Field geological outcrop crack intelligent identification method based on fusion algorithm

Abstract

The invention relates to the technical field of oil exploration and development, in particular to an intelligent field geological outcrop crack identification method based on a fusion algorithm, namely, a Canny edge detection algorithm is adopted to identify a crack distribution result in a field geological outcrop image; obtaining the size of a crack bounding box in the field geological outcrop image by adopting a YOLOv5 algorithm; and obtaining the crack length and the crack width according to the crack bounding box size. According to the method, the Canny edge detection algorithm and the YOLOv5 algorithm are fused to carry out field geological outcrop crack distribution identification and crack size calculation, so that the labor cost is greatly saved, the operation is simple, the identification is accurate, and the efficiency is improved for accurately describing the field geological outcrop crack.

Description

Technical Field

[0001] The invention relates to the technical field of petroleum exploration and development, and in particular to a method for intelligently identifying cracks in field geological outcrops based on a fusion algorithm. Background Art

[0002] Fractures are natural discontinuities formed in rocks by tectonic deformation or physical diagenesis. Fractures can store oil and gas and are also the main seepage channels for oil and gas underground. Accurately identifying fractures can help engineers better understand the fluid flow characteristics of oil reservoirs. During oil and gas production, oil flows in the reservoir through fractures. If the distribution of fractures cannot be clearly identified, it is difficult to accurately simulate and predict the migration path of oil, resulting in problems such as low production efficiency.

[0003] The identification of fractures is of vital importance in the field of petroleum engineering. The identification of reservoir fractures can be roughly divided into geological identification methods and geophysical identification methods. Among them, geological outcrop observation is one of the main technical means of fracture identification. Geological outcrop observation is the most intuitive method to study the morphological characteristics of fracture development. Through the observation of field geological outcrops, the actual morphology, occurrence and distribution characteristics of fractures on rock outcrops can be directly observed, including the length, width, strike, dip and their mutual relationship, etc., which can obtain the most real and accurate first-hand information and avoid the errors that may be caused by indirect measurement or speculation. The formation and development of fractures are often closely related to regional geological tectonic movements. Through the observation and analysis of field outcrop fractures, the stress state, structural deformation process and mechanical property changes of rocks in the geological history period can be inferred, which provides an important basis for studying regional geological evolution. In addition to obtaining the geometric characteristics of the fractures themselves through field geological outcrop observations, the filling type, filling degree, rock lithology changes, bedding structure and relationship with other geological structures of the fractures can also be observed. This information is of great value for a deeper understanding of the formation mechanism and development history of fractures and their relationship with oil and gas migration and storage.

[0004] With the continuous development of science and technology, people tend to rely more on computer technology to deal with problems in a faster and more effective way. In traditional field geological outcrop observation and research, human eye recognition is the main method, and the crack characteristics and distribution status of the geological outcrop area are described through observation and measurement by personnel. However, in the actual work process, the large workload will bring certain crack description errors through human observation. Therefore, how to use computer means to solve the field geological outcrop identification is a research with practical significance.

[0005] The Chinese patent "Digital geological outcrop crack extraction method and device" with patent number CN112307803B discloses a digital geological outcrop crack extraction method and device. The method uses the Beamlet transformation method to extract linear features of geological outcrop images after image processing, and connects the crack line segments according to the distance and angle difference between the crack line segments to form geological outcrop cracks.

[0006] The Chinese patent number CN112287718B, "Method, device, electronic equipment and storage medium for identifying fractures and microfractures", discloses a method, device, electronic equipment and storage medium for identifying fractures and microfractures, which realizes the identification of fractures by conventional logging under oil-based mud conditions. By comparing with the degree of fracture development, the fracture identification threshold is comprehensively determined, and fracture identification is carried out accordingly.

[0007] The Chinese patent number CN109872318A, "A method for producing a geological outcrop fracture dataset for deep learning", provides a method for producing a geological outcrop fracture dataset for deep learning. The outcrop fracture samples and non-outcrop fracture samples are marked; the quality of the marked sample dataset is evaluated, and finally a high-quality geological outcrop fracture dataset can be obtained.

[0008] The above-mentioned patents are not comprehensive enough in identifying crack information, and none of them have conducted research on methods for calculating crack size parameters of field geological outcrops. Summary of the invention

[0009] The present invention aims to provide an intelligent method for identifying cracks in field geological outcrops based on a fusion algorithm. It innovatively proposes a method for identifying crack distribution and a method for calculating crack size, aiming to solve the problems of low efficiency and poor reliability in identifying cracks in field geological outcrops.

[0010] The present invention uses the Canny edge detection algorithm to identify the crack distribution results in the field geological outcrop image; uses the YOLOv5 algorithm to obtain the crack boundary box size in the field geological outcrop image; and obtains the crack length and crack width according to the crack boundary box size. The specific steps are: Grayscale the field geological outcrop color pictures to obtain grayscale images; Among them, the conversion formula used for grayscale processing is: ; In the formula, RGB is the industrial color standard, referring to red (R), green (G), and blue (B). A variety of colors can be obtained by mixing these three colors.

[0011] (2) Use a Gaussian filter to smooth the grayscale image, remove the noise in the grayscale image, and remove other factors that interfere with identification other than cracks, so as to obtain an image that only contains the cracks.

[0012] (3) Based on the image containing only the cracks, the gradient and direction are calculated using the Sobel filter to obtain the crack boundary in the image; Among them, the calculation method used by the Sobel filter to calculate the gradient and direction is: ; ; In the formula, G x It is the Sobel operator in the horizontal direction, used to detect y The edge of the direction; G y It is the Sobel operator in the vertical direction, used to detect x The edge of the direction; θ is the direction of gray value change.

[0013] (4) After obtaining the crack boundary in the image, the non-maximum suppression technique NMS is used to sharpen the image to obtain a clearer crack boundary recognition image; the pixels whose gradient intensity is greater than the maximum value are retained, and the pixels with other values ​​are replaced, so that the recognized boundary in the image is clearer; Among them, the calculation method used by the non-maximum suppression technology NMS is: ; In the formula, M(x,y) Represents the gradient of pixels at different locations; T Represents the maximum value of the set gradient strength.

[0014] Based on a clearer fracture boundary recognition image, a double threshold method is used to detect and connect fracture boundaries, and a fracture distribution result image in a field geological outcrop image is obtained; Specifically, a high threshold and a low threshold are set, and a crack boundary contour map is obtained according to the high threshold; if the generated crack boundary image is not closed, new crack boundaries are collected through the recognition of the low threshold, so as to close the crack boundaries that are actually connected in the image.

[0015] (6) Select the locations of cracks and holes in the crack distribution result image, and label them with categories and numbers, and further divide the identified boundaries in the image into holes and cracks.

[0016] (7) A convolutional neural network is used to extract features from the image selected and annotated in step (6), and a convolutional layer and a fully connected layer are used to identify the bounding boxes of cracks and holes in the image to obtain the size of the crack bounding box.

[0017] (8) Based on the image after the cracks and holes are selected in step (7), the length and width of the selected cracks are calculated based on the size of the crack boundary box, given the image size and scale.

[0018] Compared with the prior art, the present invention has the following advantages: 1. The Canny edge detection algorithm in digital image processing is used to identify the distribution of cracks in field geological outcrop image information. The crack size information (crack length and width) is obtained based on the YOLOv5 algorithm in target detection. The two algorithms are combined to identify field geological outcrop cracks more finely and accurately. It has the advantages of simple operation, accurate identification and high efficiency, and is very suitable for accurately identifying field geological outcrop cracks.

[0019] 2. The present invention proposes an intelligent identification method for crack size, and further divides the information contained in the field geological outcrop pictures into cracks and holes. It can not only obtain the distribution status of cracks in the field geological outcrop, but also further obtain the distribution of holes, providing a basis for subsequent research.

[0020] 3. The present invention proposes an integrated identification method for crack distribution and crack size, which can use computer technology to obtain more information from field geological outcrop pictures, greatly improving the efficiency of crack identification in field geological outcrops. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The geological outcrop pictures are obtained during the field geological survey in the embodiments.

[0022] Figure 2 This is a crack distribution result diagram obtained by using the Canny algorithm in the embodiment.

[0023] Figure 3 It is a crack boundary box identified by using the YOLOv5 algorithm in the embodiment. DETAILED DESCRIPTION

[0024] Example 1 Taking the H area as the survey object, a geological survey was conducted in the area, and color pictures of the field geological outcrops in the area were taken. The picture size was set to 2cm×3cm, and the image scale was 1cm=100cm, which was used for the crack feature analysis of the block. The intelligent recognition of field geological outcrop cracks based on the fusion algorithm was carried out through the obtained field geological outcrop color photos. The specific crack recognition steps are as follows: Step 1: Based on the color pictures of field geological outcrops, grayscale the color pictures while ensuring the picture size to obtain grayscale pictures for subsequent work.

[0025] Among them, the conversion formula used for grayscale processing is: ; In the formula, RGB is the industrial color standard, referring to red (R), green (G), and blue (B). A variety of colors can be obtained by mixing these three colors.

[0026] Step 2: Based on the grayscale image obtained in step 1, a Gaussian filter is applied to smooth the image, remove the noise in the image, and remove other factors that interfere with identification except cracks, so as to obtain an image containing only cracks.

[0027] Step 3: Based on the image containing only cracks, use the Sobel filter to calculate the gradient and direction. The gradient represents the degree and direction of change in the grayscale value, and the location with strong grayscale change is the boundary, so as to find the crack boundary in the image.

[0028] Among them, the calculation method used by the Sobel filter to calculate the gradient and direction is as follows: ; ; In the formula, G x is the Sobel operator in the horizontal direction; G y is the Sobel operator in the vertical direction; θ is the direction of gray value change.

[0029] Step 4: After obtaining the crack boundary in the image, apply the non-maximum suppression technique NMS to clarify the image to obtain a clearer crack boundary identification image. T Set to 110, the grayscale value of non-boundary points is 0, so that the pixels with gradient intensity greater than the maximum value are retained, making the boundaries recognized in the image clearer.

[0030] Among them, the calculation method used by the non-maximum suppression technology NMS is as follows: ; In the formula, M(x,y) Represents the gradient of pixels at different locations; T Represents the maximum value of the set gradient strength.

[0031] Step 5: Based on a clearer fracture boundary recognition image, a double threshold method is used to connect the fracture boundaries; a high threshold of 125 and a low threshold of 110 are set to close the fracture boundaries that are actually connected in the image, and obtain the fracture distribution result image in the field geological outcrop image.

[0032] Step 6: Select and annotate the crack distribution result image obtained in step 5, select the locations of cracks and holes, and annotate the categories and numbers.

[0033] Step 7: Use a convolutional neural network to extract features from the image selected and annotated in step 6, and use convolutional layers and fully connected layers to identify the bounding boxes of cracks and holes in the image to obtain the size of the crack bounding box.

[0034] Step 8: Based on the image after the cracks and holes are framed in step 7, the size of the selected cracks is calculated based on the size of the bounding box, according to the image size of 2cm×3cm and the scale of 1cm:100cm, including the crack length and crack width. The two largest scale crack size results obtained in the crack distribution image are shown in Table 1: Table 1 Crack size results .

[0035] By comparing the fracture distribution results obtained by the fusion algorithm with the artificial field geological outcrop fracture description results, it is shown that the fracture distribution results obtained by the fusion algorithm are more detailed, reducing the fracture description errors caused by human factors. Taking the fracture number "1" as an example, the size of the field geological outcrop fracture number "1" was actually measured, and the fracture length was 2.56m and the fracture width was 0.02m. The fitting rates of the fracture length and fracture width of the fracture number "1" obtained in step 8 were 95.17% and 90.91% respectively, with high fitting accuracy, indicating the reliability of the field geological outcrop fracture distribution and size recognition results based on the fusion algorithm.

[0036] The above description is only an illustrative embodiment of the present invention and is not intended to limit the scope of the present invention. The implementation steps and related parameters may vary. Any equivalent changes and modifications made without departing from the technical solution of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A method for intelligent identification of cracks in field geological outcrops based on a fusion algorithm, characterized in that: The Canny edge detection algorithm is used to identify the crack distribution results in the field geological outcrop images. The YOLOv5 algorithm is used to obtain the crack bounding box size in the field geological outcrop images. The crack length and width are obtained according to the crack bounding box size.

2. The intelligent identification method for field geological outcrop cracks based on a fusion algorithm according to claim 1 is characterized in that: The specific steps are: (1) Convert the field geological outcrop color image into grayscale image; (2) Use a Gaussian filter to smooth the grayscale image, remove the noise in the grayscale image, remove other factors that interfere with identification other than cracks, and obtain an image that only contains cracks; (3) Based on the image containing only the cracks, the gradient and direction are calculated using the Sobel filter to obtain the crack boundary in the image; (4) After obtaining the crack boundary in the image, the non-maximum suppression technique (NMS) is used to sharpen the image to obtain a clearer crack boundary recognition image; (5) Based on a clearer fracture boundary recognition image, a double threshold method is used to detect and connect fracture boundaries, and an image of fracture distribution results in the field geological outcrop image is obtained; (6) Select the locations of cracks and holes in the crack distribution result image, mark them with categories and numbers, and further divide the identified boundaries in the image into holes and cracks; (7) Using a convolutional neural network to extract features from the image selected and annotated in step (6), and using a convolutional layer and a fully connected layer to identify the bounding boxes of cracks and holes in the image to obtain the size of the crack bounding box; (8) Based on the image after the cracks and holes are selected in step (7), the length and width of the selected cracks are calculated based on the size of the crack boundary box, given the image size and scale.

3. The intelligent identification method for field geological outcrop cracks based on a fusion algorithm according to claim 2 is characterized in that: In step (1), the conversion formula used for grayscale processing is: ； In the formula, RGB is the industrial color standard, referring to red R, green G, and blue B respectively. A variety of colors can be obtained by mixing these three colors.

4. The intelligent identification method for field geological outcrop cracks based on a fusion algorithm according to claim 2 is characterized in that: In step (3), the gradient represents the degree and direction of change in the grayscale value, and the location where the grayscale change is strong is the boundary.

5. The method for intelligent identification of field geological outcrop cracks based on a fusion algorithm according to claim 2, characterized in that: In step (3), the calculation method used by the Sobel filter to calculate the gradient and direction is: ； ； In the formula, G x is the Sobel operator in the horizontal direction; G y is the Sobel operator in the vertical direction; θ is the direction of gray value change.

6. The intelligent identification method of field geological outcrop cracks based on fusion algorithm according to claim 2 is characterized in that: In step (4), the calculation method used by the non-maximum suppression technique NMS is: ； In the formula, M(x,y) Represents the gradient of pixels at different locations; T Represents the maximum value of the set gradient strength.

7. The intelligent identification method of field geological outcrop cracks based on fusion algorithm according to claim 2 is characterized in that: The specific steps of step (5) are: setting a high threshold and a low threshold, and obtaining a crack boundary contour map according to the high threshold; if the generated crack boundary image is not closed, collecting new crack boundaries through low threshold recognition, thereby closing the crack boundaries that are actually connected in the image.

Images