CN116012316A - Crack identification method for drilling core roller scanning picture - Google Patents

Abstract

The invention provides a crack identification method for a drilling core roller scanning picture, which comprises the following steps: step 1, obtaining a picture through an industrial camera, performing format conversion on the imported original picture, and converting RGB into HSV color space domain graphics. And step 2, finding out and extracting the crack characteristics of white and bright colors in the image. And 3, removing the core label. And 4, converting the picture into an RGB format and graying the picture. And 5, performing local threshold binarization processing based on image blocking. And 6, performing morphological image processing on the binary image. Aiming at a drilling core roller scanning picture, the method solves the problems that the identification of cracks is seriously affected due to the existence of labels, false cracks, red arrows and soil pollution on the surface of a core in the picture, and solves the problems that the background of a core roller picture is complex and the target and the background are difficult to segment to a certain extent.

Description

Crack identification method for drilling core roller scanning picture

Technical Field

The invention relates to the technical field of petroleum exploration and development, in particular to a crack identification method for a drilling core roller scanning picture.

Background

In the field of oil and gas reservoir exploration and development, an oil and gas reservoir containing a fracture and affected by the fracture has an important proportion in the ascertained oil and gas reserves. The fracture of an underground medium without significant displacement under the action of a stress field is called a fracture. Cracks are an important channel for migration and aggregation of oil and gas, and widely exist in tight sandstones. The total amount of the domestic main basin compact oil geological resources is about 110-135 hundred million tons, and the high-yield energy area is greatly influenced by crack development. On the other hand, shale oil resources have huge potential, and an important factor influencing the yield of the shale oil resources is the development degree of cracks. It is important to study the development characteristics of the reservoir fracture, and the fracture identification can obtain the development characteristic information of the fracture, fault and the like by using the seismic wave propagation theory and the seismic exploration method principle and using the seismic data interpretation and the geological theory and law as guidance. Seismic attribute techniques are currently a common means of predicting crack development, which can extract attributes from a seismic data volume or other data volumes produced thereby in order to obtain more characteristic information about changes in cracks, faults, etc. On the other hand, the development characteristics of the cracks in the research area can be summarized through core observation, slice identification, conventional and imaging logging methods. Core observation often uses manual observation data to calculate core cracks, so that deviation is large. The method realizes accurate description and observation of the reservoir core cracks, and has important significance for improving the development efficiency of the oil and gas reservoirs.

Image crack identification is very important not only in the field of oil and gas exploration, but also to related knowledge in tunnel engineering, railway traffic, bridge construction, medical imaging and the like. In recent years, in the field of recognition of cracks based on computer vision, the direction of research is roughly divided into two parts: firstly, based on digital image processing, the crack characteristics are identified manually, and a plurality of crack characteristic rules such as frequency, edge, HOG, gray scale, texture, entropy and the like are utilized to design some characteristic identification conditions to identify the crack characteristics; the traditional image recognition technology mainly defines disease areas through characteristics such as image gray values, such as an OTSU (on-the-fly) method, an edge detection and area growth algorithm and the like, has a good recognition effect on disease images with simple backgrounds and large gray differences, and has a high recognition error rate under complex backgrounds. The conventional image recognition technology has the common problems of low accuracy and high false alarm rate, can not recognize diseases at pixel level, processes pictures under different complex backgrounds, is dead, can not be flexibly recognized, needs to manually extract features, and the preprocessing method directly influences the recognition effect. And secondly, based on deep learning, a convolutional network is established, and seam characteristics are automatically found by utilizing the network, so that a machine continuously adjusts itself according to a certain rule to realize the effect that input data and output data approximate to a label. The method based on deep learning is characterized in that a network model is built to train a large number of pictures so as to achieve the identification purpose, the method is high in identification rate, but needs to fully train a large amount of data, the built model is large, the operation speed is low, the time consumption is long, and quantitative treatment on characteristic parameters of cracks is not performed yet.

Disclosure of Invention

The invention provides a method for identifying cracks by scanning a core roller, which aims to overcome the defect of digital image processing and can improve the accuracy and efficiency of crack extraction under a complex background.

In order to achieve the purpose of accurately extracting crack characteristics under the condition of strong complex background interference, the specific steps of the crack identification method of the core roller scanning picture include:

and step 1, acquiring a picture through an industrial camera, and converting the imported original picture from an RGB format to an HSV space domain.

And 2, recognizing and extracting crack characteristics.

And step 3, converting the image into an RGB format and graying the image.

And step 4, removing the core label, and sticking an identification label on the drilling core to seriously influence the identification of the crack.

And 5, local binarization processing based on image blocking.

And 6, morphological image processing.

Drawings

FIG. 1 shows a crack identification method for a scanned picture of a drilling core barrel

FIG. 2 is a drawing of a raw image of a borehole core drum scan

FIG. 3 is a photograph of HSV color space processed

FIG. 4 is a HSV color space domain picture after label removal

FIG. 5 is a region binarization map of image tiles

FIG. 6 is a diagram of a generic global threshold binarization

FIG. 7 is a picture after processing by the method of the present invention

Detailed Description

Preferred embodiments of the present invention will be described in more detail below. While the preferred embodiments of the present invention are described below, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

According to the invention, a process (figure 1) of a method for identifying cracks of a scanned picture of a drilling core roller comprises the following detailed steps:

in step 1, the collected complete core is fixed, and the core is rotated by an industrial camera for one circle, so that a core roller scanning image is obtained. The obtained scan pictures are three different color receptors of retina, namely red (R), green (G) and blue (B), from the viewpoint of neurophysiology. In the RGB color space, there are R, G, B three channels. Each color channel ranges between 0,255, the RGB space can represent 256 x 256 colors, the most common color space in image processing is the RGB model, which is often used for color display and image processing. The central axis from the origin to the white vertex is the gray line. The conventional image processing is to perform gray processing based on RGB images, and features of colors such as hue, brightness, saturation and the like are put together to represent the colors. When extracting the crack features, the crack obviously has unique hue, brightness and saturation, if RGB image processing is used, which is equivalent to discarding the crack features in the color space. Therefore, to better extract the crack features, the RGB image is transferred to an HSV color space domain image. H (Hue), i.e., H channel, represents Hue, with a range of values from 0 to 360, and it is understood that angulation, i.e., a closed loop range of values, is contemplated. S (Saturation), namely an S channel, represents Saturation, the value range is [0,1], the Saturation refers to the vividness of the color, and the higher the Saturation is, the more vivid the color of the image is and the stronger the visual effect is; whereas the lower the saturation, the weaker the visual effect. V (Value), i.e., V channel, represents brightness, the range of values is 0,1, and represents brightness of color, brightness 0 is black, and brightness 1 is white. The invention converts RGB image into HSV image, and processes the HSV color space image. The principle of converting an image from an RGB color space to an HSV color space is as follows:

Dmax＝max(R′，G′，B′) (4)

Dmin＝min(R′，G′，B′) (5)

E＝Dmax-Dmin (6)

V＝Dmax (7)

where R ', G', B 'represent the corresponding ratio of converting each pixel value in the three channels to 0,1, dmax, dmin represent the maximum and minimum values in R', G ', B', E is the difference between the maximum and minimum values, and H, S, V represents the hue value, saturation value, and brightness value, respectively. The RGB image is converted into HSV image so as to extract the basic characteristics of the crack in the next step.

Further, in step 2, the cracks in the scanned pictures of the drilling core roller are mostly white and bright, and the color features are used as basic features of the cracks to perform feature extraction. The original picture (figure 2) obtained and imported by the industrial camera can show that the target crack presents white and bright color characteristics, but the background is complex, the black core surface, red arrows and labels are present, and artificial cracks formed during manual operation exist in the core obtaining process, and other substances such as soil are infected with objects. After the RGB image is converted into the HSV color space image by the method, corresponding H, S, V numerical parameters are set, and the characteristics that cracks in the original image are white and bright are locked and extracted. After HSV color extraction, the background that does not meet the crack features is filled in black (FIG. 3). It can be seen that the cracks can be extracted completely and the interference of artificial false cracks, red arrows, complex background colors is avoided.

Further, in step 3, we perform subsequent processing on the HSV rock-roll-through pan picture after the color extraction, and further need to convert the HSV image into an RGB image, and then convert the RGB image into a gray scale image. The gray image is represented by black with different saturation levels for each image point, and the "gray" level is represented by a [0,255] number. Three channels, i.e. "three dimensions", of the RGB image are converted into a "two-dimensional" image with only one gray scale channel. The conversion relation between the RGB values and the gradation values is as follows:

Grey＝0.299×R+0.587×G+0.114×B (10)

where Grey is a gray value, ranging from [0,255], R, G, B represents the values in the RGB three channels, respectively. According to the formula, R, G, B values of all pixel points are sequentially read, gray values are calculated, the gray values are assigned to corresponding positions of the image, and after all the pixel points are traversed once, the gray of the RGB image is completed.

Further, in step 4, the drilling core is labeled, and the scanned image still has a label, which seriously affects crack identification. The color features of the label are similar to the slits, so the label is not removed in the HSV color space image, but the label is a regular rectangular shape and is much larger than the width of the slits. By setting a suitable sliding window of omega x omega, if the sliding window slides to the label, the gray value in the window is 255, at this time, the identified label is extracted, and the gray map of the drilling core drum with the label removed is obtained (fig. 4).

Further, in step 5, a region binarization process based on image blocking is performed on the obtained gray-scale image. Image binarization is a very important basic method in image processing, and can be generally used as a preprocessing technology of a plurality of image processing methods, such as edge extraction, target recognition, shape processing, image segmentation, optical character recognition and the like, and the image can be binarized and then subjected to subsequent processing. The binarization is simply to find a threshold value in the gray values of all pixels of the gray image, if the threshold value is larger than the threshold value, the gray value is assigned to be 1, and if the threshold value is smaller than the threshold value, the gray value is assigned to be 0, and the pixels of the image are traversed to obtain the binary image. The current binarization method based on the threshold value can be divided into a global threshold value method and a local threshold value method. The global thresholding method is to use a fixed threshold for the whole image, compare the gray value of the pixels in the image with the threshold, and divide the image into a background or a target. The method has good image effect aiming at the obvious distinction between the target and the background pixel gray level, but can not achieve the purpose of distinguishing the target when the background is complex or other interference objects pollute the background. In order to overcome the defect of the global threshold, the former scholars propose a plurality of methods related to the local threshold, the methods do not select a fixed threshold any more, but compare the pixel value of a point with the pixel value of the surrounding local neighborhood, and perform the self-adaptive adjustment of the threshold according to the local pixel gray condition, and the pixel point is further binarized according to the self-adaptive threshold. Comparing different core barrel scanning pictures, it is known that in the HSV color space domain, the disease with the same characteristics as the crack in terms of color characteristics still exists, but the interference of a plurality of disease areas is less than that of RGB image processing. According to the definition of the binary image in (fig. 4), a black area with a gray level of 0 is used as a background, and a white area with a gray level of 1 is used as a target crack and disease area. And carrying out local threshold binarization processing on the image. The gray scale map is first divided into m×n block areas, and the values of m and n can be selected to be appropriate values according to the original image size. A parameter value mu is introduced representing the average absolute deviation of the gray values in a region to determine the gray level variation of the pixels in a region. The formula is as follows:

wherein S is the sum of gray values of pixels in a block region, N is the number of pixels in each block region, g _i The gray value of the i-th pixel of a block region, v represents the pixel average value of a block region. μ is the average absolute deviation of the pixel gray values of a block region.

Selecting a proper value

By->

And detecting whether each block has a disease area. Mu (m, n) represents the average absolute deviation of the regional block (m, n), and the average absolute deviation can represent whether the regional gray level change is severe or not, and when a crack exists, the gray level change is severe from the background, and the larger the value is. On the contrary, if there is a diseased region or a background region in the region block, the gradation changes slowly, and the smaller the value thereof. Thus, when->

When the method is used, the region is indicated to be a target background region, namely, a region with cracks, and the region with cracks is large in gap between the cracks and background gray level changes and easy to distinguish due to the fact that the color of the HSV space domain is extracted. Therefore, the Otsu binarization, also called the discipline method, is a representative of global thresholding. Is an adaptive threshold determination method proposed in 1979 by Japanese scholars. The algorithm assumes that the image pixels can be separated into background and object portions according to a threshold. The optimal threshold is calculated to distinguish between the two classes of pixels such that the two classes of pixels are maximally distinguished. The following is the Otsu algorithm principle formula:

τ＝ω0×τ0+ω1×τ1 (16)

g＝ω0×(τ-τ0) ² +ω1×(τ-τ1) ² (17)

g＝ω0×ω1×(τ1-τ2) ² (18)

it is assumed that the threshold for object and background segmentation can be denoted as T, and that pixel gray values greater than the threshold T are objects, and vice versa. Then N ₀ 、N ₁ N is the total number of pixels of the (m, N) th block area, which is the number of pixels belonging to the object and the background. ω0 and ω1 are the total number of pixels of the target and the total number of pixels of the background. τ0 and τ1 represent the target pixel gray average value and the background pixel gray average value. τ is the pixel gray average value of the whole area block. g is the inter-class variance. The equation (16) is brought into equation (17), resulting in a final reduced equation for equation (18).

When (when)

When the region is in the background region or the background disease region. There are many methods for selecting a local thresholding method to perform thresholding and local thresholding, in which a niback algorithm is used, which is a binary algorithm based on local features of a gray image, and the principle is as follows:

T(x，y)＝k(x，y)+β·l(x，y) (21)

wherein, (x, y) is the center pixel point of a w×w window, k (x, y) is the average value of samples in w×w neighbors of the point, l (x, y) is the standard deviation in w×w neighbors of the point, and β is the correction coefficient selection range of [0,1]. g (x, y) is the pixel gray value of the point, and T (x, y) represents the threshold value of the binarization of the pixel point. The Niblack algorithm may

To dynamically determine the threshold value for each region, it is advantageous to take advantage of the phenomenon of uneven image gradation, etc.

Therefore, the gray map is binarized for the core barrel scanning by a local binarization method based on the segmented image. After binarization (fig. 5), the diseased area is pressed to some extent. The binarization process was directly performed without HSV spatial color extraction (fig. 6), and it can be seen that although the crack characteristics were well represented, it was difficult to remove the background interference.

Further, in step 6

In morphological image processing. In the above step, a local thresholding map of the area blocks is obtained (fig. 5), and it can be seen that there are still many white noise points. Therefore, morphology processing is required for the core barrel binary image, wherein the morphology processing is to extract image components useful for expressing and describing the shape of the region from the image by using digital morphology as a tool, such as expansion, corrosion, open and close operation, connected region processing and the like. Corrosion may eliminate boundary points and eliminate boundary discrete points. The expansion is to expand the highlighted portion of the image, so that the adjacent but broken slits can be connected to form a connected region. Firstly, the binary image is processed by open operation, and the open operation is an operation process of firstly etching and then expanding. The method can eliminate small objects, separate the objects at the fineness, smooth large objects, and remove small particle noise without obviously changing the area of the objects. And then performing a closing operation, connecting the places with discontinuous crack characteristics, and filling the tiny holes. And finally, removing the area threshold of the connected region, setting an area threshold by manually observing the area of the crack region in the image, deleting the connected region smaller than the area threshold, and finishing morphological processing of the binary image.

The crack identification method for the scanned picture of the drilling core roller has a very good identification effect on cracks with obvious color characteristics. (fig. 7) is a final image after using the crack recognition method of the present invention, and the core of the present invention is to convert an RGB image into an HSV color space domain image before conventional digital image processing, and accurately recognize and extract cracks by extracting color features of a specific color, i.e., cracks. And then converting to RGB image to be image digital image. According to the method, a binarization method of a local threshold value under the image area segmentation is added into a conventional binarization method, so that the segmentation of the crack and the background is more obvious, a large-area interference area is eliminated, and the crack is accurately reserved. Finally, noise points in the binary image are pressed through morphological treatment, and finally, cracks are accurately identified.

Claims (7)

1. A crack identification method of a drilling core roller scanning picture is characterized in that the method comprises the following steps:

step 1, acquiring a picture through an industrial camera, and converting an imported original picture from an RGB format to an HSV space domain;

step 2, identifying crack characteristics and extracting;

step 3, converting the image into RGB format and graying the image;

step 4, removing the core labels, and sticking identification labels to the drilling cores to seriously influence the identification of cracks;

step 5, local binarization processing based on image blocking;

and 6, morphological image processing.

2. The process (fig. 1) for identifying cracks in a scanned picture of a drill core drum according to claim 1, wherein the complete core is collected and fixed in step 1, and the scanned picture of the core drum is obtained by rotating the core one revolution using an industrial camera. The obtained scanned pictures are three different color receptors of retina, namely red (R), green (G) and blue (B), from the neurophysiologic perspective, three primary colors are in an RGB color space, wherein R, G, B three channels exist, each color channel ranges between [0,255], the RGB space can represent 256 multiplied by 256 colors in total, the traditional image processing is based on RGB images for gray processing, and the characteristics of colors such as hue, brightness, saturation and the like are put together for representation. When the crack characteristics are extracted, the crack obviously has unique Hue, brightness and Saturation, if RGB image processing is used, the characteristic of the crack in a color space is abandoned, therefore, for better extraction of the crack characteristics, an RGB image is transferred to an HSV color space domain image, H (Hue) is an H channel which represents the Hue, the value range is 0-360 degrees, the angle can be understood, namely, the value range of a closed loop, S (Saturation) is an S channel which represents the Saturation, the value range is [0,1], the Saturation is the vividness of the color, and the higher the Saturation is, the more vivid the color of the image is, the stronger the visual effect is; conversely, the lower the saturation, the weaker the visual effect, the V (Value), i.e., the V channel, represents the brightness, the range of values is also [0,1], the brightness of the color is represented, the brightness is 0, the color is black, and the brightness is white when both the brightness are 1.

3. The method for identifying cracks of a scanned picture of a drilling core roller according to claim 2 is characterized in that in step 2, an original picture (fig. 2) obtained and imported by an industrial camera can be seen that target cracks are white and bright, but the background is complex, artificial cracks formed during manual operation exist on the surface of a black core, red arrows and labels in the process of obtaining the core, other substances such as soil are infected with objects, the conventional digital image processing technology is not specially used for processing the picture of the scanning surface of the drilling core roller, the common pretreatment is that the characteristics of the cracks are strengthened after the histogram equalization and Gaussian filtering method are compared and limited self-adaptive histogram equalization, but red arrows and artificial false cracks in the corresponding original picture are also strengthened, the background is complex because of uneven color of the surface of the core, by the method, after the RGB images are converted into color space images, the corresponding H, S, V numerical parameters are set, the characteristics of the cracks which are white and bright in the original image are locked, the characteristics of the cracks are extracted, and after the HSV color is extracted, the background cracks which are not in accordance with the characteristics of the cracks are filled into black cracks (fig. 3), and the background cracks can be completely extracted and the complicated background cracks can be seen through the HSV color.

4. The method for identifying cracks in a scanned picture of a drill core drum according to claim 3, wherein in step 3, we use the HSV rock-roll-through scanned picture after color extraction for subsequent processing, and further need to convert the HSV image into an RGB image, and in converting the RGB image into a gray image, the gray image represents each image point with black with different saturation, and the gray degree is represented with a number of [0,255 ]. Three channels of RGB images, namely 'three-dimensional' are converted into 'two-dimensional' images with only one gray channel, and the conversion relation between RGB values and gray values is shown in the following formula:

Grey＝0.299×R+0.587×G+0.114×B

where Grey is a gray scale value ranging from [0,255], R, G, B represents values in RGB three channels, respectively. According to the formula, R, G, B values of all pixel points are sequentially read, gray values are calculated, the gray values are assigned to corresponding positions of the image, and after all the pixel points are traversed once, the gray of the RGB image is completed.

5. The method for identifying cracks in a scanned picture of a drill core barrel according to claim 4, wherein in step 4, the drill core is labeled, the scanned image still has the label, the crack identification is seriously affected, the color characteristics of the label are similar to those of the crack, so the label is not removed from the HSV color space image, but the label is in a regular rectangular shape and is far greater than the width of the crack, and if the label is slid to the position, the window is all 255 gray values, and the identified label is extracted at the moment, so that the gray map of the drill core barrel with the label removed is obtained (fig. 4).

6. The method for identifying cracks in a scanned picture of a drill core drum according to claim 5, wherein the obtained gray scale image is subjected to image block-based region binarization processing in step 5. Selecting a proper value

By->

And detecting whether each block has a disease area. Mu (m, n) represents the average absolute deviation of the area block (m, n), which can represent whether the gray level change of the area is severe or not, and when a crack exists, the gray level change is severe with the background, the larger the value is, whereas if a disease area or a background area exists in the area block, the gray level change is slow, the smaller the value is, therefore, when the crack exists

The method is characterized in that the region is a target background region, namely a region with cracks, and the region with cracks is subjected to HSV space domain color extraction, so that the crack is large in difference between the change of the background gray level and easy to distinguish, and therefore, the method is binarized by adopting an Otsu method, also called a big law method, and is a representative of global threshold binarization, is an adaptive threshold determination method proposed by Japanese scholars in 1979, an algorithm assumes that image pixels can be divided into a background part and a target part according to a threshold, the optimal threshold is calculated to distinguish the two types of pixels, so that the distinction degree of the two types of pixels is maximum, and the following is an Otsu algorithm principle formula:

τ＝ω0×τ0+ω1×τ1

g＝ω0×(τ-τ0) ² +ω1×(τ-τ1) ²

g＝ω0×ω1×(τ1-τ2) ²

assuming that the threshold for object and background segmentation can be denoted as T, a pixel gray value greater than the threshold T is the object, and vice versa, N ₀ 、N ₁ N is the total number of pixels of the (m, N) th block area for the number of pixels belonging to the object and the background; ω0 and ω1 are respectively the total number of pixels of the target and the total number of pixels of the background; τ0 and τ1 represent the average value of the gray scale of the target pixel and the average value of the gray scale of the background pixel; τ is the pixel gray average value of the whole area block; g is the inter-class variance, when

When the region is in a background region or a background disease region, a local threshold method is selected for threshold segmentation, and a plurality of methods exist for local threshold binarization, wherein a Niblack algorithm is adopted, and the Niblack algorithm is a binarization algorithm based on local features of gray images, and the principle is as follows:

T(x，y)＝k(x，y)+β·l(x，y)

wherein, (x, y) is the central pixel point of a w×w window, k (x, y) is the average value of samples in w×w neighbors of the point, l (x, y) is the standard deviation in w×w neighbors of the point, and beta is the correction coefficient selection range of [0,1]; g (x, y) is the pixel gray value of the point, and T (x, y) represents the binarization threshold value of the pixel point; the Niblack algorithm can dynamically determine the threshold value of each region, is good for the phenomenon of uneven image gray level and the like, and can be used for pressing a disease region to a certain extent after the gray level map is binarized by scanning the core roller by a local binarization method based on a segmented image (figure 5), and the binarization processing is directly carried out without HSV space color extraction (figure 6), so that the background interference is difficult to remove although the crack characteristic is well represented.

7. The method for identifying the cracks of the scanned pictures of the drilling core roller according to claim 6 is characterized in that the crack identification method of the scanned pictures of the drilling core roller in step 6 has very good effect on identifying the cracks with obvious color characteristics, (fig. 7) is a final image after the crack identification method is used.

Images