Method for Expanding Analysis of Geological Survey Data Samples in Karst Cave Areas for Pile Foundation Detection

Abstract

The present invention discloses a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection. A survey instrument is used to detect the geology at the beveled position. By separately using the beveled geological vector and combining two beveled geological vectors to form a triangular projection image, the geology at the unexplored depth is detected respectively. Through a temporal convolutional network and the judgment of shape changes, the geology is judged more accurately. After obtaining the depth prediction geological image for predicting a certain depth, a horizontal cutting method is adopted to perform sample interpolation on the geology at the same horizontal plane, so as to increase the samples. Due to the beveling and triangular combination prediction methods, the technical effect of making the added samples of geology at different positions more accurate is achieved. And the geological data samples are expanded, so as to improve the accuracy and reliability of pile foundation design.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more particularly, to a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection. Background Art

[0002] Currently, the karst cave area in pile foundation design refers to the place where engineering foundation design is carried out under geological conditions where there may be geological instability factors such as underground karst caves and karst. In such a special environment, the main challenges faced in pile foundation construction include: Geological complexity: Karst caves may cause the soil to be loose, with uncertain bearing capacity, posing a threat to the stability of the pile. The pile penetrates through the karst cave: If the pile needs to pass through the karst cave, it may encounter strata with high porosity and strong permeability, making it difficult to ensure the quality of the pile body. The geological survey data samples in the karst cave area are usually limited in number and cannot fully reflect the geological conditions, resulting in certain risks in engineering design. Therefore, more accurate samples need to be added. How to use an effective method to expand geological data samples, fill data gaps, and improve the coverage and accuracy of detection data has become an urgent technical problem to be solved. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection to solve the above problems existing in the prior art.

[0004] An embodiment of the present invention provides a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection, including:

[0005] Obtaining skew geological vectors at multiple horizontal positions and corresponding tilt angles; the values in the skew geological vectors represent geological data detected by a geological survey instrument when it penetrates into the ground at an inclination angle at a horizontal position.

[0006] Based on the horizontal positions and tilt angles corresponding to the skew geological vectors, pairwise skew geological vectors are subjected to triangular projection to obtain a plurality of triangular projection images.

[0007] Through a triangular discrimination network, based on the skew geological vectors at multiple horizontal positions and a plurality of triangular projection images, a depth prediction geological image is obtained; the depth prediction geological image contains an image of the predicted geological conditions at the unmeasured depth.

[0008] According to a plurality of skew geological vectors and the depth prediction geological image, data on the same horizontal plane is extracted to obtain a detection image; a plurality of detection images are obtained corresponding to a plurality of horizontal planes.

[0009] Based on the detection image and the depth prediction geological image, the geology on the same horizontal plane is detected to obtain additional sample data.

[0010] Optionally, obtaining the depth prediction geological image based on the oblique cutting geological vectors at multiple horizontal positions and multiple triangular projection images through the triangular discrimination network includes:

[0011] Predicting the geology of the unmeasured depth based on the multiple oblique cutting geological vectors through a single-line discrimination network to obtain a predicted oblique cutting depth vector;

[0012] Based on the predicted oblique cutting depth vector, marking the geology corresponding to the predicted oblique cutting depth vector on the first projection line and the second projection line in the triangular projection image to obtain a predicted depth image;

[0013] Predicting the area of the geology of the unmeasured depth based on the multiple predicted depth images through the triangular discrimination network to obtain a depth prediction geological image.

[0014] Optionally, obtaining the depth prediction geological image based on the multiple predicted depth images through the triangular discrimination network to predict the area of the geology of the unmeasured depth includes:

[0015] Dividing and marking the geology of the areas with the same geology based on the triangular projection image according to the oblique cutting geological vectors to obtain a segmented triangular image;

[0016] Successively extracting the segmented areas in the segmented triangular image to obtain multiple segmented triangular area images; one segmented triangular area image corresponds to one segmented area;

[0017] Judging the morphological change differences based on the multiple segmented triangular area images and predicting geological changes to obtain a predicted triangular geological morphology image;

[0018] Successively inputting the multiple segmented triangular area images into a second temporal convolutional network in sequence according to the depth to extract the quality of the segmented triangular areas to obtain a depth prediction geological image.

[0019] Optionally, predicting the geology of the unmeasured depth based on the multiple oblique cutting geological vectors through the single-line discrimination network to obtain a predicted oblique cutting depth vector includes:

[0020] Inputting the oblique cutting geological vector into a first temporal convolutional network to predict the geology of the unmeasured depth to obtain a first single-line geology;

[0021] Adding the first single-line geology to the oblique cutting geological vector;

[0022] According to the new oblique cutting geological vector, obtaining a second single-line geology through the first temporal convolutional network;

[0023] Obtaining single-line geologies at multiple depths correspondingly through the first temporal convolutional network multiple times;

[0024] The single-line geology at multiple depths forms a predicted inclined cutting depth geology vector in order from the earliest acquisition time to the latest acquisition time.

[0025] Optionally, the predicting the geological change and obtaining the predicted triangular geological morphology image based on the morphological change difference judgment of the multiple segmented triangular region images includes:

[0026] Based on the multiple segmented triangular region images, an adjacent change shape, a first geological region, and a second geological region are obtained; an adjacent change shape corresponds to two adjacent geology.

[0027] Obtain the labeled adjacent change shape in the database; the labeled adjacent change shape is the shape of the geological change corresponding to the adjacent geology stored.

[0028] If the adjacent change shape corresponding to two adjacent geology in the predicted depth image exists, input the first geological region and the adjacent change shape into the geological morphology network, extract the shape change, and obtain the predicted triangular geological morphology image;

[0029] If the adjacent change shape corresponding to two adjacent geology in the predicted depth image does not exist, input the first geological region and the labeled adjacent change shape into the geological morphology network, extract the morphological information, and obtain the predicted triangular geological morphology image.

[0030] Optionally, the obtaining the adjacent change shape based on the multiple segmented triangular region images includes:

[0031] Extract the boundaries of the geological regions segmented in two adjacent segmented triangular region images to obtain a first geological region and a second geological region; the depth corresponding to the first geological region is lower than the depth corresponding to the second geological region;

[0032] Obtain the midpoint of the first geological region to obtain a first midpoint; obtain the midpoint of the second geological region to obtain a second midpoint;

[0033] Match the first midpoint and the second midpoint, overlap the first geological region and the second geological region, and obtain a difference region; the difference region is the non-overlapping region of the first geological region and the second geological region;

[0034] Multiple triangular projection images respectively obtain multiple difference regions;

[0035] Overlay the multiple difference regions, extract the skeleton of the overlapping part, and obtain the adjacent change shape.

[0036] Optionally, the detecting the geology at the same horizontal plane based on the detection image and the depth predicted geology image to obtain additional sample data includes:

[0037] Arrange the geology at different depths in sequence according to the depth prediction geological image to obtain a geological vector;

[0038] Obtain a first detection position and a second detection position according to the detection image;

[0039] Obtain a first detected geology corresponding to the first detection position; obtain a second detected geology corresponding to the second detection position;

[0040] Find the midpoint between the first detection position and the second detection position in the detection image as the third detection position;

[0041] Find the geology corresponding to the midpoint between the first detected geology and the second detected geology in the geological vector to obtain the third detected geology;

[0042] Use the third detection position and the corresponding third detected geology as additional sample data.

[0043] Optionally, based on the horizontal position and the inclination angle corresponding to the skew geological vector, perform triangular projection on pairwise skew geological vectors to obtain a plurality of triangular projection images, including:

[0044] Fit a three-dimensional straight line according to the horizontal position and the inclination angle to obtain a first skew straight line and a second skew straight line;

[0045] Use the plane perpendicular to the ground and passing through the horizontal position corresponding to the first skew straight line and the horizontal position corresponding to the second skew straight line as the projection plane;

[0046] Project the first skew straight line and the second skew straight line onto the projection plane to obtain a first projection straight line and a second projection straight line;

[0047] Detect the vertical position of the intersection point of the first projection straight line and the second projection straight line to obtain the projection vertical depth;

[0048] If the projection vertical depth is greater than the depth threshold, use the triangular region formed by the first projection straight line and the second projection straight line on the projection plane to form a triangular projection image.

[0049] Optionally, the second temporal convolutional network is a temporal convolutional network that performs convolution using a two-dimensional convolutional kernel.

[0050] Optionally, according to the skew geological vector, based on the triangular projection image, divide and label the regions with the same geology to obtain a segmented triangular image, including:

[0051] Obtain a depth range according to the depths of the same geology in the skew geological vector;

[0052] Segment the area within the depth range in the triangular projection image, mark the geology, and obtain a segmented triangular image.

[0053] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0054] The embodiments of the present invention also provide a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection. The method includes: obtaining skew geological vectors at multiple horizontal positions and corresponding inclination angles; the values in the skew geological vectors represent geological data detected by a geological survey instrument when probing into the ground at an inclination angle at a horizontal position; based on the horizontal positions and inclination angles corresponding to the skew geological vectors, pairwise skew geological vectors are subjected to triangular projection to obtain multiple triangular projection images; through a triangular discrimination network, based on the skew geological vectors at multiple horizontal positions and multiple triangular projection images, a depth prediction geological image is obtained; the depth prediction geological image contains an image of the geology at the predicted un-surveyed depth; according to multiple skew geological vectors and the depth prediction geological image, data at the same horizontal plane is extracted to obtain a detection image; multiple horizontal planes correspond to multiple detection images; based on the detection image and the depth prediction geological image, the geology at the same horizontal plane is detected to obtain increased sample data.

[0055] In the present invention, a geological survey instrument is used to detect the geology at the skew positions. By separately using the skew geological vectors and combining two skew geological vectors to form a triangular projection image, the geology at the un-surveyed depth is detected respectively. Through a temporal convolutional network and the judgment of shape changes, the geology is judged more accurately. After obtaining the depth prediction geological image predicting a certain depth, a horizontal cutting method is adopted to perform sample interpolation on the geology at the same horizontal plane, thereby increasing samples. Due to the skew and triangular combination prediction method, the technical effect of making the increased samples of the geology at different positions more accurate is achieved. And the geological data samples are expanded, thereby improving the accuracy and reliability of pile foundation design. Description of the Drawings

[0056] Figure 1 is a flowchart of a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection provided by an embodiment of the present invention.

[0057] Figure 2 is a schematic diagram of a depth prediction geological image in a method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection provided by an embodiment of the present invention. Detailed Embodiments

[0058] The present invention will be described in detail below with reference to the accompanying drawings.

[0059] Embodiment 1

[0060] As Figure 1 shown, an embodiment of the present invention provides a method for expanding the analysis of geological survey data samples in a karst cave area for pile foundation detection, and the method includes:

[0061] S101: Obtain skew geological vectors at multiple horizontal positions and corresponding inclination angles; the values in the skew geological vectors represent geological data detected by a geological survey instrument when it penetrates the ground at an inclination angle at a horizontal position.

[0062] Among them, the geological data detected by the geological exploration instrument when it penetrates below the ground surface, such as the geological data in the karst cave area from top to bottom in depth are limestone deposits, hollow, water, and limestone deposits. The skew geological vectors are formed in sequence from top to bottom in depth according to the geological data.

[0063] Among them, the inclination angle represents the angle of the included angle with the straight line perpendicular to the ground surface.

[0064] S102: Based on the horizontal positions and inclination angles corresponding to the skew geological vectors, perform triangular projection on two skew geological vectors pairwise to obtain a plurality of triangular projection images.

[0065] S103: Through a triangular discrimination network, based on the skew geological vectors at multiple horizontal positions and a plurality of triangular projection images, obtain a depth prediction geological image. The depth prediction geological image includes an image of the predicted geology at the unmeasured depth.

[0066] Among them, the schematic diagram of the depth prediction geological image is as Figure 2 shown.

[0067] S104: According to a plurality of skew geological vectors and the depth prediction geological image, extract data on the same horizontal plane to obtain a detection image; a plurality of horizontal planes correspond to obtain a plurality of detection images.

[0068] S105: Based on the detection image and the depth prediction geological image, detect the geology on the same horizontal plane to obtain additional sample data.

[0069] Among them, the additional sample data includes the geology corresponding to the added positions on the horizontal plane corresponding to a certain depth. The same horizontal plane is at the same depth.

[0070] Optionally, the step of obtaining a depth prediction geological image through a triangular discrimination network based on the skew geological vectors at multiple horizontal positions and a plurality of triangular projection images includes:

[0071] Through a single-line discrimination network, based on the plurality of skew geological vectors, predict the geology at the unmeasured depth to obtain a predicted skew depth vector.

[0072] Among them, the predicted bevel depth vector is a vector obtained by adding the geology of the unmeasured depth as an element to the bevel depth vector.

[0073] Based on the predicted bevel depth vector, mark the geology corresponding to the predicted bevel depth vector on the first projection line and the second projection line in the triangular projection image to obtain a predicted depth image.

[0074] Among them, because the first projection line and the second projection line in the triangular projection image mark the geology corresponding to the bevel geology vector, but the corresponding extended parts are not marked. After obtaining the predicted bevel depth vector, mark according to the predicted bevel depth vector.

[0075] Through a triangular discrimination network, based on multiple predicted depth images, predict the area of the geology of the unmeasured depth to obtain a depth-predicted geology image.

[0076] Among them, predict the geology from two aspects: lines and planes.

[0077] Optionally, the step of obtaining a depth-predicted geology image by predicting the area of the geology of the unmeasured depth based on multiple predicted depth images through a triangular discrimination network includes:

[0078] According to the bevel geology vector, based on the triangular projection image, divide and mark the areas with the same geology to obtain a segmented triangular image.

[0079] Extract the segmented areas in the segmented triangular image in sequence to obtain a plurality of segmented triangular area images; one segmented triangular area image corresponds to one segmented area.

[0080] Among them, the position of a geology area in the segmented triangular area image is the same as the position of the area with the same geology in the triangular projection image.

[0081] Based on the plurality of segmented triangular area images, judge the morphological change difference, predict the geological change, and obtain a predicted triangular geology morphology image.

[0082] Among them, the predicted triangular geology morphology image is an image containing the morphology of the geology at the next depth. The next depth refers to the geology that is unmeasured and adjacent to the measured depth. The depth represents the vertical distance from the ground. In this embodiment, it is represented by the established coordinate system. The corresponding vertical coordinate is 0, and the position 200 meters deep from the horizontal plane has a vertical coordinate of 200.

[0083] Input the plurality of segmented triangular area images into the second temporal convolutional network in sequence according to the depth, extract the segmented triangular area quality, and obtain a depth-predicted geology image.

[0084] Among them, the second temporal convolutional network is a temporal convolutional network (TCN), which uses a two-dimensional convolutional kernel for convolution.

[0085] Optionally, the predicting the geology of the unmeasured depth based on the multiple oblique cutting geological vectors through the single-line discrimination network to obtain a predicted oblique cutting depth vector includes:

[0086] Input the oblique cutting geological vector into the first temporal convolutional network to predict the geology of the unmeasured depth and obtain the first single-line geology.

[0087] Among them, the first temporal convolutional network is a temporal convolutional network (TCN), which uses neurons for feature extraction.

[0088] Among them, predicting the geology at a fixed depth as the first single-line geology. In this embodiment, for example, the depth interval between each element in the oblique cutting geological vector is 1 meter, and the element with subscript 2 is 1 meter deeper than the element with subscript 1. Therefore, the depth corresponding to the first single-line geology is 1 meter deeper than the depth of the last element in the oblique cutting geological vector.

[0089] Add the first single-line geology to the oblique cutting geological vector.

[0090] Among them, add an element at the end of the oblique cutting geological vector, and the value of the element is the first single-line geology.

[0091] According to the new oblique cutting geological vector, obtain the second single-line geology through the first temporal convolutional network.

[0092] Among them, the input quantity of the first temporal convolutional network is fixed. If the input quantity is 10, the values of the last 10 elements of the oblique cutting geological vector each time are input into the first temporal convolutional network in ascending order of subscript.

[0093] Obtain the single-line geology at multiple depths by passing through the first temporal convolutional network correspondingly multiple times.

[0094] Among them, the single-line geology at multiple depths includes the first single-line geology and the second single-line geology.

[0095] The single-line geology at multiple depths constitutes a predicted oblique cutting depth geology vector in sequence from the earliest acquisition time to the latest acquisition time.

[0096] Optionally, the judging the morphological change difference based on the multiple segmented triangular region images, predicting the geological change, and obtaining a predicted triangular geological morphology image includes:

[0097] Based on the multiple segmented triangular region images, obtain adjacent change shapes, a first geological region, and a second geological region; two adjacent geologies correspond to an adjacent change shape.

[0098] Among them, the adjacent changing shape depends on two adjacent geological formations in the predicted depth image.

[0099] Obtain the labeled adjacent changing shape in the database; the labeled adjacent changing shape is the shape of the geological change corresponding to the adjacent geological formations stored.

[0100] Among them, the labeled adjacent changing shape in the database includes the shapes of geological changes between multiple adjacent geological formations.

[0101] If the adjacent changing shape corresponding to two adjacent geological formations in the predicted depth image exists, input the first geological region and the adjacent changing shape into the geological morphology network, extract the shape change, and obtain the predicted triangular geological morphology image.

[0102] Among them, in this embodiment, the geological morphology network is a Convolutional Neural Networks (CNN).

[0103] If the adjacent changing shape corresponding to two adjacent geological formations in the predicted depth image does not exist, input the first geological region and the detected difference region into the geological morphology network, extract the morphological information, and obtain the predicted triangular geological morphology image.

[0104] Optionally, the predicting the geological change and obtaining the predicted triangular geological morphology image by judging the morphological change difference based on the multiple segmented triangular region images includes:

[0105] Extract the boundaries of the geological regions segmented in two adjacent segmented triangular region images to obtain a first geological region and a second geological region; the depth corresponding to the first geological region is lower than the depth corresponding to the second geological region.

[0106] Obtain the midpoint of the first geological region to get a first midpoint; obtain the midpoint of the second geological region to get a second midpoint.

[0107] Among them, the midpoint of the smallest bounding box containing the first geological region is used as the midpoint of the first geological region; the midpoint of the smallest bounding box containing the second geological region is used as the midpoint of the second geological region.

[0108] Match the first midpoint and the second midpoint, overlap the first geological region and the second geological region, and obtain a difference region; the difference region is the non-overlapping region of the first geological region and the second geological region.

[0109] Multiple difference regions are obtained corresponding to multiple triangular projection images;

[0110] Overlay the multiple difference regions, extract the skeleton of the overlapping part, and obtain the adjacent changing shape.

[0111] Among them, the overlapping part represents the part that exists in all the difference regions. The Zhang-Suen thinning algorithm is used for skeleton extraction, and the extracted skeleton is used as the adjacent changing shape.

[0112] Optionally, detecting the geology at the same horizontal plane based on the detected image and the depth-predicted geology image to obtain additional sample data includes:

[0113] Arranging the geology at different depths in sequence according to the depth-predicted geology image to obtain a geology vector.

[0114] Obtaining a first detection position and a second detection position according to the detected image.

[0115] Obtaining a first detected geology corresponding to the first detection position; obtaining a second detected geology corresponding to the second detection position.

[0116] Finding the midpoint between the first detection position and the second detection position in the detected image as the third detection position.

[0117] Finding the geology corresponding to the midpoint between the first detected geology and the second detected geology in the geology vector to obtain the third detected geology.

[0118] Taking the third detection position and the corresponding third detected geology as the additional sample data.

[0119] Optionally, performing triangular projection on two-by-two oblique geological vectors based on the horizontal position and the inclination angle corresponding to the oblique geological vector to obtain a plurality of triangular projection images, including:

[0120] Fitting a three-dimensional straight line according to the horizontal position and the inclination angle to obtain a first oblique cutting line and a second oblique cutting line.

[0121] Among them, the three-dimensional straight line equation is Ax + By + Cz + D = 0, where x is the value of the abscissa, y is the value of the ordinate, z is the value of the vertical coordinate, and A, B, C, and D are the corresponding parameters. By solving the equation, the values of A / B, B / C, and A / C can be calculated from the inclination angle, and the value of D can be calculated from the horizontal position, so as to obtain the values of A, B, C, and D.

[0122] Taking the plane perpendicular to the ground and passing through the horizontal positions corresponding to the first oblique cutting line and the second oblique cutting line as the projection plane.

[0123] Projecting the first oblique cutting line and the second oblique cutting line onto the projection plane to obtain a first projection line and a second projection line.

[0124] Among them, the three-dimensional first oblique cutting line and the second oblique cutting line can be transformed into two-dimensional first projection line and second projection line.

[0125] Detect the vertical position of the intersection point of the first projection line and the second projection line to obtain the projection vertical depth.

[0126] If the projection vertical depth is greater than the depth threshold, draw the triangular region formed by the first projection line and the second projection line on the projection plane as a triangular projection image.

[0127] Among them, in this embodiment, the depth threshold is 100 meters.

[0128] Among them, the plot function is used for image drawing.

[0129] Optionally, the second temporal convolutional network is a temporal convolutional network (TCN), and two-dimensional convolutional kernels are used for convolution.

[0130] Optionally, the step of segmenting and labeling the geology of the regions with the same geology based on the triangular projection image according to the oblique geological vector to obtain a segmented triangular image includes:

[0131] Obtain the depth range according to the depths of the same geology in the oblique geological vector.

[0132] Among them, if the depth of 50-60 meters is the same geology, that is, the cave is empty, then the depth range is 50-60 meters. Segment the region within the depth range in the triangular projection image, label the geology, and obtain a segmented triangular image.

Claims

1. A method for expanding and analyzing geological survey data samples in a karst cave area for pile foundation detection, characterized in that: include: Obtaining oblique geological vectors and corresponding inclination angles at multiple horizontal positions; The value in the oblique geological vector represents the geological data detected by the geological survey instrument penetrating into the ground at an inclined angle at a horizontal position; Based on the horizontal position and the tilt angle corresponding to the oblique geological vector, the oblique geological vectors are triangulated and projected in pairs to obtain a plurality of triangulated projection images; A depth prediction geological image is obtained based on the oblique geological vectors at multiple horizontal positions and multiple triangular projection images through a triangulated discriminant network; the depth prediction geological image includes an image of the geology at a predicted unsurveyed depth; According to multiple oblique geological vectors and depth prediction geological images, data of the same horizontal plane is extracted to obtain a detection image; multiple detection images are obtained corresponding to multiple horizontal planes; Based on the detection image and the depth prediction geological image, the geology of the same horizontal plane is detected to obtain additional sample data; The method of obtaining a depth prediction geological image based on oblique geological vectors at multiple horizontal positions and multiple triangular projection images through a triangular discriminant network includes: Predicting the geology of an unsurveyed depth based on the plurality of oblique geological vectors by a single line discriminant network to obtain a predicted oblique depth vector; Based on the predicted oblique cutting depth vector, marking the geology corresponding to the predicted oblique cutting depth vector with the first projection straight line and the second projection straight line in the triangular projection image to obtain a predicted depth image; Using a triangulated discriminant network, based on a plurality of predicted depth images, a geological area of ​​an unsurveyed depth is predicted to obtain a depth predicted geological image; The method of detecting the geology of the same horizontal plane based on the detection image and the depth prediction geological image to obtain additional sample data includes: According to the depth prediction geological image, the geology at different depths is arranged in sequence to obtain a geological vector; According to the detection image, obtaining a first detection position and a second detection position; A first detection geology is obtained according to the first detection position; a second detection geology is obtained according to the second detection position; Find the midpoint between the first detection position and the second detection position in the detection image as the third detection position; Find the geology corresponding to the midpoint between the first detection geology and the second detection geology in the geology vector to obtain the third detection geology; The third detection position and the corresponding third detection geology are used as additional sample data.

2. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 1, characterized in that: The method predicts the geological area of ​​the unsurveyed depth based on the multiple predicted depth images through the triangulated discriminant network to obtain the depth predicted geological image, including: According to the oblique geological vector, based on the triangular projection image, the geologically identical area is segmented and marked, thereby obtaining a segmented triangular image; Sequentially extracting the segmented regions in the segmented triangular image to obtain a plurality of segmented triangular region images; one segmented triangular region image corresponds to one segmented region; Based on the multiple segmented triangular area images, determine the morphological change differences, predict geological changes, and obtain a predicted triangular geological morphological image; Multiple segmented triangular area images are sequentially input into the second time convolution network according to the depth, the quality of the segmented triangular area is extracted, and the depth prediction geological image is obtained.

3. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 1, characterized in that: The method of predicting the geology of an unsurveyed depth based on the plurality of oblique geological vectors through a single line discrimination network to obtain a predicted oblique depth vector includes: Inputting the oblique geological vector into the first time convolutional network, predicting the geology at the unsurveyed depth, and obtaining the first single-line geology; adding the first single-line geology to the oblique geological vector; According to the new oblique geological vector, the second single-line geology is obtained through the first time convolution network; Multiple passes through the first time convolutional network correspond to obtaining single-line geology at multiple depths; The single-line geology at multiple depths constitutes the predicted oblique cutting depth geological vector in sequence from early to late according to the acquisition time.

4. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 2, characterized in that: The method of determining the difference in morphological changes based on the plurality of segmented triangular region images, predicting geological changes, and obtaining a predicted triangular geological morphological image comprises: Based on the multiple segmented triangular region images, adjacent change shapes, a first geological region and a second geological region are obtained; two adjacent geological regions correspond to one adjacent change shape; Obtaining the annotated adjacent change shape in the database; the annotated adjacent change shape is the shape of the geological change corresponding to the stored adjacent geological features; If adjacent change shapes corresponding to two adjacent geologies in the predicted depth image exist, the first geological region and the adjacent change shapes are input into the geological morphological network, the shape changes are extracted, and a predicted triangular geological morphological image is obtained; If the adjacent change shapes corresponding to two adjacent geologies in the predicted depth image do not exist, the first geological region and the annotated adjacent change shapes are input into the geological morphological network to extract morphological information and obtain the predicted triangulated geological morphological image.

5. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 4, characterized in that: The method of obtaining adjacent changed shapes based on a plurality of segmented triangular region images includes: Extracting the boundaries of the geological regions segmented in two adjacent segmented triangular region images to obtain a first geological region and a second geological region; the depth corresponding to the first geological region is lower than the depth corresponding to the second geological region; Obtaining the midpoint of the first geological region to obtain a first midpoint; obtaining the midpoint of the second geological region to obtain a second midpoint; matching the first midpoint and the second midpoint, overlapping the first geological region and the second geological region, and obtaining a difference region; the difference region is a region where the first geological region and the second geological region do not overlap; Multiple triangular projection images correspond to obtaining multiple difference areas; Multiple difference regions are superimposed, and the skeleton of the overlapping parts is extracted to obtain adjacent variation shapes.

6. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 1, characterized in that: Based on the horizontal position and the tilt angle corresponding to the oblique geological vector, the oblique geological vectors are triangulated and projected in pairs to obtain a plurality of triangulated projection images, including: According to the horizontal position and the inclination angle, a three-dimensional straight line is fitted to obtain a first oblique straight line and a second oblique straight line; A plane perpendicular to the ground and passing through the horizontal position corresponding to the first oblique straight line and the horizontal position corresponding to the second oblique straight line is used as the projection plane; Projecting the first oblique cutting line and the second oblique cutting line onto the projection surface to obtain a first projection line and a second projection line; detecting the vertical position of the point where the first projection line and the second projection line intersect to obtain a projection vertical depth; If the projection vertical depth is greater than a depth threshold, a triangular area formed by the first projection straight line and the second projection straight line on the projection surface is used to form a triangular projection image.

7. The method for expanding and analyzing the sample of geological survey data in a karst cave area for pile foundation detection according to claim 2, characterized in that: The second temporal convolutional network is a temporal convolutional network that uses a two-dimensional convolution kernel for convolution.

8. The method for expanding and analyzing geological survey data samples in karst cave areas for pile foundation detection according to claim 2, characterized in that: The method of segmenting and marking the geologically identical areas according to the oblique geological vector and based on the triangular projection image to obtain a segmented triangular image includes: Obtaining a depth range according to the depth of the same geology in the oblique geological vector; The area within the depth range in the triangular projection image is segmented, the geology is marked, and a segmented triangular image is obtained.

Images