CN106845498A - With reference to the single width mountain range remote sensing images landslide detection method of elevation - Google Patents
With reference to the single width mountain range remote sensing images landslide detection method of elevation Download PDFInfo
- Publication number
- CN106845498A CN106845498A CN201710043591.4A CN201710043591A CN106845498A CN 106845498 A CN106845498 A CN 106845498A CN 201710043591 A CN201710043591 A CN 201710043591A CN 106845498 A CN106845498 A CN 106845498A
- Authority
- CN
- China
- Prior art keywords
- landslide
- region
- remote sensing
- feature
- mountain range
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
Abstract
The invention discloses a kind of single width mountain range remote sensing images landslide detection method of combination elevation.The method to landslide extracted region feature and is pre-processed first, using machine learning training pattern, detects doubtful landslide region；Then in conjunction with the elevation information figure calculated based on dark principle, the elevation information around analysis suspicious region, based on histogram calculation standard deviation, judges whether the region occurs landslide.The present invention can accurately detect the landslide region in the remote sensing images of single width mountain range.
Description
Technical field
The invention belongs to remote sensing image processing and analysis field, the single width mountain range remote sensing images of particularly a kind of combination elevation
Landslide detection method.
Background technology
Remote sensing images have application widely in civilian and military field, generate huge society and economic effect
Benefit.Estimate that elevation information can apply to the detection of the natural calamities such as landslide, mud-rock flow according to single width remote sensing image.China is one
The multiple country of individual natural calamity, therefore, with reference to elevation single width mountain range remote sensing images landslide detection method in remote sensing
There is important application value in image procossing.
The purpose of remote sensing images research is exactly that, for practical application, it can turn into a kind of natural calamity and investigate and monitoring hand
Section.In recent years, with the continuous improvement continued to develop with remote sensing images resolution ratio of satellite, geological disaster researcher starts to close
The practical value of remote sensing images is noted.Until the phase at the end of the nineties in last century, stereoscope aerial photograph interpretation is still landslide system
Figure and monitoring Landslide Features and influence factor most common method, Remote Sensing For Landslides interpretation mainly by the gray scale of remote sensing images,
The features such as texture, tone, landform, landforms recognize landslide.In recent years, Photoshop, ArcGIS, Coreldraw are started with
Man-machine interaction mode interpretation is carried out with reference to DEM, this interpretation means can strengthen image Deng on software platform, and after being superimposed DEM
Certain Spatial Characteristics of Landslide is more directly perceived.But these methods are required for extra auxiliary information, and true without analysis this area
Real relative elevation information is come simple and reliable.
From the foregoing, it will be observed that in the prior art on the premise of unknown dem data, still can not exactly detect that single width mountain range is distant
Landslide region in sense image.
The content of the invention
It is an object of the invention to provide a kind of single width mountain range remote sensing images landslide detection method of combination elevation,
It is capable of detecting when the landslide region in the remote sensing images of single width mountain range.
The technical solution for realizing the object of the invention is：A kind of single width mountain range remote sensing images landslide debris of combination elevation
Stream detection method, comprises the following steps：
Step 1, the feature for determining landslide region simultaneously introduce slow signature analysis function and are pre-processed, and input is supported
Vector machine training obtains model；
Step 2, the model using output in step 1, are processed the feature that mountain range remote sensing images are extracted, and obtain mountain range
The doubtful landslide region template of remote sensing images；
Step 3, with reference to the elevation information figure obtained based on dark principle and Shadows Processing method, determine to doubt in step 2
Like the periphery in landslide region；
The height value of outer peripheral areas in step 4, extraction step 3, based on histogram calculation standard deviation, judges according to threshold value T
Landslide region.
Compared with prior art, its remarkable advantage is the present invention：1) present invention can relatively accurately detect that single width is distant
Landslide region in sense mountain range image；2) method of the present invention, can be direct by image it can be readily appreciated that simple to operate
Obtain the landslide region in image；3) degree of accuracy of method of the present invention operation result is high, and be true to life conjunction.
The present invention is described in further detail below in conjunction with the accompanying drawings.
Brief description of the drawings
Fig. 1 is the schematic diagram of doubtful mud-rock flow region detection in the present invention, wherein figure (a) is the remote sensing figure with mud-rock flow,
Figure (b) is the doubtful mud-rock flow region template for detecting.
Fig. 2 is the schematic diagram of doubtful landslide area periphery and peripheral elevation Distribution value in the present invention, wherein Fig. 2
A (), (c), (e) are the peripheries of each connected region, Fig. 2 (b), (d), (f) are area periphery elevation value histograms.
Fig. 3 is the schematic diagram of the detection of Zhouqu County's mudstone stream picture one in the present invention, and wherein Fig. 3 (a) is the remote sensing with mud-rock flow
Figure, Fig. 3 (b) is the doubtful landslide region template for detecting, Fig. 3 (c) is the elevation information figure extracted from original image,
Fig. 3 (d) is the artificial landslide region demarcated, and Fig. 3 (e) is the landslide region that the present invention is detected, Fig. 3 (f)
It is the error between the result and GT of inventive algorithm (Dark grey represents flase drop region, and light gray represents missing inspection region).
Fig. 4 is the flow chart of the single width mountain range remote sensing images landslide detection method that the present invention combines elevation.
Specific embodiment
A kind of single width mountain range remote sensing images landslide detection method of combination elevation of the invention, including following step
Suddenly：
Step 1, the feature for determining landslide region simultaneously introduce slow signature analysis function and are pre-processed, and input is supported
Vector machine training obtains model；Specially：
Step 1-1, landslide region is analyzed, extracts the feature in these regions：RGB color feature, image
The related contrast of textural characteristics, correlation, energy and homogeney；
Step 1-2, the contrast, correlation, the energy that introduce the above-mentioned image texture characteristic correlation of slow signature analysis function pair
Pre-processed with homogeney, contrast, correlation, energy and homogeney are specifically input into slow feature point with column vector form
Analysis function, the first column data in output is discrimination feature higher；
Step 1-3, the discrimination that will be obtained in RGB color feature and step 1-2 feature input SVMs higher
Training obtains model.
Step 2, the model using output in step 1, are processed the feature that mountain range remote sensing images are extracted, and obtain mountain range
The doubtful landslide region template of remote sensing images；Specially：
Step 2-1, the characteristic that input mountain range remote sensing images are extracted using the method in step 1 are simultaneously pre-processed, and are passed through
The point of doubtful landslide in the model process decision chart picture that step 1 is obtained；
Step 2-2, the region by area S less than area threshold M exclude, areas of the Retention area S more than or equal to area threshold M
Domain；
Step 2-3, the space that doubtful landslide intra-zone is removed using closing operation of mathematical morphology, obtain mountain range remote sensing
The region of doubtful landslide in image.
Step 3, with reference to the elevation information figure obtained based on dark principle and Shadows Processing method, determine to doubt in step 2
Like the periphery in landslide region；Specially：
Step 3-1, each piece of doubtful landslide region outwards expanded into N number of pixel respectively；
Step 3-2, the image before and after expansion is made the difference, obtain each piece of periphery in doubtful landslide region.
It is existing skill with reference to the elevation information figure obtained based on dark principle and Shadows Processing method in above-mentioned steps
Art, in patent《Single width mountain range remote sensing images height value extracting method based on dark principle》(the patent No.：
201510534980.8) disclosed in.
The height value of outer peripheral areas in step 4, extraction step 3, based on histogram calculation standard deviation, judges according to threshold value T
Landslide region.Specially：
The height value of step 4-1, the doubtful landslide area periphery of extraction, and calculate the standard deviation of height value；
Step 4-2, judged, then judgement of the standard deviation more than threshold value T is landslide region, otherwise by the region
Exclude.
Wherein, the value of area threshold M is 200.The value of number of pixels N is 3.The value of standard deviation threshold method T is 0.11.
The present invention can relatively accurately detect the landslide region in the image of single width remote sensing mountain range；It is of the invention
Method can directly obtain the landslide region in image by image it can be readily appreciated that simple to operate.
Further detailed description is done to the present invention with reference to embodiment：
Embodiment
The present invention combines the single width mountain range remote sensing images landslide detection method of elevation, and step is as follows：
The first step, is trained by machine learning and obtains model.Step is as follows：Divide by typical landslide mud-rock flow region
Analysis, determines two big key characters：RGB color feature, image texture characteristic (specially contrast, correlation, energy and homogeneity
Property)；Then introduce slow signature analysis, for pretreatment image textural characteristics, by contrast, correlation, energy and homogeney with
Column vector form is input into slow signature analysis function, and the first column data in output is discrimination feature higher；Finally by RGB face
Color characteristic and discrimination characteristic input SVMs training higher obtain model.
Second step, detects doubtful landslide region.It is implemented as follows：Extract the feature of input mountain range remote sensing images
Data are simultaneously pre-processed, by the doubtful landslide debris flow point in the model process decision chart picture that is obtained in the first step；It is being judged to landslide
In the region of mud-rock flow, there are some smaller and sparse parts, according to priori, it is known that they are not landslide debris
Stream region.Therefore, according to experimental result, a threshold value 200 is set to its area S, if S ＞ 200, retains this doubtful landslide mud
Rock glacier region；There is a defect in the doubtful landslide region for retaining, and inside has gap, using closing operation of mathematical morphology,
Post-etching is first expanded, shaded interior space is eliminated as much as, relatively optimal doubtful landslide region template is obtained, such as schemed
Shown in 1 (b).
3rd step, determines the periphery in doubtful landslide region.The doubtful landslide mud obtained in second step is extracted first
Each connected region in rock glacier region template.Because our information to area periphery are interested, therefore can be by expansion
Obtain the periphery in each region.It should be noted that outer peripheral areas are not the bigger the better, because distance is more remote, outermost regions
Elevation very likely changes, then can not be used for the elevation in approximate doubtful landslide region.Compared by experiment effect,
Present invention determine that outwards expanding 3 pixels, expansion effect such as Fig. 2 (a), (c), (e) are shown.
4th step, determines landslide region.The height value corresponding to outer peripheral areas is taken out, and counts these height values
Distribution, make histogram, such as Fig. 2 (b), (d), shown in (f).Judge what landslide occurs one big foundation was detected
Elevation has a downward trend from high to low around doubtful landslide region, and this shows as data distribution in histogram
More dispersed, the dispersion degree can be described with standard deviation.The downward trend of elevation is estimated with standard deviation, standard deviation is got over
Greatly, height value is more discrete, then substantially, judgement occurs landslide herein for downward trend；Standard deviation is smaller, and height value gets over collection
In, then downward trend is weaker, and judgement does not occur landslide herein.
Fig. 2 is the schematic diagram of a doubtful landslide area periphery and peripheral elevation Distribution value, and the first row represents company
The peripheral and correspondence elevation value histogram in logical region one, the second row represents connected region two, and the third line is connected region three.Its
In, the peripheral height value histogram distribution of connected region three is disperseed the most, and standard deviation is 0.168, and connected region one is taken second place, standard
Difference is 0.1306, and connected region two is most concentrated, and standard deviation is 0.1031.Therefore judge that connected region one, three occurs landslide debris
Stream, not there is landslide in region two.
Fig. 3 is one group of experimental result of quantitative analysis.Fig. 3 (a) is the remote sensing figure with mud-rock flow, and Fig. 3 (b) is detected
Doubtful landslide region template, Fig. 3 (c) is the elevation information figure extracted from original image, and Fig. 3 (d) is artificial demarcation
Landslide region, Fig. 3 (e) is the landslide region that the present invention is detected, Fig. 3 (f) is the result of inventive algorithm
Error between GT (Dark grey represents flase drop region, and light gray represents missing inspection region).Observation Fig. 3 is obtained, and is examined in major part
Survey it is correct in the case of, flase drop and missing inspection appear at greatly the edge in landslide region, and this and GT are by manually demarcating
In the presence of certain relation.
From the foregoing, it will be observed that the present invention is by determining machine learning training pattern, the doubtful landslide in detection mountain range remote sensing images
Mud-rock flow region, in conjunction with the elevation information figure calculated based on dark principle, analyzes the elevation information around suspicious region,
Based on histogram calculation standard deviation, the landslide area in the remote sensing images of single width mountain range finally can be relatively accurately detected
Domain.
Claims (8)
1. the single width mountain range remote sensing images landslide detection method of a kind of combination elevation, it is characterised in that including following step
Suddenly：
Step 1, the feature for determining landslide region simultaneously introduce slow signature analysis function and are pre-processed, and are input into supporting vector
Machine training obtains model；
Step 2, the model using output in step 1, are processed the feature that mountain range remote sensing images are extracted, and obtain mountain range remote sensing
The doubtful landslide region template of image；
Step 3, with reference to the elevation information figure obtained based on dark principle and Shadows Processing method, determine doubtful cunning in step 2
The periphery in slope mud-rock flow region；
The height value of outer peripheral areas in step 4, extraction step 3, based on histogram calculation standard deviation, judges to come down according to threshold value T
Mud-rock flow region.
2. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 1, its feature
It is that step 1 determines the feature in landslide region and introduce slow signature analysis function to be pre-processed, and is input into supporting vector
Machine training obtains model and is specially：
Step 1-1, landslide region is analyzed, extracts the feature in these regions：RGB color feature, image texture
The related contrast of feature, correlation, energy and homogeney；
Step 1-2, introduce the related contrast of the above-mentioned image texture characteristic of slow signature analysis function pair, correlation, energy and same
Matter is pre-processed, and contrast, correlation, energy and homogeney specifically are input into slow signature analysis letter with column vector form
Number, the first column data in output is discrimination feature higher；
The feature input SVMs training higher of step 1-3, the discrimination that will be obtained in RGB color feature and step 1-2
Obtain model.
3. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 1, its feature
It is that step 2 is processed the feature that mountain range remote sensing images are extracted, obtains the doubtful landslide area of mountain range remote sensing images
Domain template, specially：
Step 2-1, the characteristic that input mountain range remote sensing images are extracted using the method in step 1 are simultaneously pre-processed, by step 1
The point of doubtful landslide in the model process decision chart picture for obtaining；
Step 2-2, the region by area S less than area threshold M exclude, regions of the Retention area S more than or equal to area threshold M；
Step 2-3, the space that doubtful landslide intra-zone is removed using closing operation of mathematical morphology, obtain mountain range remote sensing images
In doubtful landslide region.
4. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 1, its feature
It is that the periphery that doubtful landslide region is determined in step 3 is specially：
Step 3-1, each piece of doubtful landslide region outwards expanded into N number of pixel respectively；
Step 3-2, the image before and after expansion is made the difference, obtain each piece of periphery in doubtful landslide region.
5. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 1, its feature
It is to determine that landslide region is specially in step 4：
The height value of step 4-1, the doubtful landslide area periphery of extraction, and calculate the standard deviation of height value；
Step 4-2, judged, then judgement of the standard deviation more than threshold value T is landslide region, otherwise arranges the region
Remove.
6. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 3, its feature
It is that the value of area threshold M is 200 in step 2-2.
7. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 4, its feature
It is that the value of number of pixels N is 3 in step 3-1.
8. the single width mountain range remote sensing images landslide detection method of combination elevation according to claim 5, its feature
It is that the value of step 4-2 Plays difference limen values T is 0.11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710043591.4A CN106845498A (en) | 2017-01-19 | 2017-01-19 | With reference to the single width mountain range remote sensing images landslide detection method of elevation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710043591.4A CN106845498A (en) | 2017-01-19 | 2017-01-19 | With reference to the single width mountain range remote sensing images landslide detection method of elevation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106845498A true CN106845498A (en) | 2017-06-13 |
Family
ID=59119910
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710043591.4A Pending CN106845498A (en) | 2017-01-19 | 2017-01-19 | With reference to the single width mountain range remote sensing images landslide detection method of elevation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106845498A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635726A (en) * | 2018-12-11 | 2019-04-16 | 陕西科技大学 | A kind of landslide identification method based on the symmetrical multiple dimensioned pond of depth network integration |
| CN112418363A (en) * | 2021-01-25 | 2021-02-26 | 中国地质大学（武汉） | Complex background region landslide classification model establishing and identifying method and device |
| CN114049565A (en) * | 2021-11-08 | 2022-02-15 | 中国公路工程咨询集团有限公司 | Geological disaster identification method and device based on remote sensing image and DEM data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632155A (en) * | 2013-12-16 | 2014-03-12 | 武汉大学 | Remote-sensing image variation detecting method based on slow characteristic analysis |
| CN105096275A (en) * | 2015-08-27 | 2015-11-25 | 南京理工大学 | Method for extracting elevation value of single mountain remote sensing images on basis of dark channel principle |
-
2017
- 2017-01-19 CN CN201710043591.4A patent/CN106845498A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632155A (en) * | 2013-12-16 | 2014-03-12 | 武汉大学 | Remote-sensing image variation detecting method based on slow characteristic analysis |
| CN105096275A (en) * | 2015-08-27 | 2015-11-25 | 南京理工大学 | Method for extracting elevation value of single mountain remote sensing images on basis of dark channel principle |
Non-Patent Citations (5)
| Title |
|---|
| 孙洪雨: "基于数学形态学的图像增强与边缘检测方法的研究", 《中国优秀硕士学位论文全文数据库》 * |
| 徐俊峰: "多特征融合的遥感影像变化检测技术研究", 《中国优秀硕士学位论文全文数据库》 * |
| 许高程, 张文君，王卫红: "支持向量机技术在遥感影像滑坡体提取中的应用", 《安徽农业科学》 * |
| 谢飞，杨树文，李轶鲲，刘涛: "基于SPOT5图像的泥石流自动提取方法", 《国土资源遥感》 * |
| 鄢圣藜，霍宏，方涛: "基于SFA和GLCM的影像特征提取方法", 《计算机工程》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635726A (en) * | 2018-12-11 | 2019-04-16 | 陕西科技大学 | A kind of landslide identification method based on the symmetrical multiple dimensioned pond of depth network integration |
| CN109635726B (en) * | 2018-12-11 | 2023-03-24 | 陕西科技大学 | Landslide identification method based on combination of symmetric deep network and multi-scale pooling |
| CN112418363A (en) * | 2021-01-25 | 2021-02-26 | 中国地质大学（武汉） | Complex background region landslide classification model establishing and identifying method and device |
| CN112418363B (en) * | 2021-01-25 | 2021-05-04 | 中国地质大学（武汉） | Complex background region landslide classification model establishing and identifying method and device |
| CN114049565A (en) * | 2021-11-08 | 2022-02-15 | 中国公路工程咨询集团有限公司 | Geological disaster identification method and device based on remote sensing image and DEM data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111080629B (en) | Method for detecting image splicing tampering | |
| CN110929607B (en) | Remote sensing identification method and system for urban building construction progress | |
| CN103150549B (en) | A kind of road tunnel fire detection method based on the early stage motion feature of smog | |
| CN109447068A (en) | A method of it separating seal from image and calibrates seal | |
| CN106951898B (en) | Vehicle candidate area recommendation method and system and electronic equipment | |
| WO2018076138A1 (en) | Target detection method and apparatus based on large-scale high-resolution hyper-spectral image | |
| CN111611861B (en) | Image change detection method based on multi-scale feature association | |
| CN110826429A (en) | Scenic spot video-based method and system for automatically monitoring travel emergency | |
| CN105279772A (en) | Trackability distinguishing method of infrared sequence image | |
| Zhao et al. | Tampered region detection of inpainting JPEG images | |
| CN104966054A (en) | Weak and small object detection method in visible image of unmanned plane | |
| CN106845498A (en) | With reference to the single width mountain range remote sensing images landslide detection method of elevation | |
| Janalipour et al. | A novel and automatic framework for producing building damage map using post-event LiDAR data | |
| Kumar et al. | Comparative analysis for edge detection techniques | |
| Adu-Gyamfi et al. | Functional evaluation of pavement condition using a complete vision system | |
| Gong et al. | Earthquake building damage detection with object-oriented change detection | |
| Manandhar et al. | Segmentation based building detection in high resolution satellite images | |
| CN116543308A (en) | Landslide detection early warning model and early warning method based on multi-model fusion | |
| Liu et al. | An automatic high confidence sets selection strategy for SAR images change detection | |
| Gecen et al. | Road detection from high and low resolution satellite images | |
| Si-ming et al. | Moving shadow detection based on Susan algorithm | |
| Zhan et al. | An image splicing detection method based on PCA minimum eigenvalues. | |
| Yuan et al. | Graph neural network based multi-feature fusion for building change detection | |
| CN112307873A (en) | Automatic illegal building identification method based on full convolution neural network | |
| Abraham et al. | A fuzzy based automatic bridge detection technique for satellite images |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |
|
| RJ01 | Rejection of invention patent application after publication |