A holographic stereoscopic landslide intelligent early warning method based on multiple monitoring occasions

Abstract

The present invention relates to the technical field of early warning methods for low-quality disasters, and in particular to a holographic stereoscopic networking landslide intelligent early warning method based on multiple monitoring occasions. The present invention relates to a holographic stereoscopic networking landslide intelligent early warning method based on multiple monitoring occasions, by carrying out holographic and all-round monitoring of landslides in multiple monitoring fields in the air, sky, ground and underground, and constructing a landslide comprehensive early warning criterion that couples multi-dimensional characteristics such as surface deformation displacement, velocity and acceleration, deep deformation displacement, velocity and acceleration, groundwater level, microseismic intensity, rainfall, soil moisture content, etc., effectively overcoming the limitations of currently common single criteria or double criteria with large errors, low precision and high false alarms, and significantly improving the early warning accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of early warning methods for low-quality disasters, and in particular to an intelligent early warning method for landslides based on a holographic stereoscopic network in multiple monitoring occasions. Background Art

[0002] At present, the main difficulties in early warning of landslide geological disasters are the complexity of the inducing mechanism of landslides, the lack of comprehensive investigation and monitoring big data and analysis methods, etc. The traditional geological disaster early warning model has problems such as limited early warning accuracy, high false alarm, and insufficient refinement. Due to the large number of related factors and complex conditions involved in the landslide system, the disaster-causing mechanism, early warning criteria and thresholds of different types of landslides are very different. Traditional landslide early warning models and methods cannot be adapted to all types of landslides, and can only play an early warning role for landslides of a certain type or a certain evolution stage; most of the early warning criteria and parameters have limitations and cannot accurately and comprehensively reflect the deformation law and disaster mechanism of landslides. Therefore, it is necessary to study the comprehensive early warning criteria of coupled deformation characteristics and disaster-causing characteristics to improve the accuracy of landslide early warning and reduce the false alarm rate. In recent years, with the rapid development of big data and artificial intelligence technology, monitoring technology has made a leap forward, and an integrated air-ground-space monitoring system has been gradually established, making it possible to accurately predict landslide early warnings. At the same time, machine learning technology has gradually been widely used in geological disaster early warning and has achieved good results.

[0003] At present, the integrated monitoring system of air, space, ground and underground mainly relies on remote sensing satellites, UAV remote sensing and the Global Navigation Satellite System (GNSS). As an important spatiotemporal information infrastructure, the global satellite navigation system plays a very important role in the construction of the national economy. On July 31, 2020, my country's BeiDou-3 global satellite navigation system (BeiDou System, BDS) was fully completed and opened, further promoting the depth and breadth of BeiDou / GNSS applications. In addition to basic services such as navigation, positioning and timing, GNSS technology has been widely used in the field of surface deformation monitoring because of its advantages of all-day and all-weather, no need for line of sight between stations, and the ability to obtain high-precision three-dimensional absolute coordinate information. As a point observation system, GNSS technology requires the deployment of stations in the area to be monitored. Due to the limitations of the number and density of receivers, it is impossible to obtain surface deformation information over a large area with high spatial density. In engineering practice, due to the complex environment such as the high mountains and deep valleys of hydropower projects and high-level hidden landslides, only scattered and discontinuous single monitoring points are deployed in key areas, and the deformation information of discrete points is collected by total station, GNSS and other measurement methods according to the method of point-to-surface and local-to-whole for data processing. It is inevitable to miss some major safety hazards, which will cause casualties, economic losses and negative social impacts in serious cases. The advantages of large-scale synchronous observation of remote sensing satellites and UAV remote sensing can effectively make up for some defects of GNSS. By combining the two, the ability to monitor surface deformation can be improved in both space and time.

[0004] With the development of big data and machine learning, combined with the forecast data of meteorological stations over the years (geological environment, precipitation, actual occurrence of disasters), it provides a new development opportunity for the research on landslide disaster early warning models. We can make full use of big data mining and analysis technology, rely on big data such as historical landslide disasters and the geological environment and precipitation that induce landslide disasters, and explore landslide disaster early warning models based on machine learning, which has important theoretical significance and practical application value. Summary of the invention

[0005] In order to solve the problems of the prior art, the present invention provides a holographic stereoscopic networking landslide intelligent early warning method based on multiple monitoring occasions, which can effectively and automatically warn of landslides.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A holographic stereoscopic landslide intelligent early warning method based on multiple monitoring occasions.

[0008] The beneficial effects brought by the technical solution provided by the present invention are:

[0009] (1) The present invention proposes a holographic stereoscopic network landslide intelligent early warning method based on multiple monitoring occasions, conducts holographic and all-round monitoring of landslides in multiple monitoring fields in the air, sky, ground and underground, and constructs a comprehensive landslide early warning criterion that couples multiple characteristics such as surface deformation displacement, velocity and acceleration, deep deformation displacement, velocity and acceleration, groundwater level, microseismic intensity, rainfall, soil moisture content, etc., effectively overcoming the limitations of the currently commonly used single criterion or dual criterion with large errors, low accuracy and high false alarms, and significantly improving the early warning accuracy;

[0010] (2) The present invention combines machine learning with comprehensive landslide early warning criteria, uses machine learning to automatically mine comprehensive landslide early warning criteria, and realizes automatic early warning of landslides, thereby reducing the consumption of manpower and material resources caused by group monitoring and group prevention;

[0011] (3) The present invention constructs an intelligent landslide early warning model based on ensemble learning, which can automatically learn deformation rules and automatically calculate early warning thresholds, reducing the errors caused by currently manually setting thresholds and improving the credibility of the results;

[0012] (4) The present invention adopts a weighted analysis method to effectively integrate and optimize the results of machine learning, thereby ensuring the accuracy of the results;

[0013] (5) The present invention uses weather forecast data to achieve early warning of landslides and ensure the timeliness of the warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 The present invention is a flow chart of a holographic stereoscopic networking landslide intelligent early warning method based on multiple monitoring occasions.

[0016] Figure 2 is the monitoring data used in the embodiment of the present invention, wherein Figure 2(a) is a deformation rate diagram, Figure 2(b) is a tangent angle diagram, Figure 2(c) is a deep displacement diagram, Figure 2(d) is a rainfall diagram, Figure 2(e) is a soil moisture diagram, and Figure 2(f) is a water level gauge diagram.

[0017] Figure 3 It is the intelligent early warning result in the example of the present invention.

[0018] Figure 4(a) is the remote sensing image of the landslide, and Figure 4(b) is the 3D Google image of the landslide.

[0019] Figure 5This is the SBAS-InSAR monitoring result of the landslide.

[0020] Figure 6 This is the landslide classification warning result. DETAILED DESCRIPTION

[0021] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0022] Embodiment 1

[0023] Reference Figure 1 As shown, this embodiment provides a holographic stereoscopic network landslide intelligent early warning method based on multiple monitoring occasions, which specifically includes the following steps:

[0024] S1. Obtain all-round monitoring data from the air, sky, ground and underground, and pre-process the monitoring data based on statistical methods;

[0025] This embodiment adopts high-resolution satellite remote sensing, satellite SAR (synthetic aperture radar), UAV remote sensing, airborne LiDAR (laser detection and ranging system), ship-borne mobile photography, ground professional monitoring equipment and underground professional monitoring equipment in view of the geological background, topography and meteorological conditions of the landslide monitoring area, so as to realize all-round, full-information three-dimensional monitoring of landslide disasters.

[0026] In terms of remote sensing monitoring, high-resolution satellite remote sensing is used to obtain abnormal data on the macroscopic morphology of landslides, satellite SAR is used to obtain surface deformation monitoring data, UAV remote sensing is used to obtain high-precision topographic data of disaster points, and airborne LiDAR is used to obtain landslide morphology and deformation data in high vegetation coverage areas. For steep terrain areas such as the Three Gorges Reservoir area, a comprehensive use of methods such as close-up photography by drones and mobile photography on ships is used to obtain remote sensing information on the side of the landslide. In terms of professional monitoring, the GNSS (Global Navigation Satellite System) monitoring integrated machine provides cumulative displacement monitoring and deformation rate monitoring, water level meters provide groundwater level monitoring, rain gauges provide rainfall monitoring, and soil moisture monitoring provided by soil moisture sensors. The data is collected at intervals of 24 hours.

[0027] The preprocessing method adopted in this embodiment is the quartile method, which calculates the monitoring data of multiple fields, obtains segmented data, and determines whether there are abnormal values ​​in the data. Therefore, the present invention can effectively eliminate the errors and abnormal data caused by the inherent noise of professional monitoring instruments through preprocessing, thereby improving the accuracy and reliability of the data.

[0028] S2. Analyze the monitoring data of landslides based on mathematical statistics to obtain the correlation between landslide deformation and various disaster-causing factors, and combine the macroscopic deformation characteristics of landslides obtained from field investigations to determine the disaster-causing mechanism and deformation and failure mode of landslides;

[0029] This embodiment uses SPSS software (statistical products and service solution software) and Origin software (function drawing software) to normalize the monitoring data, eliminate the influence of dimension, perform correlation analysis on the deformation rate of the landslide and various disaster factors (rainfall, groundwater level, soil moisture, etc.), and use bivariate analysis or multivariate analysis in SPSS software to clarify the significant correlation between various disaster factors and the deformation of the landslide.

[0030] Secondly, in the field investigation, the monitoring points with larger deformation and displacement in InSAR monitoring and surface and underground monitoring are combined to clarify the macroscopic deformation characteristics of the landslide (such as cracks, steps, and dislocations), and analyze the cause mechanism of the landslide and the corresponding deformation and failure mode.

[0031] S3. According to the landslide disaster factors and deformation and damage patterns, the driving force of landslide disasters is clarified, and the comprehensive landslide early warning criteria are constructed based on the holographic data of multiple monitoring sites;

[0032] In this embodiment, the driving force of landslide disasters (such as rainfall-type landslides, whose driving force is related to rainfall, groundwater, and irrigation) is first analyzed based on the correlation of landslide disaster factors obtained in step S2 and the macroscopic characteristics of field investigations. The data related to the driving force is used as one of the component features of the comprehensive criterion, and combined with deformation characteristic criteria such as deformation rate, acceleration, inclination angle, deep displacement, etc. of the landslide, a comprehensive landslide early warning criterion that couples deformation characteristics and disaster characteristics is constructed.

[0033] S4. Determine the graded warning standards for landslides by integrating the landslide surface and underground displacement monitoring data and the macroscopic deformation characteristics of landslides;

[0034] The graded warning standard used in this embodiment is mainly based on: (1) the cumulative deformation curve of the landslide, according to which the deformation rate, tangent angle, acceleration, etc. of the landslide are obtained, and [0, 45°], [45°, 80°], [80°, 85°], [85°, 90°] are used as the tangent angle judgment intervals. Combined with the field investigation of the degree of penetration and size of the cracks, they are divided into blue warning, yellow warning, orange warning, and red warning in sequence.

[0035] S5. Construct a landslide intelligent early warning model based on integrated machine learning, establish an integrated learning algorithm that couples support vector machine, random forest, and C5.0 machine learning, automatically learn deformation laws and automatically calculate warning thresholds;

[0036] In this embodiment, the machine learning method is introduced into the landslide early warning. The ensemble learning algorithm can integrate weak learners into a strong learner, which can fully integrate the advantages of each machine learning algorithm, thereby significantly improving the accuracy of the landslide early warning results. In addition, ensemble learning has the characteristics of self-learning, self-organization and self-reasoning, which can support intelligent landslide early warning.

[0037] S5.1. Construct SVM (Support Vector Machine) machine learning method to realize automatic graded warning of initial landslide by inputting comprehensive warning criteria;

[0038] In this embodiment, an SVM machine learning algorithm is constructed, which takes comprehensive warning criteria (rainfall, groundwater level, soil moisture, deformation rate, acceleration, tangent angle, etc.) as input, and four warning levels of blue warning, yellow warning, orange warning, and red warning as output. It automatically learns deformation laws and calculates warning thresholds to achieve automatic graded warning of landslides.

[0039] S5.2. Construct RDF (Random Decision Forests) machine learning method to realize the second landslide automatic graded warning by inputting comprehensive warning criteria;

[0040] In this embodiment, an RDF machine learning algorithm is constructed, which takes comprehensive warning criteria (rainfall, groundwater level, soil moisture, deformation rate, acceleration, tangent angle, etc.) as input, and four warning levels of blue warning, yellow warning, orange warning, and red warning as output. It automatically learns deformation laws and calculates warning thresholds to achieve automatic graded warning of landslides.

[0041] S5.3, construct C5.0 machine learning method, and realize the third landslide automatic graded warning by inputting comprehensive warning criteria;

[0042] In this embodiment, a C5.0 machine learning algorithm is constructed, which takes comprehensive warning criteria (rainfall, groundwater level, soil moisture, deformation rate, acceleration, tangent angle, etc.) as input, and four warning levels of blue warning, yellow warning, orange warning, and red warning as output. It automatically learns deformation laws and calculates warning thresholds to achieve automatic graded warning of landslides.

[0043] S5.4. Based on the automatic weighted algorithm, a comprehensive analysis is conducted on the automatic graded warning results obtained by the three machine learning methods to automatically obtain graded warning results with high accuracy and good effects as the final automatic graded warning results of the landslide.

[0044] In this embodiment, weighted calculation is performed on the results obtained by the three machine learning in steps S5.1, S5.2 and step S5.3. The learner with the highest accuracy is assigned the largest weight. In order of accuracy from high to low, the three learners are assigned weights of 0.5, 0.3, and 0.2 respectively according to their accuracy. On this basis, weighted calculation is performed and the result with the highest weight is used as the warning result.

[0045] S6. Based on steps S3, S4 and S5, automatic graded early warning is implemented for landslides, and real-time graded early warning is implemented for landslides through real-time updated monitoring data.

[0046] The monitoring data used in this embodiment is real-time and dynamically updated. By adding the subsequent time data into the existing data set, the dynamic update of the data can be completed. The trained intelligent early warning model can be used to realize dynamic landslide early warning. The rainfall data of the weather forecast is coupled with other multi-field monitoring data to realize early warning of landslides.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A holographic stereoscopic network landslide intelligent early warning method based on multiple monitoring occasions, comprising the following steps: S1. Obtaining all-round monitoring data from the air, sky, ground and underground, and preprocessing the monitoring data based on statistical methods; S2. Analyze the monitoring data based on mathematical statistics to obtain the correlation between landslide deformation and disaster-causing factors, and determine the disaster-causing mechanism and deformation and failure mode of the landslide in combination with the macroscopic deformation characteristics of the landslide obtained through field investigation; S3. According to the landslide disaster factors and deformation and damage patterns, the driving force of landslide disasters is clarified, and the data of multiple monitoring sites are integrated to construct a comprehensive landslide early warning criterion that couples deformation characteristics and disaster characteristics; S4. Establish a graded warning standard for landslides based on the monitoring data of landslide surface and deep displacement and the macroscopic deformation characteristics of landslides; S5. Construct a landslide intelligent early warning model based on integrated machine learning, couple the integrated learning of the machine learning algorithm, and use weather forecast data to achieve early landslide forecast and warning; S6, based on the steps S3, S4 and S5, automatic graded early warning is implemented for the landslide, and real-time graded early warning is implemented for the landslide through the real-time updated monitoring data; The step S5 specifically includes the following steps: C1. Construct an SVM machine learning method to automatically learn the deformation law of landslides and automatically calculate the graded warning thresholds by inputting comprehensive warning criteria, thus realizing automatic graded warning of initial landslides. C2. Construct RDF machine learning method, input comprehensive warning criteria, and realize the second landslide automatic graded warning through self-learning, self-organization and self-reasoning mechanism; C3, construct C5.0 machine learning method, input comprehensive warning criteria, and realize the third landslide automatic graded warning through self-learning and self-reasoning; C4. Based on the voting mechanism and automatic weighting algorithm, a comprehensive analysis is conducted on the automatic graded warning results obtained by the three machine learning methods, and the graded warning results with high accuracy and good effect are automatically obtained as the final automatic graded warning results of the landslide.

2. According to the holographic stereoscopic network landslide intelligent early warning method based on multiple monitoring occasions according to claim 1, in the step S1, the method of pre-processing the monitoring data based on the statistical method specifically comprises: The monitoring data is analyzed by using the quartile method to obtain the positions of three split points for judging whether the monitoring data is reasonable, and the abnormal data is eliminated by comparing the monitoring data with the split point data.

3. According to the method for intelligent landslide early warning based on holographic stereoscopic networking in multiple monitoring occasions according to claim 1, the step S2 specifically comprises the following steps: A1. Normalize the landslide monitoring data to eliminate the dimension effect, use mathematical statistics method to analyze the correlation between the deformation displacement of the landslide and the disaster-causing factors, and clarify the disaster-causing factors of the landslide; A2. Through field investigation of landslides, we can obtain the macroscopic deformation characteristics of landslides, conduct comprehensive analysis with the results of indoor mathematical analysis, and determine the disaster-causing mechanism and deformation and failure mode of landslides.

4. According to the method for intelligent landslide early warning based on holographic stereoscopic networking in multiple monitoring occasions as claimed in claim 1, the step S3 specifically comprises the following steps: B1. Through the disaster-causing mechanism and deformation and damage mode of the landslide in step S2, the corresponding early warning criteria are established. Combined with the commonly used criteria in landslide early warning, the C5.0 algorithm is used to automatically mine the comprehensive landslide early warning criteria, and the comprehensive landslide early warning criteria that couples deformation characteristics and disaster-causing characteristics are constructed.

5. According to the method for intelligent landslide early warning based on holographic stereoscopic networking in multiple monitoring occasions as claimed in claim 1, the step S4 specifically comprises the following steps: By analyzing the comprehensive criteria obtained in step S3, using the deformation displacement, deformation rate, tangent angle, and macroscopic deformation characteristics of the landslide, four graded warning standards of blue warning, yellow warning, orange warning, and red warning are clarified.

6. According to the method for intelligent landslide early warning based on holographic stereoscopic networking in multiple monitoring occasions as claimed in claim 1, the step S6 specifically comprises the following steps: The input comprehensive judgment criteria are updated through real-time updated monitoring data, and the landslide intelligent early warning model constructed in step S5 is used to realize real-time automatic graded landslide early warning. The weather forecast rainfall data is input into the landslide intelligent early warning model to realize early landslide warning.

Images