CN112686192B - Landslide stability classification method based on fine terrain features - Google Patents
Landslide stability classification method based on fine terrain features Download PDFInfo
- Publication number
- CN112686192B CN112686192B CN202110011403.6A CN202110011403A CN112686192B CN 112686192 B CN112686192 B CN 112686192B CN 202110011403 A CN202110011403 A CN 202110011403A CN 112686192 B CN112686192 B CN 112686192B
- Authority
- CN
- China
- Prior art keywords
- landslide
- elevation
- area
- stability
- calculating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
Landscapes
- Pit Excavations, Shoring, Fill Or Stabilisation Of Slopes (AREA)
Abstract
The invention discloses a fine terrain feature-based landslide stability classification method, and belongs to the field of geological disaster assessment. The method optimizes the sample characteristics by extracting the topographic characteristics of the landslide sample data, so that the extracted characteristics can better reflect the landslide condition, and the landslide classification result has stronger interpretability. In order to improve the accuracy of landslide stability classification, landslide is better imaged; furthermore, the method for calculating the stability of the landslide by adopting the average sunshine amount in unit time is beneficial to detecting the state of the landslide in real time and calculating the stability of the current landslide more accurately, because the probability of the landslide in the night is much higher than that of the landslide in the day, the probability characteristic is introduced by adopting the method of the average sunshine amount per hour, so that the stability of the landslide at the current moment is calculated more accurately, and the capability of calculating the stability of the landslide is realized. Can be used for disaster prevention and reduction, national resource exploration, engineering construction and the like.
Description
Technical Field
The invention belongs to the field of geological disaster assessment, and particularly relates to a landslide stability classification method based on fine terrain features.
Background
In order to reduce the loss of the geological disaster to the life and property safety of people, a batch of major geological disaster treatment works are carried out in national resource halls of Sichuan province. With the development of landslide control work, a geological disaster control department and a survey design department accumulate a large amount of landslide control engineering data, and the landslide control engineering data are gradually formed. At present, in landslide control engineering, only a part of landslide survey data is used, and the previous landslide control engineering data is not combined.
In the stability evaluation of landslide, the traditional method is to extract influence factors of landslide from a geological disaster field survey table and a landslide survey map. For the extraction of the topographic features, generally, only the slope, elevation interval, elevation difference, slope direction, slope length, slope width, slope morphology and the like of the landslide are considered. These landslide characteristics can only reflect the overall situation of the landslide and cannot finely reflect the local characteristics of the landslide.
Disclosure of Invention
The invention aims to provide a landslide stability classification method based on fine terrain features, which is used for solving the problem of fine extraction of the terrain features in the existing landslide stability research and can quickly obtain the landslide stability condition by means of a constructed classification model. The method solves the problems that in the traditional landslide stability research, landslide terrain conditions are not sufficiently considered, and a stability result can be obtained only in a long time due to excessive dependence on geological survey and analysis calculation.
In order to realize the purpose, the technical scheme of the invention is as follows: a landslide stability classification method based on fine terrain features comprises the following steps:
step 1: acquiring existing landslide range image data;
step 2: generating an irregular triangular grid according to the actually measured elevation point data in the landslide range, and further generating a digital elevation image according to the irregular triangular grid;
and step 3: calculating the gradient, the slope direction and the curvature of the landslide area according to the digital elevation image of the landslide area;
and 4, step 4: extracting detailed characteristics;
step 4.1: calculating the average elevation, the maximum elevation, the minimum elevation, the elevation variance and the elevation difference of the landslide area according to the digital elevation image obtained in the step 2;
step 4.2: calculating the maximum value, the minimum value and the gradient change trend of the gradient according to the gradient data obtained in the step 3;
step 4.3: calculating the sunlight intensity and the sunlight time of a landslide area according to the slope direction data and the landslide position;
step 4.4: calculating the lengthwise length, the transverse length and the shape of the landslide according to the landslide range;
and 5: extracting the stability condition of the landslide as a label corresponding to the landslide range image according to the document of the landslide control project;
step 6: extracted in step 4: taking the average elevation, the maximum elevation, the minimum elevation, the elevation variance and the elevation difference of the elevation, the maximum value, the minimum value and the gradient change trend of the gradient, the sunshine intensity and the sunshine time of a landslide area, the curvature, the lengthwise length, the crosswise length and the shape of the landslide as the input of a classifier, taking the label defined in the step 5 as the output, and training the classifier;
and 7: and during actual classification, collecting images of the landslide area, and then classifying by adopting a trained classifier.
Further, the gradient trend in the step 4.2 is to only intercept 1/a area of the slope surface close to the edge part, wherein a is larger than 2.
Further, the classifier adopted in the step 6 is an XGBoost classifier.
Further, in step 4.3, the sunshine intensity and sunshine duration in the landslide area are replaced by the average sunshine amount in unit time, and the average sunshine amount in unit time is calculated by the following method:
step 4.3.1: setting the geographic coordinates: eight directions of north, northeast, east, southeast, south, southwest, west and northwest, wherein each direction comprises a sector area with 45 degrees;
step 4.3.2: calculating a sector area with the highest landslide area ratio, wherein the corresponding direction of the sector area is the main landslide direction;
step 4.3.3: calculating the average sunshine amount in unit time of the area according to the weather condition of the area from the landslide to the current moment;
step 4.3.3: and calculating the average sunshine amount in the unit time born by the landslide area according to the main slope direction of the landslide, the geographical position of the landslide and the average sunshine amount in the unit time of the area.
Further, the average amount of sunshine is an average amount of sunshine per hour.
The XGboost classifier integrated with learning in the machine learning algorithm is adopted to classify the landslide stability condition, and the method has the advantages of high classification precision and high speed; the invention also optimizes the sample characteristics by extracting the topographic characteristics of the landslide sample data, so that the extracted characteristics can better reflect the landslide condition, and the landslide classification result has stronger interpretability. In order to improve the accuracy of landslide stability classification, landslide is better imaged; furthermore, the method for calculating the stability of the landslide by adopting the average sunshine amount in unit time is beneficial to detecting the state of the landslide in real time and calculating the stability of the current landslide more accurately, because the probability of the landslide in the night is much higher than that of the landslide in the day, the probability characteristic is introduced by adopting the method of the average sunshine amount per hour, so that the stability of the landslide at the current moment is calculated more accurately, and the capability of calculating the stability of the landslide is realized. The invention provides a novel landslide stability classification method based on fine terrain features. Starting from fine images of landslide terrain, the method shows the terrain conditions of all parts of the landslide, extracts features according to the influence of the landslide on the landslide stability, researches and formulates a set of landslide fine terrain feature extraction flow, and constructs a landslide stability classification model. Can be used for disaster prevention and reduction, national soil resource exploration, engineering construction and the like
Drawings
FIG. 1 is a flow chart of a method of landslide stability classification based on fine terrain features of the present invention;
FIG. 2 is an elevation profile of an example of a grade of a central primary school landslide of the eight temple town of the invention in the sunny region of Cinchong in the Bazhong;
FIG. 3 is a graph showing a slope profile of an example of a grade of a central primary school of an eight temple town of the present invention in the sunny region of Cinchun, Bazhong;
FIG. 4 is a slope diagram of an example of a landslide of the central primary school of the eight temple town of the Bazhong city, Enyang, according to the present invention;
FIG. 5 is a graph of curvature distribution of an example of a grade of a central primary school of the eight temple town on the sunny side of the city in the Bazhong of the present invention;
FIG. 6 is a schematic diagram of a landslide range of an example of a landslide of the central primary school of the eight temple town of the Yangyang district in Cizhong.
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
The technical scheme of the invention is illustrated as figure 1, taking the landslide of the central primary school of the eight temple town on the sunny area of city in Ba as an example, and comprises the following steps:
step 1: data processing
Firstly, a plane layout map of a landslide control project of a central primary school of eight temple towns in the Yangyang area in the city in Ba is opened by ArcGIS, and a required elevation point and landslide range map layer are derived. And calculating high-precision elevation Data (DEM) of a landslide area according to the elevation image of the landslide range, wherein the resolution ratio reaches 2 meters. And then inputting the DEM of the landslide area through tools such as gradient, slope direction and curvature in the grid surface, and outputting the data of the gradient, the slope direction and the curvature of the obtained landslide area. And then, utilizing a space analysis tool of ArcGIS, sequentially opening extraction analysis- > according to a mask for extraction, inputting the DEM, the gradient, the slope direction, the curvature and the landslide range of the landslide region, and outputting the DEM, the gradient, the slope direction and the curvature of the cut landslide.
Step 2: feature extraction
After the topographic data of the landslide is obtained, the topographic data is read into a two-dimensional array by a program, and then characteristic extraction is started. For elevation data, the impact characteristics include mean elevation, maximum elevation, minimum elevation, elevation variance, and elevation difference. Relative elevation and height differences may better reflect the stability of the landslide relative to absolute elevation. For the gradient data, not only the maximum value and the minimum value thereof but also the tendency of the gradient to change in the landslide direction is noted. The size and distribution of the slope have a great influence on the stability of the landslide. For example, when the slope of the leading edge of a landslide is large, a blank surface is easily generated, which is very disadvantageous to the stability of the landslide. For the slope data, the distribution of the landslide slope is obtained by performing statistical analysis to obtain histogram features. The slope value is distributed between 0 and 360, and the range of the slope is divided into eight parts by taking 45 degrees as one part: [22.5, 67.5, 112.5, 157.5, 202.5, 247.5, 292.5, 337.5 ]. These eight parts represent eight directions north, northeast, east, southeast, south, southwest, west and northwest. Wherein the slope direction with the highest proportion is the main slope direction of the landslide. The slope direction influences the sunshine intensity and time of the landslide, and further influences the stability of the landslide. For curvature data, the absolute value of curvature and the positive and negative distributions are extracted. The positive or negative of the curvature value may determine whether a particular portion of the landslide surface is convex or concave. This is related to ground water and surface runoff. For a range of landslides, the lengthwise and widthwise lengths and shape of the landslide are determined by row-column scanning. And extracting the topographic features of each landslide according to the feature extraction mode, and storing the extracted features into an array.
Further, the average amount of sunshine, which is the average amount of sunshine per hour, is used instead of the sunshine intensity and time.
And 3, step 3: generating a landslide stability label
And looking up landslide management engineering documents, finding corresponding landslide stability calculation results, comparing the calculated stability coefficient with the safety coefficient, marking the landslide label with the stability coefficient smaller than the safety coefficient as unstable, and marking the landslide label with the stability coefficient larger than the safety coefficient as stable. And storing the stability conditions of all the landslides into an array as a label of stability classification.
And 4, step 4: establishing an XGboost classification model;
for all landslides with the feature extraction and label generation completed, sample data is randomly segmented according to the proportion of 8:2 to generate a training set and a testing set, namely 80% of data is used for training, and 20% of data is reserved for testing. The main parameters related to the XGBoost algorithm include the type of the basic model, the learning rate, the maximum depth of the decision tree, the learning goal, and the like. The learning rate and the maximum depth of the decision tree are the more important parameters in the model, and the specific parameter cases are shown in table 1. And building an XGboost classifier according to the selected parameters, inputting the characteristics and the labels of the training set into the built XGboost classifier, outputting the predicted classification result and accuracy of the training set, and storing the trained model.
TABLE 1 partial hyperparameters involved in the XGboost model
And 5: analyzing the model and evaluating the accuracy;
inputting the characteristics of the test sample to obtain a classification label of the test sample, comparing a prediction label of the test set with an original label of the test set sample, and calculating the prediction accuracy by using accuracy _ score of a metrics packet in a sklern library. And comparing the y _ predicted obtained by importing the test set into the model with the actual y _ test (actual accurate label of the test set) to obtain corresponding accuracy, and adjusting the evaluation model according to the accuracy.
Step 6: and (5) verifying the result.
Seen from a landslide of a central primary school of eight temple towns in a city sunny region in a Bazhong city, the landslide is divided into stable landslides by the XGboost classifier constructed by the invention in a test set, and the landslide is consistent with the original label stability of the landslide. Looking over its topography characteristic map, fig. 2 to 5, can see that its topography characteristic is that northwest is high, the southeast is low, and landslide leading edge and trailing edge slope are comparatively mild, and the middle part slope is steeper, and main slope is the southeast direction, and the curvature value does not have the place of sudden change, and the stability condition is better, accords with the result that the classifier reachd.
The method has the beneficial effect that the landslide stability can be quickly and accurately obtained according to the fine terrain features. The XGboost classifier trained by the method provided by the invention has the accuracy rate of 90% for landslide stability classification, and the result has higher reliability. Can be used as the basis and reference for landslide disaster control and prediction.
Claims (3)
1. A landslide stability classification method based on fine terrain features comprises the following steps:
step 1: acquiring existing landslide range image data;
and 2, step: generating an irregular triangular grid according to the actually measured elevation point data in the landslide range, and further generating a digital elevation image according to the irregular triangular grid;
and step 3: calculating the gradient, the slope direction and the curvature of the landslide area according to the digital elevation image of the landslide area;
and 4, step 4: extracting detailed characteristics;
step 4.1: calculating the average elevation, the maximum elevation, the minimum elevation, the elevation variance and the elevation difference of the landslide area according to the digital elevation image obtained in the step 2;
step 4.2: calculating the maximum value, the minimum value and the gradient change trend of the gradient according to the gradient data obtained in the step 3;
step 4.3: calculating the sunlight intensity and the sunlight time of a landslide area according to the slope direction data and the landslide position;
4.3, replacing the sunshine intensity and sunshine time of the landslide area by the average sunshine amount in unit time, wherein the average sunshine amount in unit time is calculated by the following method:
step 4.3.1: setting reference geographic coordinates: eight directions of north, northeast, east, southeast, south, southwest, west and northwest, wherein each direction comprises a sector area with 45 degrees;
step 4.3.2: calculating a sector area with the highest landslide area ratio, wherein the corresponding direction of the sector area is the main landslide direction;
step 4.3.3: calculating the average sunshine amount in unit time of the area according to the weather condition of the area from the landslide to the current moment;
step 4.3.3: calculating the average sunshine amount in unit time born by a landslide area according to the main slope direction of the landslide, the geographic position of the landslide and the average sunshine amount in unit time of the area;
step 4.4: calculating the lengthwise length, the transverse length and the shape of the landslide according to the landslide range;
and 5: extracting the stability condition of the landslide as a label corresponding to the landslide range image according to the document of the landslide control project;
step 6: extracted in step 4: taking the average elevation, the maximum elevation, the minimum elevation, the elevation variance and the elevation difference of the elevation, the maximum value, the minimum value and the gradient change trend of the gradient, the sunshine intensity and the sunshine time of a landslide area, the curvature, the lengthwise length, the crosswise length and the shape of the landslide as the input of a classifier, taking the label defined in the step 5 as the output, and training the classifier;
the classifier adopted in the step 6 is an XGboost classifier; for all landslides with the feature extraction and label generation completed, sample data is randomly segmented according to the proportion of 8:2 to generate a training set and a testing set, namely 80% of data is used for training, and 20% of data is reserved for testing
And 7: and during actual classification, collecting images of the landslide area, and then classifying by adopting a trained classifier.
2. A method for fine terrain feature-based classification of landslide stability as claimed in claim 1 wherein the trend of slope change in step 4.2 is to intercept only 1/a area of the landslide slope near the edge where a is greater than 2.
3. A method for fine terrain feature-based landslide stability classification as claimed in claim 1 wherein the average amount of insolation is an average amount of insolation per hour.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110011403.6A CN112686192B (en) | 2021-01-06 | 2021-01-06 | Landslide stability classification method based on fine terrain features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110011403.6A CN112686192B (en) | 2021-01-06 | 2021-01-06 | Landslide stability classification method based on fine terrain features |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112686192A CN112686192A (en) | 2021-04-20 |
| CN112686192B true CN112686192B (en) | 2022-05-31 |
Family
ID=75455794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110011403.6A Active CN112686192B (en) | 2021-01-06 | 2021-01-06 | Landslide stability classification method based on fine terrain features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686192B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101191384A (en) * | 2006-11-21 | 2008-06-04 | 钟显亮 | Technique for applying negative energy consumption environment air conditioner to town stock building |
| CN103455711A (en) * | 2013-08-15 | 2013-12-18 | 广州地理研究所 | Small watershed region-oriented landslide hazard risk division method based on mechanism analysis |
| CN110362949A (en) * | 2019-07-23 | 2019-10-22 | 电子科技大学 | A kind of landslide sensitivity assessment method neural network based |
| CN111222466A (en) * | 2020-01-08 | 2020-06-02 | 武汉大学 | Remote sensing image landslide automatic detection method based on three-dimensional space-channel attention mechanism |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
| WO2020151226A1 (en) * | 2019-01-25 | 2020-07-30 | 深圳大学 | Landslide prediction method and device based on training model |
| CN111739259A (en) * | 2020-06-15 | 2020-10-02 | 中国科学院、水利部成都山地灾害与环境研究所 | Slope unit local clustering damage judgment method and regional landslide early warning method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008209992A (en) * | 2007-02-23 | 2008-09-11 | Fuji Xerox Co Ltd | Image processor and program |
-
2021
- 2021-01-06 CN CN202110011403.6A patent/CN112686192B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101191384A (en) * | 2006-11-21 | 2008-06-04 | 钟显亮 | Technique for applying negative energy consumption environment air conditioner to town stock building |
| CN103455711A (en) * | 2013-08-15 | 2013-12-18 | 广州地理研究所 | Small watershed region-oriented landslide hazard risk division method based on mechanism analysis |
| WO2020151226A1 (en) * | 2019-01-25 | 2020-07-30 | 深圳大学 | Landslide prediction method and device based on training model |
| CN110362949A (en) * | 2019-07-23 | 2019-10-22 | 电子科技大学 | A kind of landslide sensitivity assessment method neural network based |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
| CN111222466A (en) * | 2020-01-08 | 2020-06-02 | 武汉大学 | Remote sensing image landslide automatic detection method based on three-dimensional space-channel attention mechanism |
| CN111739259A (en) * | 2020-06-15 | 2020-10-02 | 中国科学院、水利部成都山地灾害与环境研究所 | Slope unit local clustering damage judgment method and regional landslide early warning method |
Non-Patent Citations (5)
| Title |
|---|
| Ground-based multi-view photogrammetry for the monitoring of landslide deformation and erosion;A.Stumpf等;《Geomorphology》;ELSEVIER;20150215;第231卷;第130-145页 * |
| 四川省通江县地质灾害特征及影响因素探讨;梁源等;《中国新技术新产品》;20150210(第03期);第139-140页 * |
| 基于聚类－信息量耦合模型下的广元市;何静等;《测绘与空间地理信息》;20201218;第43卷(第12期);第25-31页 * |
| 贵州省毕节市崩塌、滑坡地质灾害易发性评价研究;翟克礼;《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》;中国学术期刊（光盘版）电子杂志社;20200315(第3期);A011-49 * |
| 黄土—基岩滑坡滑带土特性及形成机理研究 ————以延安二庄科滑坡为例;王建伟;《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》;中国学术期刊（光盘版）电子杂志社;20170215(第2期);A011-135 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112686192A (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103294792B (en) | Based on the polarization SAR terrain classification method of semantic information and polarization decomposing | |
| CN108257142A (en) | Ramp unit extracting method in DEM | |
| CN108985327A (en) | A kind of Terrain Matching self-organizing Optimum Classification method based on factorial analysis | |
| Huang et al. | Differentiating tower karst (fenglin) and cockpit karst (fengcong) using DEM contour, slope, and centroid | |
| Sun et al. | Coastline extraction using remote sensing: A review | |
| CN114036841A (en) | Landslide incidence prediction method and system based on semi-supervised support vector machine model | |
| Pawłuszek et al. | Landslides identification using airborne laser scanning data derived topographic terrain attributes and support vector machine classification | |
| CN115272866A (en) | Mountain landform form classification method based on positive and negative landforms and application | |
| CN114882373A (en) | Multi-feature fusion sandstorm prediction method based on deep neural network | |
| Straumann et al. | Delineation of valleys and valley floors | |
| CN112686192B (en) | Landslide stability classification method based on fine terrain features | |
| Ayana et al. | Land use change analysis using remote sensing and Markov modeling in Fincha watershed, Ethiopia | |
| Carlón Allende et al. | Hydrogeographical regionalisation: An approach for evaluating the effects of land cover change in watersheds. A case study in the Cuitzeo Lake Watershed, Central Mexico | |
| Lin et al. | Accurate and automatic mapping of complex debris‐covered glacier from remote sensing imagery using deep convolutional networks | |
| Arundel et al. | The effect of resolution on terrain feature extraction | |
| Schumann et al. | Evaluating uncertain flood inundation predictions with uncertain remotely sensed water stages | |
| Afshinfar et al. | RS-based assessment: spatial-temporal changes in water basins of Tar and Havir Lakes | |
| Hua et al. | The research of artificial shoreline extraction based on airborne LIDAR data | |
| Shaw et al. | A methodology for preserving channel flow networks and connectivity patterns in large‐scale distributed hydrological models | |
| CN114998719A (en) | Forest fire prediction method based on deep learning and multi-source remote sensing data | |
| BARRINGER et al. | National mapping of landform elements in support of S-Map, a New Zealand soils database | |
| Wang et al. | Classification of landslide stability based on fine topographic features | |
| CN114387501A (en) | Remote sensing intelligent identification method and device for deposition of flood-washing fan | |
| Du et al. | Landslide susceptibility mapping using support vector machine model | |
| Tassetti et al. | Use of remote sensing data and GIS technology for assessment of landslide hazards in Susa valley, Italy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |