A high-level landslide risk monitoring method and system

Abstract

The present invention provides a method and system for monitoring the risk of high-level landslides, which relates to the field of landslide monitoring technology and ensures the detail level of the characterization data of the first feature splicing vector. The present invention then calculates the risk monitoring data of the mountain object based on the first feature splicing vector, so that the risk monitoring data can characterize the sharing coefficient between the contents of several mountain risk data sets to be processed, thereby improving the accuracy of risk monitoring of the data. The present invention performs multi-task training of artificial intelligence analysis threads based on the first regression analysis type of the example mountain object, the second regression analysis type, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector of the hidden danger event information of several example mountain risk data sets, and the first reference type and the second reference type of each example mountain risk data set, which not only improves the thread analysis efficiency, but also can accurately determine the risk of high-level landslides, thereby improving the accuracy of monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of landslide monitoring, and in particular to a method and system for monitoring high-level landslide risks. Background Art

[0002] Due to the large vertical drop between the shear outlet and the front edge of the high-position long-distance landslide, the source area is mostly located on the top of a ridge with a steep slope, so that the sliding body hits the slope after starting, collide and disintegrate, and is severely fragmented. During the movement, the dynamic erosion of loose surface deposits causes a sharp increase in volume. In addition, the movement path of the sliding body may turn multiple times, and the sliding body will also show a collision acceleration process at the turning point. High-position long-distance landslides are generally characterized by high concealment, fast movement speed, large volume, fluidization, block fragmentation and dynamic erosion.

[0003] High-altitude landslides have always been a disaster that is difficult to monitor in geology. High-altitude landslide disasters are extremely dangerous and pose huge safety hazards. Therefore, a monitoring technology is urgently needed to improve the above technical problems. Summary of the invention

[0004] In order to improve the technical problems existing in the related technologies, the present invention provides a high-level landslide risk monitoring method and system.

[0005] In a first aspect, a high-level landslide risk monitoring method is provided, comprising:

[0006] Obtaining hidden danger event information and mountain risk data information of several to-be-processed mountain risk data sets corresponding to the target mountain object;

[0007] Determine, by using an artificial intelligence analysis thread, hidden danger event feature vectors corresponding to hidden danger event information of the plurality of mountain risk data sets to be processed, and mountain risk data set feature vectors corresponding to mountain risk data information of the plurality of mountain risk data sets to be processed;

[0008] Using the artificial intelligence analysis thread to perform splicing processing on the hidden danger event feature vector and the mountain risk data set feature vector to obtain a first feature splicing vector of the target mountain object;

[0009] Determining risk monitoring data of the target mountain object by using the first feature splicing vector, wherein the risk monitoring data is used to represent a sharing coefficient between types of the plurality of mountain risk data sets to be processed;

[0010] Among them, the artificial intelligence analysis thread is obtained by training the original analysis thread using the regression analysis data and the first reference type and the second reference type of each example mountain risk data set after processing the hidden danger event information and mountain risk data information of several example mountain risk data sets corresponding to the example mountain object using the original analysis thread to obtain the regression analysis data of the example mountain object. The regression analysis data includes the first regression analysis type and the second regression analysis type of the example mountain object, the target hidden danger event feature vector of the hidden danger event information of the several example mountain risk data sets, and the regression analysis hidden danger event feature vector; the second reference type is a local type of the first reference type, and the second regression analysis type is a local type of the first regression analysis type.

[0011] In the present invention, the step of determining the risk monitoring data of the target mountain object by using the first feature splicing vector includes:

[0012] The first feature concatenation vector is processed by using different convolution modules in the first decision unit in the artificial intelligence analysis thread to obtain a plurality of hidden danger description feature vectors; different convolution units exist between two random convolution modules;

[0013] Determine the difference coefficients between the plurality of hidden danger description feature vectors, and determine the risk monitoring data of the target mountain object according to the difference coefficients.

[0014] In the present invention, the step of determining the difference coefficients between the plurality of hidden danger description feature vectors and determining the risk monitoring data of the target mountain object according to the difference coefficients includes:

[0015] Performing function calculation on the plurality of hidden danger description feature vectors to obtain difference coefficients of the plurality of hidden danger description feature vectors;

[0016] The difference coefficient is subjected to dimensionless simplification processing, and the dimensionless simplified difference coefficient is determined as the risk monitoring data of the target mountain object.

[0017] In the present invention, the step of determining the risk monitoring data of the target mountain object by using the first feature splicing vector includes:

[0018] Performing key description content analysis on the first feature splicing vector to obtain key description content of the first feature splicing vector;

[0019] The risk monitoring data of the target mountain object is determined according to the key description content of the first feature stitching vector.

[0020] In the present invention, the method further comprises:

[0021] Obtaining hidden danger event information, mountain risk data information and marking information of a plurality of example mountain risk data sets corresponding to the example mountain objects; the marking information includes a first reference category and a second reference category of each example mountain risk data set;

[0022] Determine, by using the original analysis thread, a target hidden danger event feature vector of the hidden danger event information of the plurality of example mountain risk data sets and a mountain risk data set feature vector of the mountain risk data information of the plurality of example mountain risk data sets;

[0023] Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the several example mountain risk data sets are processed to obtain the regression analysis data of the example mountain object;

[0024] The original analysis thread is trained using the regression analysis data, the first reference category and the second reference category of each example mountain risk data set to obtain an artificial intelligence analysis thread.

[0025] In the present invention, the original analysis thread is used to process the target hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the several example mountain risk data sets to obtain the regression analysis data of the example mountain object, including:

[0026] Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the several example mountain risk data sets are processed to obtain regression analysis hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and splicing feature vectors of the several example mountain risk data sets, wherein the regression analysis hidden danger event feature vector is generated based on the hidden danger hidden feature vector of the target hidden danger event feature vector;

[0027] Using the original analysis thread, the spliced ​​feature vectors of the several example mountain risk data sets are split and processed to obtain second feature splicing vectors corresponding to the hidden danger event information of the several example mountain risk data sets and third feature splicing vectors corresponding to the mountain risk data information of the several example mountain risk data sets, and determining a fourth feature splicing vector of the example mountain object according to the second feature splicing vector and the third feature splicing vector;

[0028] Processing the third feature splicing vector using the second decision unit in the original analysis thread to obtain a first regression analysis type of the example mountain object, and processing the fourth feature splicing vector using the first decision unit in the original analysis thread to obtain a second regression analysis type of the example mountain object;

[0029] The first regression analysis type, the second regression analysis type, the target hidden danger event feature vector, and the regression analysis hidden danger event feature vector are determined as regression analysis data of the example mountain object.

[0030] In the present invention, the original analysis thread is trained using the regression analysis data, the first reference type and the second reference type of each example mountain risk data set to obtain an artificial intelligence analysis thread, including:

[0031] Determining a first data comparison result using the first regression analysis type and a first reference type of each of the example mountain risk data sets;

[0032] Determining a second data comparison result using the second regression analysis type and a second reference type of each of the example mountain risk data sets;

[0033] Determine a third data comparison result using the target hidden danger event feature vector and the regression analysis hidden danger event feature vector;

[0034] The thread coefficient of the original analysis thread is debugged using the first data comparison result, the second data comparison result and the third data comparison result to obtain an artificial intelligence analysis thread.

[0035] In the present invention, the step of debugging the thread coefficient of the original analysis thread by using the first data comparison result, the second data comparison result and the third data comparison result to obtain the artificial intelligence analysis thread includes:

[0036] Performing key description content analysis on the fourth feature stitching vector to obtain key description content of the fourth feature stitching vector;

[0037] Determine a weight coefficient using key description content of the fourth feature splicing vector;

[0038] Determine a target data comparison result using the weight coefficient, the first data comparison result, the second data comparison result, and the third data comparison result;

[0039] The target data comparison result is used to debug the thread coefficient of the original analysis thread to obtain an artificial intelligence analysis thread.

[0040] In the present invention, the step of obtaining the hidden danger event information and mountain risk data information of a plurality of to-be-processed mountain risk data sets corresponding to the target mountain object includes:

[0041] Obtaining several to-be-processed mountain risk data sets corresponding to the target mountain object;

[0042] Perform hidden danger event information screening processing on each of the mountain risk data sets to be processed, and obtain hidden danger event information of each of the mountain risk data sets to be processed, wherein the hidden danger event information includes one or more of the landslide hidden danger directory information of the mountain risk data set to be processed and the hidden danger event screening information of the landslide data node;

[0043] Mountain risk data information screening processing is performed on each of the mountain risk data sets to be processed to obtain mountain risk data information of each of the mountain risk data sets to be processed, wherein the mountain risk data information includes landslide risk element information of the landslide data nodes of the mountain risk data sets to be processed.

[0044] In a second aspect, a high-level landslide risk monitoring system is provided, comprising a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the above method.

[0045] A high-level landslide risk monitoring method and system provided by an embodiment of the present invention processes the hidden danger event information and mountain risk data information of several to-be-processed mountain risk data sets corresponding to a mountain object through an artificial intelligence analysis thread to obtain a first feature splicing vector corresponding to the mountain object. Since the first feature splicing vector is obtained based on the splicing of multi-dimensional data of the mountain object, it can better characterize the feature information of the overall content of the mountain object, and ensure the detail level of the characterization data of the first feature splicing vector. The present invention calculates the risk monitoring data of the mountain object based on the first feature splicing vector, so that the risk monitoring data can characterize the sharing coefficient between the contents of several to-be-processed mountain risk data sets, thereby improving the accuracy of risk monitoring of the data. In addition, the present invention performs multi-task training of the artificial intelligence analysis thread based on the first regression analysis type of the example mountain object, the second regression analysis type, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector of the hidden danger event information of several example mountain risk data sets, and the first reference type and the second reference type of each example mountain risk data set, which not only improves the thread analysis efficiency, but also can accurately determine the high-level landslide risk, thereby improving the monitoring accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 The present invention provides a flowchart of a method for monitoring high-altitude landslide risk. DETAILED DESCRIPTION

[0048] In order to better understand the above technical scheme, the technical scheme of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical scheme of the present invention, rather than limitations on the technical scheme of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0049] See also Figure 1 , shows a method for monitoring high-level landslide risks, which may include the technical solutions described in the following steps S201-S204.

[0050] S201. Obtain hidden danger event information and mountain risk data information of a plurality of to-be-processed mountain risk data sets corresponding to a target mountain object.

[0051] For example, the mountain object is a mountain with a high altitude and a large slope. Under such conditions, it is difficult for relevant technicians to bring precision equipment to conduct specific surveys, so the survey data obtained is inaccurate. When monitoring, there will be problems of inaccurate monitoring. Therefore, high-altitude landslides are a problem that is currently difficult to overcome. Therefore, relevant technicians conceived and implemented the present invention, which is of great help to the problems currently faced and makes a huge contribution in this field.

[0052] In the embodiment of the present invention, a plurality of to-be-processed mountain risk data sets corresponding to the target mountain object are first obtained. Since each to-be-processed mountain risk data set may include hidden danger event information and landslide risk element information, such as landslide data nodes in the mountain risk data set, the hidden danger event information and mountain risk data information of each to-be-processed mountain risk data set are then obtained. By obtaining multi-dimensional information of the to-be-processed mountain risk data set of the target mountain object for subsequent mountain object monitoring, the data detail level can be guaranteed, and the accuracy and credibility of the risk monitoring data of the mountain object can be improved.

[0053] In a possible implementation example, the above-mentioned obtaining of hidden danger event information and mountain risk data information of a plurality of to-be-processed mountain risk data sets corresponding to the target mountain object may be implemented according to the following steps:

[0054] (a1) Obtain several unprocessed mountain risk data sets corresponding to the target mountain object.

[0055] (a2) Screening and processing the hidden danger event information of each mountain risk data set to be processed to obtain the hidden danger event information of each mountain risk data set to be processed, wherein the hidden danger event information includes one or more of the landslide hidden danger directory information of the mountain risk data set to be processed and the hidden danger event screening information of the landslide data node.

[0056] In one possible implementation, for hidden danger event information, the landslide hidden danger directory information, theme information, etc. corresponding to each mountain risk data set to be processed can be directly obtained to determine as the hidden danger event information of the mountain risk data set to be processed; or the landslide data nodes of each mountain risk data set to be processed can be identified and processed to obtain hidden danger event screening information, and the hidden danger event screening information can be determined as the hidden danger event information of the mountain risk data set to be processed.

[0057] (a3) screening and processing the mountain risk data information of each mountain risk data set to be processed to obtain the mountain risk data information of each mountain risk data set to be processed, wherein the mountain risk data information includes the landslide risk factor information of the landslide data node of the mountain risk data set to be processed.

[0058] In a possible implementation embodiment, for mountain risk data information, mountain risk data information screening processing can be performed on each mountain risk data set to be processed to obtain one or several landslide data nodes screened for each mountain risk data set to be processed, and then the landslide risk element information of the landslide data node extracted from each mountain risk data set to be processed is determined as the mountain risk data information of the mountain risk data set to be processed.

[0059] S202. Determine, using an artificial intelligence analysis thread, hidden danger event feature vectors corresponding to hidden danger event information of several mountain risk data sets to be processed, and mountain risk data set feature vectors corresponding to mountain risk data information of several mountain risk data sets to be processed.

[0060] Exemplarily, mountain risk data include: the looseness of mountain soil or rock, whether a fire has occurred in the mountain, whether there is a fault zone in the area, the amount of vegetation in the area, and rainfall information, etc. There are many factors that affect the occurrence of landslides, and I will not list them one by one again.

[0061] In an embodiment of the present invention, the first feature screening thread in the artificial intelligence analysis thread can be used to perform feature screening processing on the hidden danger event information of several mountain risk data sets to be processed, and obtain hidden danger event feature vectors corresponding to the hidden danger event information of several mountain risk data sets to be processed. The second feature screening thread in the artificial intelligence analysis thread can be used to perform feature screening processing on the mountain risk data information of several mountain risk data sets to be processed, and obtain mountain risk data set feature vectors corresponding to the mountain risk data information of several mountain risk data sets to be processed. Among them, the first feature screening thread and the second feature screening thread can be the same feature screening thread or different feature screening threads. The feature screening thread can be obtained by pre-training based on example data, so as to improve the accuracy of the feature representation of the hidden danger event feature vector and the mountain risk data set feature vector, and perform mountain object monitoring based on accurate hidden danger event feature vectors and mountain risk data set feature vectors, thereby ensuring the accuracy of risk monitoring data.

[0062] In one possible implementation, an artificial intelligence analysis thread may be used to determine a feature vector corresponding to the hidden danger event information of each mountain risk data set to be processed, and then the feature vectors corresponding to the hidden danger event information of each mountain risk data set to be processed are spliced ​​(or fused) to obtain a hidden danger event feature vector corresponding to the overall hidden danger event information of several mountain risk data sets to be processed. Similarly, an artificial intelligence analysis thread may be used to determine a feature vector corresponding to the mountain risk data information of each mountain risk data set to be processed, and then the feature vectors corresponding to the mountain risk data information of each mountain risk data set to be processed are spliced ​​(or fused) to obtain a mountain risk data set feature vector corresponding to the overall mountain risk data information of several mountain risk data sets to be processed.

[0063] S203, using an artificial intelligence analysis thread to perform splicing processing on the hidden danger event feature vector and the mountain risk data set feature vector to obtain a first feature splicing vector of the target mountain object.

[0064] Exemplarily, the characteristics of hidden danger events include: first, geological conditions and geomorphological conditions; second, the influence of internal and external forces (dynamics) and human activities.

[0065] Among them, splicing can be understood as fusion, integration, etc.

[0066] Artificial intelligence analysis threads may include neural network threads, etc.

[0067] In the embodiment of the present invention, the first feature splicing vector of the target mountain object is finally obtained by inputting the hidden danger event feature vector and the mountain risk data set feature vector into the artificial intelligence analysis thread for processing. The first feature splicing vector is used to characterize the feature information of the overall content of the target mountain object. Since the first feature splicing vector is obtained by multi-dimensional data splicing based on the hidden danger event information and the mountain risk data information of the mountain object, the first feature splicing vector can characterize the hidden danger feature information of the mountain risk data set content corresponding to several to-be-processed mountain risk data sets of the target mountain object, thereby ensuring the detail of the characterization data of the first feature splicing vector.

[0068] In a possible implementation, the hidden danger event feature vectors corresponding to the hidden danger event information of several to-be-processed mountain risk data sets can be spliced ​​to obtain the spliced ​​hidden danger event feature vector corresponding to the hidden danger event information, and the mountain risk data set feature vectors corresponding to the mountain risk data information of several to-be-processed mountain risk data sets can be spliced ​​to obtain the spliced ​​mountain risk data set feature vector corresponding to the mountain risk data information. Then, the spliced ​​hidden danger event feature vector and the spliced ​​mountain risk data set feature vector are processed using an artificial intelligence analysis thread to obtain a first feature splicing vector of the target mountain object.

[0069] Exemplarily, for hidden danger event information of several mountain risk data sets to be processed, the feature vectors of the hidden danger event information of each mountain risk data set to be processed can be spliced ​​to obtain a spliced ​​hidden danger event feature vector.

[0070] For the mountain risk data information of several mountain risk data sets to be processed, the characteristic vectors of the mountain risk data information of each mountain risk data set to be processed can be grouped based on the first category of each mountain risk data set to be processed, and then the characteristic vectors in each group can be spliced ​​according to the grouping, and then the spliced ​​feature vectors of each group are spliced ​​to obtain the spliced ​​mountain risk data set feature vector.

[0071] In a possible implementation example, an artificial intelligence analysis thread can be used to process the stitching hidden danger event feature vector and the stitching mountain risk data set feature vector based on the mountain risk data set analysis method, and finally obtain the first feature stitching vector of the target mountain object.

[0072] S204: Determine risk monitoring data of the target mountain object using the first feature splicing vector, where the risk monitoring data is used to represent a sharing coefficient between types of a plurality of mountain risk data sets to be processed.

[0073] Among them, risk monitoring data is used to describe the possibility of landslides in the area. For example, the risk levels are divided into three types: no risk, low risk and high risk. No risk can be level 0, low risk can be levels 1-5, and high risk can be levels 6-10. The possibility of landslides can be determined according to the above description, which can protect people's lives and property safety as much as possible, and can also carry out disaster chain prevention and control work in advance.

[0074] Exemplarily, the sharing coefficient can be understood as similarity. Risk types can be understood as risk types under natural conditions and risk types under artificial conditions. Specifically, the types under natural conditions include: rainfall risk types and soil structure risk types. Risk types under artificial conditions include: vibration risk types caused by large equipment operation and blasting vibration risk types.

[0075] In an embodiment of the present invention, since the first feature stitching vector is obtained based on the hidden danger event information and mountain risk data information of several mountain risk data sets to be processed corresponding to the target mountain object, the risk monitoring data of the target mountain object determined by the first feature stitching vector can, to a certain extent, characterize the sharing coefficient between the contents of several mountain risk data sets to be processed.

[0076] In a possible implementation embodiment, the artificial intelligence analysis thread is obtained by training the original analysis thread using the regression analysis data and the first reference category and the second reference category of each example mountain risk data set after processing the hidden danger event information and mountain risk data information of several example mountain risk data sets corresponding to the example mountain object using the original analysis thread to obtain the regression analysis data of the example mountain object. The second reference category is a local category of the first reference category, and the second regression analysis category is a local category of the first regression analysis category. The regression analysis data of the example mountain object includes the first regression analysis category of the example mountain object, the second regression analysis category, the target hidden danger event feature vector of the hidden danger event information of several example mountain risk data sets, and the regression analysis hidden danger event feature vector.

[0077] Based on the above embodiments, the beneficial effects of the present invention are as follows: the present invention proposes to process the hidden danger event information and mountain risk data information of several to-be-processed mountain risk data sets corresponding to the mountain object through an artificial intelligence analysis thread to obtain the first feature splicing vector corresponding to the mountain object. Since the first feature splicing vector is obtained based on the multi-dimensional data splicing of the mountain object, it can better characterize the feature information of the overall content of the mountain object, thereby ensuring the detail level of the characterization data of the first feature splicing vector. The present invention then calculates the risk monitoring data of the mountain object based on the first feature splicing vector, so that the risk monitoring data can characterize the sharing coefficient between the contents of several to-be-processed mountain risk data sets. The risk monitoring data is determined in combination with the first feature splicing vector of the mountain object, thereby improving the accuracy of risk monitoring.

[0078] The present invention does not require strategy formulation, but reflects the risk monitoring data of the mountain object through robustness, and can ultimately measure the risk monitoring data of the mountain object through dimensionless simplification processing. The first implementation method of determining the risk monitoring data of the target mountain object using the first feature splicing vector is introduced in detail below through steps S301-S302:

[0079] S301. Use different convolution modules in the first decision unit in the artificial intelligence analysis thread to process the first feature splicing vector to obtain a plurality of hidden danger description feature vectors.

[0080] S302: Determine the difference coefficients between a number of hidden danger description feature vectors, and determine the risk monitoring data of the target mountain object according to the difference coefficients.

[0081] In the embodiment of the present invention, based on the idea of ​​thread disturbance, the difference coefficients between several hidden danger description feature vectors are calculated, and then the risk monitoring data of the target mountain object is determined according to the difference coefficients. The difference coefficients between several hidden danger description feature vectors are used to represent the degree of difference between several hidden danger description feature vectors, and the difference coefficients can be the variances of several hidden danger description feature vectors.

[0082] In a possible implementation example, the above-mentioned determination of the difference coefficients between a plurality of hidden danger description feature vectors and determination of the risk monitoring data of the target mountain object according to the difference coefficients can be implemented according to the following steps:

[0083] (b1) Perform function calculation on several hidden danger description feature vectors to obtain the difference coefficients of several hidden danger description feature vectors.

[0084] In a possible implementation, a difference coefficient is obtained by performing function calculation on several hidden danger description feature vectors, so that the difference coefficient can characterize the relevance of the contents of several to-be-processed mountain risk data sets of the mountain object to a certain extent. If the difference coefficient is smaller, the risk monitoring data is smaller, indicating that the content relevance of several to-be-processed mountain risk data sets of the target mountain object is higher; if the difference coefficient is larger, the risk monitoring data is larger, indicating that the content relevance of several to-be-processed mountain risk data sets of the target mountain object is lower.

[0085] (b2) Perform dimensionless simplification on the difference coefficient, and determine the dimensionless simplified difference coefficient as the risk monitoring data of the target mountain object.

[0086] In a possible implementation example, the dimensionless simplified difference coefficient is finally determined as the risk monitoring data of the target mountain object.

[0087] In a possible implementation example, before the target mountain object is processed by the artificial intelligence analysis thread, the following thread training process may also be included:

[0088] (c1) Obtaining hidden danger event information, mountain risk data information and marking information of several example mountain risk data sets corresponding to the example mountain objects; the marking information includes the first reference category and the second reference category of each example mountain risk data set.

[0089] In an embodiment of the present invention, the example mountain object refers to a mountain object used for thread training of the original analysis thread, and several example mountain risk data sets may refer to part or all of the mountain risk data set content that has been released by the example mountain object. Several example mountain risk data sets released by the example mountain object may be obtained from a database. First, several example mountain risk data sets corresponding to the example mountain object are obtained. Since each example mountain risk data set may include hidden danger event information and landslide risk element information, such as landslide data nodes in the mountain risk data set, the hidden danger event information and mountain risk data information of each example mountain risk data set are then obtained. By obtaining multi-dimensional information of the example mountain risk data set of the example mountain object for subsequent thread training, the training effect can be improved.

[0090] Among them, the specific process of obtaining the hidden danger event information and mountain risk data information of several example mountain risk data sets corresponding to the example mountain objects can be found in the description of steps (a1-a3) in the above embodiment, which will not be repeated in this embodiment.

[0091] (c2) using the original analysis thread to determine target hidden danger event feature vectors of hidden danger event information of several example mountain risk data sets and mountain risk data set feature vectors of mountain risk data information of several example mountain risk data sets.

[0092] In an embodiment of the present invention, the first feature screening thread in the original analysis thread can be used to perform feature screening processing on the hidden danger event information of several example mountain risk data sets, and obtain hidden danger event feature vectors corresponding to the hidden danger event information of several example mountain risk data sets. The second feature screening thread in the original analysis thread can be used to perform feature screening processing on the mountain risk data information of several example mountain risk data sets, and obtain mountain risk data set feature vectors corresponding to the mountain risk data information of several example mountain risk data sets. Among them, the first feature screening thread and the second feature screening thread can be the same feature screening thread or different feature screening threads. The feature screening thread can be obtained by pre-training based on the example data, so as to improve the accuracy of the feature representation of the hidden danger event feature vector and the mountain risk data set feature vector.

[0093] (c3) Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of several example mountain risk data sets are processed to obtain the regression analysis data of the example mountain objects.

[0094] In the embodiment of the present invention, the regression analysis data includes a first regression analysis type of a sample mountain object, a second regression analysis type, a target hidden danger event feature vector of hidden danger event information of several sample mountain risk data sets, and a regression analysis hidden danger event feature vector.

[0095] In a possible implementation example, the above step (c3) can be implemented according to the following steps:

[0096] (c31) Use the original analysis thread to process the target hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of several example mountain risk data sets, and obtain the regression analysis hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the splicing feature vectors of several example mountain risk data sets. The regression analysis hidden danger event feature vector is generated based on the hidden danger hidden feature vector of the target hidden danger event feature vector.

[0097] (c32) Use the original analysis thread to split the spliced ​​feature vectors of several example mountain risk data sets to obtain the second feature splicing vectors corresponding to the hidden danger event information of several example mountain risk data sets and the third feature splicing vectors corresponding to the mountain risk data information of several example mountain risk data sets, and determine the fourth feature splicing vector of the example mountain object based on the second feature splicing vector and the third feature splicing vector.

[0098] In an embodiment of the present invention, the original analysis thread can be used to split the splicing feature vectors of several example mountain risk data sets to obtain a second feature splicing vector corresponding to the hidden danger event information and a third feature splicing vector corresponding to the mountain risk data information, and then the second feature splicing vector and the third feature splicing vector are spliced ​​to obtain a fourth feature splicing vector of the example mountain object.

[0099] (c33), using the second decision unit in the original analysis thread to process the third feature splicing vector to obtain the first regression analysis type of the example mountain object, and using the first decision unit in the original analysis thread to process the fourth feature splicing vector to obtain the second regression analysis type of the example mountain object.

[0100] In the embodiment of the present invention, different decision units can be used to process the intermediate data (including the third feature stitching vector and the fourth feature stitching vector) in the original analysis thread processing process to obtain the first regression analysis type and the second regression analysis type of the example mountain object.

[0101] (c34) determining the first regression analysis type, the second regression analysis type, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector as the regression analysis data of the example mountain object.

[0102] In the embodiment of the present invention, the regression analysis data includes the first regression analysis type and the second regression analysis type of the example mountain object, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector corresponding to the hidden danger event information of several example mountain risk data sets corresponding to the mountain object. Thread training is performed through the above regression analysis data and the marked data of each example mountain risk data set (including the first reference type and the second reference type of each example mountain risk data set).

[0103] (c4) Using the regression analysis data, the first reference category and the second reference category of each example mountain risk data set to train the original analysis thread, and obtain the artificial intelligence analysis thread.

[0104] In a possible implementation example, the above step c4 can be implemented according to the following steps:

[0105] (c41) Determine a first data comparison result using a first regression analysis type and a first reference type of each example mountain risk data set.

[0106] In a possible implementation, the first data comparison result can be determined based on the data comparison results between the first regression analysis type of the example mountain object and the first reference type of each example mountain risk data set, and then based on the data comparison results between the first regression analysis type of the example mountain object and the first reference type of each example mountain risk data set (for example, the data comparison results between the first regression analysis type of the example mountain object and the first reference type of each example mountain risk data set are added to determine the first data comparison result).

[0107] (c42) Determine a second data comparison result using a second regression analysis type and a second reference type for each example mountain risk data set.

[0108] In a possible implementation, the second data comparison result can be determined based on the data comparison results between the second regression analysis type of the example mountain object and the second reference type of each example mountain risk data set, and then based on the data comparison results between the second regression analysis type of the example mountain object and the second reference type of each example mountain risk data set (for example, the data comparison results between the second regression analysis type of the example mountain object and the second reference type of each example mountain risk data set are added to determine the second data comparison result).

[0109] (c43) Determine the third data comparison result by using the target hidden danger event feature vector and the regression analysis hidden danger event feature vector.

[0110] In a possible implementation embodiment, the target hidden danger event feature vector can be regarded as intelligent monitoring information of the regression analysis hidden danger event feature vector, and the data comparison result between the target hidden danger event feature vector and the regression analysis hidden danger event feature vector can be determined as the third data comparison result.

[0111] (c44) Debugging the thread coefficient of the original analysis thread using the first data comparison result, the second data comparison result and the third data comparison result to obtain an artificial intelligence analysis thread.

[0112] In one possible implementation, the first data comparison result, the second data comparison result and the third data comparison result may be added together to determine the total loss, and then the thread coefficient of the original analysis thread may be debugged using the total loss to obtain an artificial intelligence analysis thread.

[0113] The present invention can determine the risk monitoring data of the mountain object through the key description content of the mountain object feature vector. The second implementation method of determining the risk monitoring data of the target mountain object by using the first feature splicing vector is described in detail below through steps S501-S502:

[0114] S501: Perform key description content analysis on the first feature splicing vector to obtain key description content of the first feature splicing vector.

[0115] In the embodiment of the present invention, since the first feature stitching vector represents the hidden danger feature information of the mountain risk data set content of the mountain object, the key description content of the first feature stitching vector is obtained by performing key description content analysis on the first feature stitching vector.

[0116] S502: Determine risk monitoring data of the target mountain object according to the key description content of the first feature splicing vector.

[0117] In an embodiment of the present invention, the key description content of the first feature stitching vector can be directly determined as the risk monitoring data of the target mountain object; or the key description content of the first feature stitching vector can be dimensionally simplified to obtain a dimensionless simplified result, and then the dimensionless simplified result can be determined as the risk monitoring data of the target mountain object.

[0118] In a possible implementation example, before the target mountain object is processed by the artificial intelligence analysis thread, the following thread training process may also be included:

[0119] (d1) Obtaining hidden danger event information, mountain risk data information and marking information of several example mountain risk data sets corresponding to the example mountain objects; the marking information includes the first reference category and the second reference category of each example mountain risk data set.

[0120] (d2) Determine target hidden danger event feature vectors of hidden danger event information of several example mountain risk data sets and mountain risk data set feature vectors of mountain risk data information of several example mountain risk data sets using the original analysis thread.

[0121] (d3) Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of several example mountain risk data sets are processed to obtain the regression analysis data of the example mountain objects.

[0122] In a possible implementation example, the above step (d3) can be implemented according to the following steps:

[0123] (d31) Use the original analysis thread to process the target hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of several example mountain risk data sets to obtain the regression analysis hidden danger event feature vectors of the hidden danger event information of several example mountain risk data sets and the splicing feature vectors of several example mountain risk data sets. The regression analysis hidden danger event feature vector is generated based on the hidden danger hidden feature vector of the target hidden danger event feature vector.

[0124] (d32) Use the original analysis thread to split the spliced ​​feature vectors of several example mountain risk data sets to obtain the second feature splicing vectors corresponding to the hidden danger event information of several example mountain risk data sets and the third feature splicing vectors corresponding to the mountain risk data information of several example mountain risk data sets, and determine the fourth feature splicing vector of the example mountain object based on the second feature splicing vector and the third feature splicing vector.

[0125] (d33), using the second decision unit in the original analysis thread to process the third feature splicing vector to obtain the first regression analysis type of the example mountain object, and using the first decision unit in the original analysis thread to process the fourth feature splicing vector to obtain the second regression analysis type of the example mountain object.

[0126] (d34) Determine the first regression analysis type, the second regression analysis type, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector as the regression analysis data of the example mountain object.

[0127] (d4) Using the regression analysis data, the first reference category and the second reference category of each example mountain risk data set to train the original analysis thread, and obtain an artificial intelligence analysis thread.

[0128] In the embodiment of the present invention, the regression analysis hidden danger event feature vector is used for the MLM training task, and the regression analysis hidden danger event feature vector is generated based on the hidden danger hidden feature vector of the target hidden danger event feature vector. Specifically, the original analysis thread performs hidden processing on the target hidden danger event feature vector of the hidden danger event information of several example mountain risk data sets to obtain the hidden hidden danger event feature vector, and then uses the original analysis thread to perform regression analysis processing on the hidden hidden danger event feature vector to obtain the regression analysis hidden danger event feature vector.

[0129] In a possible implementation example, the above step d4 can be implemented according to the following steps:

[0130] (d41) Determine a first data comparison result using a first regression analysis type and a first reference type of each example mountain risk data set.

[0131] (d42) Determine a second data comparison result using a second regression analysis type and a second reference type for each example mountain risk data set.

[0132] (d43) Determine the third data comparison result by using the target hidden danger event feature vector and the regression analysis hidden danger event feature vector.

[0133] (d44) Debugging the thread coefficient of the original analysis thread using the first data comparison result, the second data comparison result and the third data comparison result to obtain an artificial intelligence analysis thread.

[0134] In a possible implementation example, the above step d44 can be implemented according to the following steps:

[0135] (e1) performing key description content analysis on the fourth feature splicing vector to obtain key description content of the fourth feature splicing vector.

[0136] (e2) Determine the weight coefficient using the key description content of the fourth feature concatenation vector.

[0137] (e3) Determine a target data comparison result using the weight coefficient, the first data comparison result, the second data comparison result and the third data comparison result.

[0138] In the embodiment of the present invention, one or more difference coefficients among the first data comparison result, the second data comparison result and the third data comparison result may be adjusted by a weight coefficient and a target data comparison result may be determined based on the adjusted difference coefficients.

[0139] (e4) Using the target data comparison result, the thread coefficient of the original analysis thread is debugged to obtain the artificial intelligence analysis thread.

[0140] Based on the above embodiments, the beneficial effects of the present invention are as follows: the present invention provides an adaptive risk monitoring data calculation method for mountain objects, which determines the verticality of mountain objects through the key description content of the mountain object feature vector. The larger the key description content, the higher the quality of the mountain object and the larger the risk monitoring data; the smaller the key description content, the worse the quality of the mountain object and the smaller the risk monitoring data. Mountain objects with larger risk monitoring data can be determined as high-quality mountain objects, thereby improving the accuracy of such mountain objects and their published mountain risk data sets.

[0141] On the basis of the above, a high-level landslide risk monitoring device is provided, the device comprising:

[0142] An information acquisition module is used to obtain hidden danger event information and mountain risk data information of a number of to-be-processed mountain risk data sets corresponding to the target mountain object;

[0143] A feature determination module, used to determine, by using an artificial intelligence analysis thread, hidden danger event feature vectors corresponding to hidden danger event information of the plurality of mountain risk data sets to be processed, and mountain risk data set feature vectors corresponding to mountain risk data information of the plurality of mountain risk data sets to be processed;

[0144] A vector splicing module, used for splicing the hidden danger event feature vector and the mountain risk data set feature vector using the artificial intelligence analysis thread to obtain a first feature splicing vector of the target mountain object;

[0145] A data monitoring module is used to determine the risk monitoring data of the target mountain object using the first feature splicing vector, wherein the risk monitoring data is used to represent the sharing coefficient between the types of the plurality of mountain risk data sets to be processed.

[0146] On the basis of the above, a high-level landslide risk monitoring system is shown, which includes a processor and a memory that communicate with each other, and the processor is used to read a computer program from the memory and execute it to implement the above method.

[0147] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0148] In summary, based on the above scheme, the hidden danger event information and mountain risk data information of several to-be-processed mountain risk data sets corresponding to the mountain object are processed by the artificial intelligence analysis thread to obtain the first feature splicing vector corresponding to the mountain object. Since the first feature splicing vector is obtained based on the splicing of the multi-dimensional data of the mountain object, it can better characterize the feature information of the overall content of the mountain object, and ensure the detail of the characterization data of the first feature splicing vector. The present invention calculates the risk monitoring data of the mountain object based on the first feature splicing vector, so that the risk monitoring data can characterize the sharing coefficient between the contents of several to-be-processed mountain risk data sets, thereby improving the accuracy of risk monitoring of the data. In addition, the present invention performs multi-task training of the artificial intelligence analysis thread based on the first regression analysis type of the example mountain object, the second regression analysis type, the target hidden danger event feature vector and the regression analysis hidden danger event feature vector of the hidden danger event information of several example mountain risk data sets, and the first reference type and the second reference type of each example mountain risk data set, which not only improves the thread analysis efficiency, but also accurately determines the high-level landslide risk, thereby improving the monitoring accuracy.

[0149] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or included in processor control codes, such as carrier media such as disks, CDs or DVD-ROMs, programmable memories such as read-only memories (firmware), or data carriers such as optical or electronic signal carriers. Such codes are provided on. The system and its modules of the present invention can be implemented not only by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software such as executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (e.g., firmware).

[0150] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

Claims

1. A method for monitoring high-level landslide risk, characterized in that: The method comprises: Obtaining hidden danger event information and mountain risk data information of several to-be-processed mountain risk data sets corresponding to the target mountain object; Determine, by using an artificial intelligence analysis thread, hidden danger event feature vectors corresponding to hidden danger event information of the plurality of mountain risk data sets to be processed, and mountain risk data set feature vectors corresponding to mountain risk data information of the plurality of mountain risk data sets to be processed; Using the artificial intelligence analysis thread to perform splicing processing on the hidden danger event feature vector and the mountain risk data set feature vector to obtain a first feature splicing vector of the target mountain object; Determining risk monitoring data of the target mountain object by using the first feature splicing vector, wherein the risk monitoring data is used to represent a sharing coefficient between types of the plurality of mountain risk data sets to be processed; Wherein, the artificial intelligence analysis thread is obtained by training the original analysis thread using the regression analysis data and the first reference type and the second reference type of each example mountain risk data set after processing the hidden danger event information and the mountain risk data information of several example mountain risk data sets corresponding to the example mountain object using the original analysis thread to obtain the regression analysis data of the example mountain object, wherein the regression analysis data includes the first regression analysis type and the second regression analysis type of the example mountain object, the target hidden danger event feature vector of the hidden danger event information of the several example mountain risk data sets, and the regression analysis hidden danger event feature vector; the second reference type is a local type of the first reference type, and the second regression analysis type is a local type of the first regression analysis type; Wherein, the determining the risk monitoring data of the target mountain object by using the first feature splicing vector includes: The first feature concatenation vector is processed by using different convolution modules in the first decision unit in the artificial intelligence analysis thread to obtain a plurality of hidden danger description feature vectors; different convolution units exist between two random convolution modules; Determining the difference coefficients between the plurality of hidden danger description feature vectors, and determining the risk monitoring data of the target mountain object according to the difference coefficients; The step of determining the difference coefficients between the plurality of hidden danger description feature vectors and determining the risk monitoring data of the target mountain object according to the difference coefficients includes: Performing function calculation on the plurality of hidden danger description feature vectors to obtain difference coefficients of the plurality of hidden danger description feature vectors; The difference coefficient is subjected to dimensionless simplification processing, and the dimensionless simplified difference coefficient is determined as the risk monitoring data of the target mountain object.

2. The method according to claim 1, characterized in that The step of determining the risk monitoring data of the target mountain object by using the first feature splicing vector includes: Performing key description content analysis on the first feature splicing vector to obtain key description content of the first feature splicing vector; The risk monitoring data of the target mountain object is determined according to the key description content of the first feature stitching vector.

3. The method according to claim 2, characterized in that The method further comprises: Obtaining hidden danger event information, mountain risk data information and marking information of a plurality of example mountain risk data sets corresponding to the example mountain objects; the marking information includes a first reference category and a second reference category of each example mountain risk data set; Determine, by using the original analysis thread, a target hidden danger event feature vector of the hidden danger event information of the plurality of example mountain risk data sets and a mountain risk data set feature vector of the mountain risk data information of the plurality of example mountain risk data sets; Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the several example mountain risk data sets are processed to obtain the regression analysis data of the example mountain object; The original analysis thread is trained using the regression analysis data, the first reference category and the second reference category of each example mountain risk data set to obtain an artificial intelligence analysis thread.

4. The method according to claim 3, characterized in that The method of using the original analysis thread to process the target hidden danger event feature vectors of the hidden danger event information of the plurality of example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the plurality of example mountain risk data sets to obtain the regression analysis data of the example mountain object includes: Using the original analysis thread, the target hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and the mountain risk data set feature vectors of the mountain risk data information of the several example mountain risk data sets are processed to obtain regression analysis hidden danger event feature vectors of the hidden danger event information of the several example mountain risk data sets and splicing feature vectors of the several example mountain risk data sets, wherein the regression analysis hidden danger event feature vector is generated based on the hidden danger hidden feature vector of the target hidden danger event feature vector; Using the original analysis thread, the spliced ​​feature vectors of the several example mountain risk data sets are split and processed to obtain second feature splicing vectors corresponding to the hidden danger event information of the several example mountain risk data sets and third feature splicing vectors corresponding to the mountain risk data information of the several example mountain risk data sets, and determining a fourth feature splicing vector of the example mountain object according to the second feature splicing vector and the third feature splicing vector; Processing the third feature splicing vector using the second decision unit in the original analysis thread to obtain a first regression analysis type of the example mountain object, and processing the fourth feature splicing vector using the first decision unit in the original analysis thread to obtain a second regression analysis type of the example mountain object; The first regression analysis type, the second regression analysis type, the target hidden danger event feature vector, and the regression analysis hidden danger event feature vector are determined as regression analysis data of the example mountain object.

5. The method according to claim 4, characterized in that The method of training the original analysis thread by using the regression analysis data, the first reference type and the second reference type of each example mountain risk data set to obtain an artificial intelligence analysis thread includes: Determining a first data comparison result using the first regression analysis type and a first reference type of each of the example mountain risk data sets; Determining a second data comparison result using the second regression analysis type and a second reference type of each of the example mountain risk data sets; Determine a third data comparison result using the target hidden danger event feature vector and the regression analysis hidden danger event feature vector; The thread coefficient of the original analysis thread is debugged using the first data comparison result, the second data comparison result and the third data comparison result to obtain an artificial intelligence analysis thread.

6. The method according to claim 5, characterized in that The step of debugging the thread coefficient of the original analysis thread by using the first data comparison result, the second data comparison result, and the third data comparison result to obtain an artificial intelligence analysis thread includes: Performing key description content analysis on the fourth feature stitching vector to obtain key description content of the fourth feature stitching vector; Determine a weight coefficient using key description content of the fourth feature splicing vector; Determine a target data comparison result using the weight coefficient, the first data comparison result, the second data comparison result, and the third data comparison result; The target data comparison result is used to debug the thread coefficient of the original analysis thread to obtain an artificial intelligence analysis thread.

7. The method according to claim 2, characterized in that The step of obtaining the hidden danger event information and mountain risk data information of a plurality of to-be-processed mountain risk data sets corresponding to the target mountain object includes: Obtaining several to-be-processed mountain risk data sets corresponding to the target mountain object; Perform hidden danger event information screening processing on each of the mountain risk data sets to be processed, and obtain hidden danger event information of each of the mountain risk data sets to be processed, wherein the hidden danger event information includes one or more of the landslide hidden danger directory information of the mountain risk data set to be processed and the hidden danger event screening information of the landslide data node; Mountain risk data information screening processing is performed on each of the mountain risk data sets to be processed to obtain mountain risk data information of each of the mountain risk data sets to be processed, wherein the mountain risk data information includes landslide risk element information of the landslide data nodes of the mountain risk data sets to be processed.

8. A high-level landslide risk monitoring system, characterized in that: The invention comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 7.

Images