Intelligent analysis and recognition method of 3D seismic point cloud data

Abstract

The present invention relates to the field of seismic wave data processing. In order to realize automatic prediction, a three-dimensional seismic wave point cloud data intelligent analysis and recognition method is provided. By directly standardizing the point cloud data generated by the seismic wave method, and through the corresponding relationship in engineering experience, the machine intelligently recognizes the geological objects and automatically generates the prediction conclusion. The method bypasses the process of raw data to image and then to manual interpretation, avoiding many problems in manual image analysis. At the same time, a large amount of test raw data is streamlined and structured, saving storage space and the amount of data processing for subsequent work, laying the foundation for machine intelligent fusion analysis between data under different prediction methods.

Description

Technical Field

[0001] The invention relates to the field of seismic wave data processing, and in particular to a method for intelligent analysis and recognition of three-dimensional seismic wave point cloud data. Background Art

[0002] At present, seismic wave reflection imaging technologies such as TRT and TSP are widely used in the field of tunnels and underground engineering. Data is received through seismic wave reflection, and the software calculates and generates images to analyze and predict the geological conditions of the unexcavated tunnel section in front of the tunnel. It is an important means of advanced geological forecasting for underground engineering. At present, the process from receiving raw data to data mapping is mostly developed and realized by domestic and foreign equipment manufacturers, while the process from image to prediction result is to manually identify and analyze the image to further determine the spatial distribution law of prediction objects such as adverse geological bodies.

[0003] The prior art has the following deficiencies:

[0004] 1. In the traditional forecasting mode, field tests are first conducted, and the test data is calculated by the built-in algorithm of the equipment to form a graph, which is the final result of the machine operation. After that, the image needs to be analyzed manually, and the forecast conclusion is obtained based on experience. Under this forecasting process, the prediction is highly subjective, the analysis of the results is subject to many interference factors, the accuracy is low, and the analysis efficiency is low.

[0005] 2. Images are difficult to analyze quantitatively, so it is difficult to convert them into a general data structure and enter the advanced forecast analysis system, and it is also difficult to conduct data fusion analysis with other digitized advanced forecasting methods.

[0006] 3. There is no standardized forecasting system and machine recognition rules.

[0007] 4. The amount of raw test data is large and difficult to analyze directly. Summary of the invention

[0008] In order to realize automatic prediction, the present application provides a method for intelligent analysis and identification of three-dimensional seismic point cloud data.

[0009] The technical solution adopted by the present invention to solve the above problems is:

[0010] The intelligent analysis and recognition method of three-dimensional seismic point cloud data includes:

[0011] Step 1, collecting tunnel 3D seismic point cloud data;

[0012] Step 2: classify the seismic point cloud data according to the reflection morphology classification rules;

[0013] Step 3, obtaining the positions of various types of reflection interfaces, and obtaining the probability of existence of bad geological bodies corresponding to each reflection morphology classification according to the relationship between the preset reflection morphology classification and the probability of existence of bad geological bodies;

[0014] Step 4: Accumulate the existence probabilities of bad geological bodies corresponding to different reflection morphology classifications at the same location to obtain the existence probability of the bad geological body at the location.

[0015] Furthermore, after obtaining the positions of various types of morphological reflection interfaces, step 3 further includes: performing linear analysis between various types of reflection interfaces, and obtaining the probability of existence of bad geological bodies corresponding to the linear relationship based on the relationship between the preset linear interface and the probability of existence of bad geological bodies.

[0016] Furthermore, after obtaining the positions of various types of morphological reflection interfaces, step 3 further includes: obtaining the wave velocity change trend at the position, and obtaining the probability of existence of the bad geological body corresponding to the wave velocity change trend according to the preset relationship between the wave velocity change trend and the probability of existence of the bad geological body.

[0017] Further, the reflection morphology classification includes:

[0018] According to the morphology, it can be divided into planar continuous reflection, planar discontinuous reflection, staggered reflection, dense point reflection and sporadic point reflection;

[0019] According to the morphological characteristics of positive and negative reflections, they are divided into positive and negative interval distribution, negative reflection as the main feature, positive reflection as the main feature, negative reflection only, and positive reflection only;

[0020] The two classification results are combined to obtain the final classification.

[0021] Furthermore, the classification rules are as follows:

[0022] Surface continuous reflection:

[0023] Detect and extract the smallest F% data from all seismic point cloud data, detect the filtered data, and remove the sporadic data with non-continuous spatial distribution according to the coordinate distribution corresponding to each data. Finally, compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it is a planar continuous reflection with only negative reflection.

[0024] Detect and extract the largest F% of data from all seismic point cloud data, detect the filtered data, and remove the sporadic data with non-continuous distribution in space according to the coordinate distribution corresponding to each data. Finally, compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it is a planar continuous reflection and only positive reflection;

[0025] Search among the detected continuous reflections of each surface, first determine the distance A, if there are both positive reflections and negative reflections within the distance range, then determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive reflections and the negative reflections;

[0026] Planar discontinuity reflection:

[0027] Detect and extract the smallest F% data from all seismic wave point cloud data, detect the filtered data, eliminate the data with non-continuous distribution in space according to the coordinate distribution, and finally compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it is a planar discontinuous reflection and only negative reflection;

[0028] Detect and extract the largest F% of data from all seismic wave point cloud data, detect the filtered data, eliminate the data with non-continuous distribution in space according to the coordinate distribution, and finally compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it is a planar discontinuous reflection and only positive reflection;

[0029] Search among the detected planar discontinuous reflections, first determine the distance C, if there are both positive and negative reflections within the range, then determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive and negative reflections;

[0030] Disconnected reflection:

[0031] On the basis of all planar discontinuous reflections, the staggered distance and distribution elevation of each positive reflection and negative reflection surface are detected respectively, specifically: a planar discontinuous reflection of a positive reflection is searched within the range of ±Dm, if another planar discontinuous reflection of a positive reflection is found, and the vertical distribution range is complementary to the detected object, then these two planar discontinuous reflections and only positive reflections are merged into staggered reflections and only positive reflections; similarly, staggered reflections and only negative reflections can be obtained;

[0032] Determine the distance E, search among the detected staggered reflections, if there are both positive and negative reflections within the distance E, determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive and negative reflections;

[0033] Dense point reflection and scattered point reflection:

[0034] After the above tests are completed, the remaining seismic wave point cloud data are screened, and the maximum and minimum G% data are retained. According to the distribution density of the coordinates of the sporadic numerical points, they are divided into dense point reflections and sporadic point reflections. According to the number of positive and negative value distributions, they are divided into positive and negative interval distribution, mainly negative reflections, mainly positive reflections, only positive reflections, or only negative reflections;

[0035] A, B, C, D, E, F, G are preset values.

[0036] Furthermore, the step 2 also includes encoding processing of reflection forms of different categories.

[0037] Furthermore, after obtaining the position of each reflection form classification, step 3 also includes encoding the position.

[0038] Furthermore, the step 4 also includes structural processing of the data based on coding before calculating the probability of the existence of the bad geological body.

[0039] Furthermore, the method further comprises step 5, adjusting the probability of existence of the bad geological body corresponding to the classification rules and the reflection morphology classification according to the actual situation of the bad geological body after excavation.

[0040] Compared with the prior art, the present invention has the following beneficial effects: the present invention directly standardizes the original point cloud data generated by the seismic wave method, and uses the corresponding relationship in engineering experience to perform machine intelligent identification of geological objects and automatically generate forecast conclusions. This method bypasses the process from original data to image and then to manual interpretation, avoiding many problems in manual image analysis, and at the same time simplifies and structures a large amount of test original point cloud data, saving storage space and the amount of data processing for subsequent work, laying the foundation for machine intelligent fusion analysis between data under different forecasting methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of the intelligent analysis and recognition method for three-dimensional seismic point cloud data. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] like Figure 1 As shown, the intelligent analysis and recognition method of three-dimensional seismic point cloud data includes:

[0044] Step 1: Collect 3D seismic point cloud data of the tunnel. This embodiment is explained based on the TRT results. The point cloud data of a single detection is a set of positive and negative values ​​with normal distribution, where the median is the normal state, and the positive and negative values ​​correspond to positive reflection, negative reflection and reflection intensity respectively; in addition, each value also corresponds to a set of spatial coordinates.

[0045] Step 2: Classify the seismic point cloud data according to the reflection morphology classification rules.

[0046] In order to improve the accuracy of classification and reduce data interference, the seismic wave data needs to be filtered before classification. This embodiment is divided into four types according to the morphology: planar continuous reflection, planar discontinuous reflection, staggered reflection, and point reflection. Among them, point reflection can be further subdivided into dense point reflection and sporadic point reflection. According to the positive and negative values, it can be divided into negative reflection and positive reflection. The relationship between positive and negative reflections can be subdivided into five types: positive and negative interval distribution, negative reflection as the main reflection, positive reflection as the main reflection, only negative reflection, and only positive reflection. Combining the two classifications can obtain the final classification result type.

[0047] The specific classification rules are:

[0048] (1) Surface continuous reflection

[0049] Detect and extract the smallest F% data in all seismic point cloud data. In this embodiment, F is taken as 20. Then the filtered data is further detected, and non-adjacent data in space are eliminated according to the coordinate distribution. Finally, the maximum value range of the Y coordinate in the filtered data is compared with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it is identified as a planar continuous reflection, and the conclusion is only negative reflection.

[0050] In order to facilitate subsequent data processing, reduce the amount of data processing, and save data storage space, each type is coded separately, and the coding can be used in subsequent processing instead of the classification of each reflection form. For example, the above-mentioned continuous reflection with only negative reflection is coded as 14.

[0051] The largest 20% of all values ​​are detected and tested. The non-adjacent data in space are eliminated according to the coordinate distribution. Finally, the maximum value range of the Y coordinate in the filtered data is compared with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it can be concluded that the surface is continuously reflected and only positive reflection is obtained, and it is coded as 15.

[0052] Further search is performed among the detected planar continuous reflections. First, the distance A is determined. If there are only multiple groups of positive reflections or multiple groups of negative reflections within the distance range, the code is retained unchanged. If there are both positive and negative reflections within the distance range, it is determined based on the quantitative relationship between the positive and negative reflections whether they are distributed at positive and negative intervals, whether negative reflections are dominant, or whether positive reflections are dominant. The corresponding codes are 11, 12, 13, and the original codes 14 or 15 within the distance are deleted. If there are only positive reflections or negative reflections, the original code is retained.

[0053] (2) Planar discontinuous reflection

[0054] Detect and extract the smallest 20% of all values, detect the filtered data, eliminate non-adjacent data in space according to the coordinate distribution, and finally compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it is determined to be a planar discontinuous reflection with only negative reflection, and the code is 24.

[0055] The largest 20% of all values ​​are detected and tested. The non-adjacent data in space are eliminated according to the coordinate distribution. Finally, the maximum value range of the Y coordinate in the filtered data is compared with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it can be concluded that the surface is discontinuously reflected and only has positive reflection, which is coded as 25.

[0056] Further search is performed among the detected planar discontinuous reflections. First, the distance C is determined. If there are only multiple groups of positive reflections or multiple groups of negative reflections within the distance range, the original code is retained unchanged. If there are both positive and negative reflections within the distance range, it is determined based on the quantitative relationship between the positive and negative reflections whether it is a positive and negative interval distribution, mainly negative reflections, or mainly positive reflections. The corresponding codes are 21, 22, 23, and the original codes 24 or 25 within the distance are deleted. If there are only positive reflections or negative reflections, the original code is retained.

[0057] (3) Disconnected reflection

[0058] On the basis of all planar discontinuous reflections, the distance and distribution elevation of each reflecting surface are detected, and the positive and negative reflecting surfaces need to be detected separately. For example, if a positive reflecting planar discontinuous reflection is searched within the range of ±Dm, D can be 10. If another positive reflecting planar discontinuous reflection is found, and the vertical distribution range is complementary to the detected object, then the two previous surfaces coded as 25 are merged into 35, and the code 34 can be obtained in the same way;

[0059] After the above detection is completed, further detect the assigned staggered reflections. According to the quantitative relationship between positive and negative reflections within the distance E, determine whether it is a positive and negative interval distribution, negative reflections are dominant, or positive reflections are dominant. The corresponding codes are 31, 32, 33, and the original codes within this distance (codes 34~35) are deleted.

[0060] (4) Dense point reflection and scattered point reflection

[0061] After the above detection is completed, the remaining numerical points are screened, and the maximum and minimum G% are retained, and G is taken as 20. At this time, according to the distribution density of the coordinates of the sporadic numerical points, they are divided into dense point reflection and sporadic point reflection (such as defining the hit points ≥50 in a 1m×1m×1m space, and searching in units of 0.2m in the direction of the tunnel axis). Further, according to the distribution ratio of positive and negative values, they are defined and coded 41~55 respectively (for example, if 55 points meet the requirements in a single search and the positive reflection points are greater than 70%, the value is assigned 43; if the proportion of positive reflection points is between 50% and 70%, the value is assigned 41, and if the proportion is greater than 90%, the value is assigned 45).

[0062] A, C, D, and E are preset distances, which can be set according to actual needs, such as 20m~50m. If the value is too large, the accuracy will be reduced, and if it is too small, it will be meaningless. B, F, and G can also be selected according to actual needs, and there is no restriction here.

[0063] Step 3: Obtain the positions of various types of reflection interfaces, and obtain the probability of existence of bad geological bodies corresponding to each reflection morphology classification based on the relationship between the preset reflection morphology classification and the probability of existence of bad geological bodies.

[0064] The above steps complete the classification and assignment of point clouds to various types of reflection space forms. In addition, the distribution space position of each code corresponding to the reflection form needs to be described and encoded. According to the preset, the starting surface of a single test (the section perpendicular to the axis of the tunnel) is defined as the starting point, and the position of the detected object can be defined according to the distribution range of the X coordinates of each object. According to the nature of the test, the effective value of a single test is generally within 100m, so a four-digit number is used for encoding, such as 4055, indicating that the distribution range of the object is 40m~55m from the starting section of the test. When retrieving the distribution range, some data on both sides are generally eliminated, such as 10% of the coordinate data on both sides, to prevent the distribution range of the retrieval from being too affected by discrete points. In addition, the real pile number information of a single test should be recorded in the system. When outputting the conclusion, the system directly adds the distribution range on the basis of the real pile number to obtain the real spatial position of each object.

[0065] The probability of existence of bad geological bodies corresponding to each reflection form classification can be obtained by looking up a table. The probability table of this embodiment is as follows:

[0066] Table 1 Probability of existence of unfavorable geological bodies corresponding to each reflection morphology classification

[0067]

[0068] Furthermore, after the above detection is completed, all planar continuous reflections and planar discontinuous reflections are subjected to linear analysis between reflection surfaces. First, under the condition of looking down (generally two of the three coordinate values ​​are extracted at a time, such as XY values, i.e., plane distribution), each reflection surface is simplified to the coordinates of two points (starting point and end point), and replaced with a straight line. Under this condition, the starting point and end point of each adjacent reflection surface are connected respectively. If the connecting line is approximately a straight line (based on the approximate judgment of the fitting degree), it is considered that there is a linear image between the reflection surfaces. At this time, according to the type of the original reflection surface constituting the straight line, it is divided into a linear structure formed between positive reflection surfaces, a linear structure formed between negative reflection surfaces, or a linear structure formed by positive and negative reflection surfaces, and the values ​​are assigned to 61 to 63 respectively. Repeat the above process under the condition of side view (i.e., XZ direction, where X is the tunnel axis direction) and assign values.

[0069] The probability of existence of unfavorable geological bodies corresponding to special structural surface types is shown in the following table:

[0070] Table 2 Probability of existence of unfavorable geological bodies corresponding to special structural surface types

[0071]

[0072] In addition to the above information, other relevant information is generally included when collecting raw data, such as wave velocity diagram. The absolute value of the wave velocity diagram is greatly affected by the collection, but the waveform trend can be used as an auxiliary judgment factor. Therefore, the change trend of the wave velocity at each reflection surface position can be counted and encoded with reference to the wave velocity of the excitation surface, as shown in Table 3. The wave velocity change trend can be preset and defined, such as defining that the average wave velocity value at the judgment position is greater than 30% compared with the average wave velocity value within 20m from the position to the excitation direction, which is a rapid rise; 10%~30% is a slight rise; less than 10% is basically consistent, etc.

[0073] Table 3 Probability of existence of unfavorable geological bodies corresponding to the wave velocity change trend

[0074]

[0075] Step 4: Accumulate the existence probabilities of bad geological bodies corresponding to different reflection morphology classifications at the same location to obtain the existence probability of bad geological bodies at that location. If it exceeds 100%, it will be calculated as 100%.

[0076] If code 11 and code 24 exist in the range of 4055 at the same time, the probability of occurrence of the geological interface in this range is superimposed and output as 85% based on the probability of occurrence of their geological interfaces (+30%, +55%) respectively.

[0077] In order to improve the prediction accuracy, the classification rules and output probabilities can be modified according to the actual characteristics of the unfavorable geological body after excavation.

[0078] For example, if 10 sections are counted and the surface detection is 11, the corresponding probability preset value is +30%. However, excavation confirms that there are only 2 sections with geological interfaces, so the actual probability is 20%. Therefore, the initial value corresponding to 11 is modified to +20%.

[0079] When multiple geophysical methods are used to predict adverse geological bodies, fusion analysis between data is involved. Similarly, the data under the geophysical methods with the same data structure are structured according to this method. At this time, the prediction weight of each method needs to be defined, and finally the probability of occurrence of each geological interface after comprehensive analysis is obtained.

[0080] After a single point cloud data scan is completed, all reflective surfaces and special structural surfaces are converted into corresponding codes according to the rules. At this time, the structure of the coded data is specified and standardized, with the aim of normalizing various data streams under different detection methods. Later, the machine can perform mutual operations on the data obtained under different methods according to the preset algorithm to achieve the purpose of comprehensive analysis. The identified data is generally arranged in a matrix format, with each row containing relevant information of a single reflective surface. The columns express different meanings according to different positions. When there is no data or meaning in a certain position, a letter is used instead. Examples are as follows:

[0081] XX YY 4 0 5 5 1 2 2

[0082] XX 6 1 3 5 4 2 BB 1

[0083] For example, if a certain project uses 11-bit coding, 1~2 bits represent the project location. Currently, no information has been input and is replaced by the letter X. 3~4 bits indicate whether it is a linear image between objects. The first line indicates no and is replaced by the letter Y, and the second line is 61. 5~8 bits are position distribution information, indicating the range of distribution position pile numbers. 9~10 bits are reflection surface feature information. Since the second line is a linear image between objects, no data is input for the feature information position and is replaced by a letter. The last bit is other information, such as the trend of wave velocity changes.

[0084] By arranging the reflective surfaces according to this rule, the reflective surfaces under certain characteristics can be quickly screened and identified, and calculations can also be performed quickly. For example, in the above example, the reflective surfaces in the first row are distributed at 40-55m, and the reflective surfaces in the second row are distributed at 35-42m. After cross calculation, it is found that both reflective surfaces in the range of 40-42m exist at the same time.

Claims

1. A method for intelligent analysis and identification of three-dimensional seismic point cloud data, characterized in that: include: Step 1, collecting tunnel 3D seismic point cloud data; Step 2: classify the seismic point cloud data according to the reflection morphology classification rules; Reflection morphology classification includes: According to the morphology, it is divided into planar continuous reflection, planar discontinuous reflection, staggered reflection, dense point reflection and sporadic point reflection; according to the morphological characteristics of positive and negative reflections, it is divided into positive and negative interval distribution, negative reflection as the main reflection, positive reflection as the main reflection, only negative reflection and only positive reflection; the two classification results are combined to obtain the final classification; Step 3, obtaining the positions of various types of reflection interfaces, and obtaining the probability of existence of bad geological bodies corresponding to each reflection morphology classification according to the relationship between the preset reflection morphology classification and the probability of existence of bad geological bodies; Step 4: Accumulate the existence probabilities of bad geological bodies corresponding to different reflection morphology classifications at the same location to obtain the existence probability of the bad geological body at the location.

2. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 1 is characterized in that: After obtaining the positions of various types of reflection interfaces, step 3 further includes: performing linear analysis between various types of reflection interfaces, and obtaining the probability of existence of bad geological bodies corresponding to the linear relationship based on the relationship between the preset linear interface and the probability of existence of bad geological bodies.

3. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 1 is characterized in that: After obtaining the positions of various types of morphological reflection interfaces, step 3 further includes: obtaining the wave velocity change trend at the position, and obtaining the probability of existence of the bad geological body corresponding to the wave velocity change trend according to the preset relationship between the wave velocity change trend and the probability of existence of the bad geological body.

4. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 1 is characterized in that: The specific classification rules are: Surface continuous reflection: Detect and extract the smallest F% data from all seismic point cloud data, detect the filtered data, and remove the sporadic data with non-continuous distribution in space according to the coordinate distribution corresponding to each data. Finally, compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it is a planar continuous reflection with only negative reflection. Detect and extract the largest F% of data from all seismic point cloud data, detect the filtered data, and remove the sporadic data with non-continuous distribution in space according to the coordinate distribution corresponding to each data. Finally, compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If it is greater than or equal to the tunnel diameter, it is a planar continuous reflection and only positive reflection; Search among the detected continuous reflections of each surface, first determine the distance A, if there are both positive reflections and negative reflections within the distance range, then determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive reflections and the negative reflections; Planar discontinuity reflection: Detect and extract the smallest F% data from all seismic wave point cloud data, detect the filtered data, eliminate the data with non-continuous distribution in space according to the coordinate distribution, and finally compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it is a planar discontinuous reflection and only negative reflection; Detect and extract the largest F% of data from all seismic wave point cloud data, detect the filtered data, eliminate the data with non-continuous distribution in space according to the coordinate distribution, and finally compare the maximum value range of the Y coordinate in the filtered data with the spatial coordinate range of the tunnel diameter. If the maximum length is less than the tunnel diameter or the preset value B, it is a planar discontinuous reflection and only positive reflection; Search among the detected planar discontinuous reflections, first determine the distance C, if there are both positive and negative reflections within the range, then determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive and negative reflections; Disconnected reflection: On the basis of all planar discontinuous reflections, the staggered distance and distribution elevation of each positive reflection and negative reflection surface are detected respectively, specifically: a planar discontinuous reflection of a positive reflection is searched within the range of ±Dm, if another planar discontinuous reflection of a positive reflection is found, and the vertical distribution range is complementary to the detected object, then these two planar discontinuous reflections and only positive reflections are merged into staggered reflections and only positive reflections; similarly, staggered reflections and only negative reflections can be obtained; Determine the distance E, search among the detected staggered reflections, if there are both positive and negative reflections within the distance E, determine whether it is positive and negative interval distribution, negative reflection is dominant or positive reflection is dominant according to the quantitative relationship between the positive and negative reflections; Dense point reflection and scattered point reflection: After the above tests are completed, the remaining seismic wave point cloud data are screened, and the maximum and minimum G% data are retained. According to the distribution density of the coordinates of the sporadic numerical points, they are divided into dense point reflections and sporadic point reflections. According to the number of positive and negative value distributions, they are divided into positive and negative interval distribution, mainly negative reflections, mainly positive reflections, only positive reflections, or only negative reflections; A, B, C, D, E, F, G are preset values.

5. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 1 is characterized in that: The step 2 also includes encoding the reflection forms of different categories.

6. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 5, characterized in that: Step 3 also includes encoding the position after obtaining the position of each reflection form classification.

7. The intelligent analysis and identification method of three-dimensional seismic point cloud data according to claim 6, characterized in that: The step 4 also includes structural processing of the data based on coding before calculating the probability of the existence of the bad geological body.

8. The method for intelligent analysis and identification of three-dimensional seismic point cloud data according to any one of claims 1 to 7, characterized in that: The method further includes step 5, adjusting the probability of existence of the bad geological body corresponding to the classification rules and the reflection morphology classification according to the actual situation of the bad geological body after excavation.

Images