0043] Among them, if the mutation coefficient is greater than or equal to the preset mutation coefficient, the differential correlation coefficient = 1 / (angle disorder value + mutation degree); If the mutation coefficient is less than the preset mutation coefficient, the differential association coefficient is negatively correlated with the adjacent difference; The method for confirming the mutation coefficient is that, for a single pixel point, the pixel point and each associated point corresponding to the pixel point are recorded as analysis points, and the standard deviation of the pixel value corresponding to each analysis point is recorded as the mutation coefficient; The value of the preset mutation coefficient can be determined by the user according to the actual application scenario. The larger the value of the preset mutation coefficient is, the greater the user's need to determine the differential correlation coefficient according to the proximity difference degree is. A value of the preset mutation coefficient is provided, and the historical records of the user determining the differential correlation coefficient according to the angle disorder value and the mutation degree are detected. The minimum value of the mutation coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset mutation coefficient; The angle confusion value is confirmed by recording the associated point whose pixel difference with the analysis point is greater than the preset pixel difference as the target associated point, and the angle confusion value = (the angle corresponding to the minimum angle of the gradient line that takes the analysis point as the endpoint and can contain each target associated point) / (the number of target associated points); The mutation degree is the maximum value of the pixel difference between the analysis point and each associated point; the pixel difference between a single associated point and the analysis point is the absolute value of the difference between the pixel value corresponding to the associated point and the pixel value corresponding to the analysis point; the value of the preset pixel difference can be determined by the user according to the actual application scenario. The greater the user's demand for improving the image processing accuracy, the smaller the value of the preset pixel difference. A preset pixel difference value is provided, and the preset pixel difference value is 50; Neighborhood difference = (maximum value among the pixel means corresponding to each reference line corresponding to the analysis point) - (minimum value among the pixel means corresponding to each reference line corresponding to the analysis point); The reference lines are the gradient lines located in the analysis area corresponding to the analysis point. The analysis area is a circle with the analysis point as the center and the reference length value as the radius. The reference length value is positively correlated with the mutation coefficient corresponding to the analysis point. The pixel mean corresponding to a single reference line is the average value of the pixel values ​​corresponding to the pixels through which the reference line passes. The gradient line is confirmed in that, for a single pixel, the ray with the position of the pixel as the endpoint and passing through the position of the single associated point corresponding to the pixel is the gradient line. It can be understood that the number of gradient lines corresponding to a single pixel is the same as the number of associated points corresponding to the pixel.

[0044] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A concrete crack recognition method based on image processing, characterized in that: include: Acquire an image to be analyzed, determine an image state of the image to be analyzed according to an estimated comparison deviation value and a noise pollution degree, and determine an image processing method according to the image state, wherein the image processing method includes a hierarchical segmentation process and an apparent segmentation process; In the hierarchical segmentation process, the number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient. Under the condition that the initial selection is completed, the selection optimization method is determined according to the recognition difficulty coefficient and the segmentation interaction value. The segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value or the associated threshold value, and the selection optimization method is to delete the segmentation pixel value or increase the segmentation pixel value; In the apparent segmentation process, the segmentation region is determined according to the associated anomaly combination, and the area of ​​the segmentation region is adjusted according to the edge influence coefficient; Determine an image refinement method according to the segmentation region type, the image refinement method is to determine whether to perform zero-value processing on the pixel point according to the neighborhood state of the pixel point, and determine a pixel optimization method according to the differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization or mean pixel optimization; When the image is refined, the crack area is determined based on the abnormal area in the refined image.

2. The method for identifying concrete cracks based on image processing according to claim 1, characterized in that: Determine the image processing method based on the image status, including: If the image status is that the estimated comparison deviation value is greater than or equal to the preset estimated comparison deviation value or the noise pollution degree is greater than or equal to the preset noise pollution degree, the image processing method is hierarchical segmentation processing; If the image status is that the estimated comparison deviation value is less than the preset estimated comparison deviation value and the noise pollution level is less than the preset noise pollution level, the image processing method is the appearance segmentation processing.

3. The method for identifying concrete cracks based on image processing according to claim 2, characterized in that: The method for confirming the estimated comparison deviation value includes: If the pixel fluctuation value is greater than or equal to the preset pixel fluctuation value, an estimated comparison deviation value is determined according to the position outlier degree and the pixel fluctuation value; If the pixel fluctuation value is less than the preset pixel fluctuation value, the estimated comparison deviation value is determined according to the recognition deviation coefficient.

4. The method for identifying concrete cracks based on image processing according to claim 2, characterized in that: The number of segmentation pixel values ​​is determined according to the crack complexity coefficient, and the segmentation pixel value selection method is determined according to the pixel value correlation coefficient; If the pixel value correlation coefficient is less than the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the feature reference value; If the pixel value correlation coefficient is greater than or equal to the preset pixel value correlation coefficient, the segmentation pixel value selection method is to select the segmentation pixel value according to the correlation threshold; The number of segmented pixel values ​​is positively correlated with the crack complexity coefficient.

5. The method for identifying concrete cracks based on image processing according to claim 4, characterized in that: The optimization method is determined based on the recognition difficulty coefficient and the segmentation interaction value, including: If the recognition difficulty coefficient is greater than or equal to the preset recognition difficulty coefficient or the segmentation interaction value is less than the preset segmentation interaction value, the optimization method is selected to increase the segmentation pixel value; If the recognition difficulty coefficient is less than the preset recognition difficulty coefficient and the segmentation interaction value is greater than or equal to the preset segmentation interaction value, the optimization method selected is to delete the segmentation pixel value.

6. The method for identifying concrete cracks based on image processing according to claim 5, characterized in that: Determine the segmentation area according to the associated abnormal combination, and increase and adjust the area of ​​the segmentation area according to the edge influence coefficient; The increase in the area of ​​a single segmented region is positively correlated with the edge influence coefficient corresponding to the segmented region.

7. The method for identifying concrete cracks based on image processing according to claim 6, characterized in that: The image refinement method is determined according to the segmentation region type, including: For a type of segmented area, the image refinement method is to determine the pixel optimization method based on the differential correlation coefficient; For the second type of segmented area, the image refinement method is to determine whether to perform 0 value processing on the pixel point according to the neighborhood status of the pixel point.

8. The method for identifying concrete cracks based on image processing according to claim 7, characterized in that: Determine whether to perform zero-value processing on the pixel point according to the neighborhood state of the pixel point, including: For a single pixel, If the neighborhood state is that the boundary correlation degree is greater than or equal to the preset boundary correlation degree or the characteristic coefficient is greater than or equal to the preset characteristic coefficient, the pixel point is processed as 0; If the neighborhood state is that the boundary correlation degree is less than the preset boundary correlation degree and the characteristic coefficient is less than the preset characteristic coefficient, no 0 value processing is performed on the pixel point.

9. The method for identifying concrete cracks based on image processing according to claim 7, characterized in that: The pixel optimization method is determined according to the differential correlation coefficient, including: For a single pixel, If the differential correlation coefficient is less than the preset differential correlation coefficient, the pixel optimization method is eigenvalue pixel optimization; If the differential correlation coefficient is greater than or equal to the preset differential correlation coefficient, the pixel optimization method is mean pixel optimization.

10. The method for identifying concrete cracks based on image processing according to claim 9, characterized in that: The confirmation methods of differential correlation coefficient include: If the mutation coefficient is greater than or equal to the preset mutation coefficient, the differential correlation coefficient is determined according to the angle disorder value and the mutation degree; If the mutation coefficient is less than the preset mutation coefficient, the differential association coefficient is determined according to the proximity difference.

Images