CN106372352B - Landslide area detection device and method - Google Patents

Landslide area detection device and method Download PDF

Info

Publication number
CN106372352B
CN106372352B CN201610821272.7A CN201610821272A CN106372352B CN 106372352 B CN106372352 B CN 106372352B CN 201610821272 A CN201610821272 A CN 201610821272A CN 106372352 B CN106372352 B CN 106372352B
Authority
CN
China
Prior art keywords
node
equal
detection
landslide
probability
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
Application number
CN201610821272.7A
Other languages
Chinese (zh)
Other versions
CN106372352A (en
Inventor
陈潇君
朱娜
蔡文红
江晓明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN201610821272.7A priority Critical patent/CN106372352B/en
Publication of CN106372352A publication Critical patent/CN106372352A/en
Application granted granted Critical
Publication of CN106372352B publication Critical patent/CN106372352B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

The invention discloses a landslide area detection device and a landslide area detection method. Screening initial data by using the existing mature data significant correlation detection technology; counting and calculating the landslide occurrence probability by using a machine learning technology; and managing the landslide area in a classified mode by an R-tree technology. The invention realizes regional detection and integrity analysis of the slice region, improves the detection accuracy and can be applied to landslide region detection.

Description

Landslide area detection device and method
Technical Field
The invention belongs to the field of geological disaster detection, and particularly relates to a landslide area detection technology.
Background
Landslide is a serious geological disaster, and if the landslide occurs in a population gathering area, huge economic loss and casualties are caused. Landslide occurs because the shear stress on one of its slip surfaces exceeds the maximum shear stress that the slope can withstand. Detection of landslide is a viable method to reduce the risk of landslide. In the field of landslide detection, the landslide detection device and method are continuously improved from traditional group detection and group defense and manual field detection, to the application of physical detection instruments and professional landslide detectors and to the application of technologies such as remote sensing technology and Internet of things in landslide. Because the landslide is generally a regional event, the detection device and the detection method which are adopted conventionally can obtain and display the sensor raw data of each independent detection point, but the regional detection cannot be carried out on the partitioned area, the integral analysis function is not provided, and the accuracy of the detection is influenced.
Disclosure of Invention
The invention aims to provide a landslide area detection device and a landslide area detection method, so that regional detection and overall analysis can be performed on a partitioned area, and the detection accuracy is improved.
In order to solve the technical problem and realize the function of detecting the landslide area on the whole, the adopted specific technical scheme is as follows:
a landslide area detection apparatus comprising: the system comprises a detection point data acquisition module, a machine learning module, a landslide area management module and a display module;
the detection point data acquisition module is connected with the machine learning module;
the machine learning module is connected with the landslide area management module;
the landslide area management module is connected with the display module;
the detection point data acquisition module is used for acquiring and preprocessing detection data;
the machine learning module is used for training historical data, counting and calculating landslide occurrence probability and realizing a detection function;
the landslide area management module is used for managing landslide areas in a classified mode;
the display module is used for displaying detection point data, a landslide area, a landslide occurrence probability value, a current state of the landslide area and other numerical values.
A landslide area detection method characterized by comprising: a training process and a detection process.
The training process is as follows:
the process S1: detection point data set DP ═ DP1,dp2,...,dpn,...dpNH, wherein the number of the detection points dpnN is more than or equal to 1 and less than or equal to N, N is the total number of detection points of the detection area, each detection point comprises a plurality of sensors, and a sensor data set D is { D ═ at each detection point1,d2,...,ds,...dSS is more than or equal to 1 and less than or equal to S, S is the total number of the nodes of the detection point sensor, and DN is equal to { D ═ D ] of the detection point sensor data set of the detection area1,D2,...,DND for each inspection point sensor datasn(ii) a Initializing s to be 0;
the process S2: s +1, calculating the significant correlation number of the sensor data of the detection pointThe values of p are,
Figure GDA0002298092740000011
the process S3: judging whether the data set is in a set safety threshold range, wherein E is a lower threshold, F is an upper threshold, whether E is more than or equal to rho and less than or equal to F is met, if so, turning to a process S4, and if not, turning to a process S5;
the process S4: discarding D in checkpoint sensor dataset DsGo to process S6;
the process S5: retention of D in checkpoint sensor dataset DsGo to process S6;
the process S6: if S is greater than or equal to 1 and less than S, go to step S2, otherwise go to procedure S7;
the process S7: obtaining a simplified detection point sensor data set DN', wherein the detection point sensor data dmn，1≤m≤M，1≤M≤S；
The process S8: extracting a data training set T from a simplified detection point sensor data set DN', wherein T is { g (x)i,ti)|xi∈DN',tiE.g. DN',1 ≦ i ≦ K }, and data [ x ≦1,x2,...xi...,xK]TFor neuronal input, data [ x ]1,x2,...xi...,xK]TT at the upper right corner represents a training set of data, and the following matrix is also defined according to the rule, and the data [ T1,t2,...ti,...,tK]TFor hidden neuron input, K is the total number of samples;
the process S9: neuron activation function: a ═ g (w)ixi+bi)，biFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]TWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
the process S10: for K training samples (x)i,ti) The entire training matrix is represented as:
Figure GDA0002298092740000021
the process S11: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) is expressed as:
Figure GDA0002298092740000022
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]TThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]T；
The process S12: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data isPij TThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
the process S13: and finishing the training.
The detection process comprises the following steps:
step S1: sensor data set DC ═ D at current time of detection point1,D2,...,DND for each inspection point sensor datamnExtracting C ═ g (x)i,ti)|xi∈DC,tiE.g. DC, i is more than or equal to 1 and less than or equal to K is used as a data detection set, x1,x2,...xi...,xK]CFor input to a neuron; [ t ] of1,t2,...ti,...,tK]CFor hidden neurons, K is the total number of detections;
step S2: neuron activation function: a ═ g (w)ixi+bi) Wherein b isiFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]CConnecting the weight of the neuron input and the weight of the ith hidden layer neuron, wixi represents the inner product of the neuron input and the ith hidden layer neuron, and a is the neuron output;
step S3: for K test samples (x)i,ti) The entire detection matrix is represented as:
Figure GDA0002298092740000031
step S4: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) can be expressed as:
Figure GDA0002298092740000032
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]CThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]C；
Step S5: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000033
Pij CThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
step S6: using a logistic regression model: z is a radical ofj＝a0+a1P1j T+a2P2j T+...,ai-1P(i-1)j T,aiPij C,ai+ 1P(i+1)j T,...+anPnj T
Figure GDA0002298092740000034
zjAs an intermediate variable parameter, a0Is a regression constant, aiCalculating the regression coefficient of j variable, i is more than or equal to 1 and less than or equal to M, j is more than or equal to 1 and less than or equal to N to obtain Pj CThe regression prediction value is the landslide occurrence probability of the area where the jth detection point is located;
step S7: prediction of key landslide area, local area RemAnd M is more than or equal to 1 and less than or equal to M, classifying all detection points of logistic regression layer by adopting a probability R-tree to obtain a local detection point region with high landslide probability, wherein the landslide probability of leaf nodes of the probability R-tree is Pp CP is the number of detection points, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to Com, and Com is the maximum number of layers of the R-tree;
step S8: generating probability R-tree bottom layer leaf Noden1Calculating the probability of landslide
Figure GDA0002298092740000035
Initializing q to 2;
step S9: non-leaf Node for generating probability R-treepqCalculating the spatial range Node of the non-leaf NodeMBRpqCalculating the probability of landslide of non-leaf nodes
Figure GDA0002298092740000041
Step S10: in the display module according to
Figure GDA0002298092740000042
The value of (a) displays different warning colors in the maximum boundary rectangle;
step S11: if q is more than or equal to 1 and less than or equal to Com, go to step S9, otherwise go to step S12;
step S12: the current moment detection process ends.
The expression method of the probability R-tree is as follows: the nodes in the probability R-tree are divided into 2 types: leaf nodes and non-leaf nodes, leaf nodesn1Storing the coordinate positions dp of all detection pointsnLo, probability value P of occurrence of landslide at detection pointn C，dpnThe data items of Lo each correspond to a coordinate position of a spatial coordinate system, where LoRepresenting coordinate positions, comprising 2 data items dpn.Lo.x,dpnY, wherein x represents an x coordinate and y represents a y coordinate; non-leaf NodepqThe storage is the maximum boundary rectangle Node including the space range of all the sensor nodesMBRpqDetection of regional landslide occurrence probability values
Figure GDA0002298092740000043
p is the number of detection points, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to Com, and Com is the maximum number of layers of the R-tree; nodepqLo represents the coordinate position of a non-leaf Node, where Node11X denotes the x coordinate of the coordinate position of a non-leaf Node, Node11Y represents the y coordinate of the coordinate position of the non-leaf node;
NodeMBRpqfor detecting the maximum bounding rectangle of a leaf Node in a region, NodeMBRpqComprises 2 data items NodeMBRpq.st,NodeMBRpqEd, the calculation method is as follows:
NodeMBRp2.st.x＝max(Node11.Lo.x,Node21.Lo.x....Noden1.Lo.x)
NodeMBRp2.ed.x＝min(Node11.Lo.x,Node21.Lo.x....Noden1.Lo.x)
NodeMBRp2.st.y＝max(Node11.Lo.y,Node21.Lo.y....Noden1.Lo.y)
NodeMBRp2.ed.y＝min(Node11.Lo.y,Node21.Lo.y....Noden1.Lo.y)
NodeMBRpq.st.x＝max(NodeMBR1(q-1).st.x,NodeMBR2(q-1).st.x....NodeMBRn(q-1).st.x)
NodeMBRpq.ed.x＝min(NodeMBR1(q-1).ed.x,NodeMBR2(q-1).ed.x....NodeMBRn(q-1).ed.x)
NodeMBRpq.st.y＝max(NodeMBR1(q-1).st.y,NodeMBR2(q-1).st.y....NodeMBRn(q-1).st.y)
NodeMBRpq.ed.y＝min(NodeMBR1(q-1).ed.y,NodeMBR2(q-1).ed.y....NodeMBRn(q-1).ed.y)。
the invention has the beneficial effect. The landslide detection point sensor data collection training method comprises the steps of extracting training samples from landslide detection point sensor data collection, and establishing a landslide model through a training process by adopting a machine learning technology. Through the detection process, the landslide occurrence probability of all detection points is generated, the landslide occurrence probability of a local area is obtained through a probability R-tree, so that a key detection landslide area and a non-key detection landslide area are distinguished, a display module is adopted to mark the key detection landslide area, related workers can conveniently distinguish the severity of landslide, prevention work is done in advance, regional detection and integrity analysis can be carried out on a fragmentation area, and the accuracy of detection is improved.
Drawings
FIG. 1 is a schematic view of the general structure of the apparatus of the present invention.
Description of reference numerals: the system comprises a 1-detection point data acquisition module, a 2-machine learning module, a 3-landslide area management module and a 4-display module.
FIG. 2 is a flow chart of the training method of the present invention.
FIG. 3 is a flow chart of the detection method of the present invention.
Fig. 4 is a real shot diagram of a detection node according to an embodiment of the present invention.
FIG. 5 is a diagram of a detecting node layout interface according to an embodiment of the present invention.
FIG. 6 is a schematic diagram of a distribution of local detection nodes of a probability R-tree according to an embodiment of the present invention.
FIG. 7 is a schematic diagram of generating a spatial range of a probability R-tree according to an embodiment of the present invention.
FIG. 8 is a schematic diagram of probability R-tree generation according to an embodiment of the present invention.
Fig. 9 is a diagram of a landslide area detection interface in accordance with an embodiment of the present invention.
FIG. 10 is a diagram of a second exemplary embodiment of a test node layout interface.
Fig. 11 is a diagram of a landslide area detection interface according to a second embodiment of the present invention.
FIG. 12 is a diagram of a layout interface of a third detecting node according to the embodiment of the present invention.
Fig. 13 is a diagram of a three-landslide area detection interface in accordance with an embodiment of the present invention.
DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
As shown in the general structural diagram of the landslide region detection device shown in fig. 1, the landslide region detection device comprises a 1-detection point data acquisition module, a 2-machine learning module, a 3-landslide region management module and a 4-display module.
The detection point data acquisition module 1 is connected with the machine learning module 2;
the machine learning module 2 is connected with the landslide area management module 3;
the landslide area management module 3 is connected with the display module 4.
In use, the function of the various components of the invention is described as follows:
the detection point data acquisition module 1 is used for acquiring and preprocessing detection data;
the machine learning module 2 is used for training historical data, counting and calculating landslide occurrence probability and realizing a detection function;
the landslide area management module 3 is used for managing landslide areas in a classified mode;
the display module 4 is used for displaying various values such as detection point data, a landslide area, a landslide occurrence probability value, a current state of the landslide area and the like. The training process of a landslide area detection method is shown in fig. 2, the detection process of a landslide area detection method is shown in fig. 3, and the following provides an example of the implementation process of the present invention for different regions.
Example 1: easy-to-slide region in hilly area
The detection node shown in fig. 4 is adopted, the solar power supply plate is arranged on the left side, the closed detection box is arranged on the right side, and the detection node layout interface diagram of the whole area is shown in fig. 5.
The training process is as follows:
the process S1: detection point data set DP of detection region { DP ═ DP1,dp2,...,dpn,...dpNH, wherein the number of the detection points dpnN is more than or equal to 1 and less than or equal to N, N is 50 which is the total number of detection points in the detection area, each detection point comprises a plurality of sensors, and each detection point sensor data set D is { D ═ D1,d2,...,ds,...dSS is not less than 1 and not more than S, S is 8 and is the total number of the nodes of the detection point sensor, a data set of the detection point sensor is selected as { soil density, vegetation coverage, lithology, temperature, humidity, rainfall, displacement and dip }, and a data set DN of the detection point sensor in the detection area is not more than { D ═ D }1,D2,...,DND for each inspection point sensor datasn(ii) a Initializing s to be 0;
the process S2: s +1, calculating the significant correlation value rho of the sensor data of the detection point,
the process S3: judging whether the data set is in a set safety threshold range, wherein E is a lower threshold limit, F is an upper threshold limit, selecting E as 0.02 and F as 0.87, and judging whether E is equal to or less than rho and is equal to or less than F, if so, turning to a process S4, and if not, turning to a process S5;
the process S4: discarding D in checkpoint sensor dataset DsGo to process S6;
the process S5: retention of D in checkpoint sensor dataset DsGo to process S6;
the process S6: if S is greater than or equal to 1 and less than S, go to step S2, otherwise go to procedure S7;
the process S7: obtaining a simplified detection point sensor data set DN', wherein the detection point sensor data dmn，1≤m≤M，1≤M≤S；
The process S8: extracting a data training set T from a detection point sensor data set DN', wherein T is { g (x)i,ti)|xi∈DN',tiE.g. DN',1 ≦ i ≦ K }, numberAccording to [ x ]1,x2,...xi...,xK]TFor neuronal input, data [ t ]1,t2,...ti,...,tK]TThe method is used for hidden layer neuron input, wherein K is the total number of samples, and K is 6;
the process S9: neuron activation function: a ═ g (w)ixi+bi)，biFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]TWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
the process S10: for K training samples (x)i,ti) The entire training matrix is represented as:
Figure GDA0002298092740000062
the process S11: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) is expressed as:
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]TThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]T；
The process S12: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data isPij TFor landslide of ith sensor node in jth detection pointAn occurrence probability value;
the process S13: and finishing the training.
The detection process is as follows:
step S1: sensor data set DC ═ D for each detection point in the detection area at the present time1,D2,...,DND for each inspection point sensor datamnExtracting C ═ g (x)i,ti)|xi∈DC,tiE.g. DC, i is more than or equal to 1 and less than or equal to K is used as a data detection set, x1,x2,...xi...,xK]CFor input to a neuron; [ t ] of1,t2,...ti,...,tK]CThe method is used for hidden layer neurons, K is the total detection number, and K is 6;
step S2: neuron activation function: a ═ g (w)ixi+bi) Wherein b isiFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]CWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
step S3: for K test samples (x)i,ti) The entire detection matrix is represented as:
Figure GDA0002298092740000072
step S4: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) can be expressed as:
Figure GDA0002298092740000073
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]CThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnSingle hidden layer feedforward neural networkIs output vector of [ o1n,o2n,...,omn]C；
Step S5: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000074
Pij CThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
step S6: using a logistic regression model: z is a radical ofj＝a0+a1P1j T+a2P2j T+...,ai-1P(i-1)j T,aiPij C,ai+ 1P(i+1)j T,...+anPnj T
zjAs an intermediate variable parameter, a0Is a regression constant, aiCalculating the regression coefficient of j variable, i is more than or equal to 1 and less than or equal to M, j is more than or equal to 1 and less than or equal to N to obtain Pj CThe regression prediction value is the landslide occurrence probability of the area where the jth detection point is located;
step S7: prediction of key landslide area, local area RemAnd M is more than or equal to 1 and less than or equal to M, classifying all detection points of logistic regression layer by adopting a probability R-tree to obtain a local detection point region with high landslide probability, wherein the landslide probability of leaf nodes of the probability R-tree is Pp CP is a detection point number, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to C, C is the maximum number of layers of the R-tree, and C is 5, as shown in FIG. 6, the probability R tree local detection node distribution;
step S8: generating probability R-tree bottom layer leaf Noden1Calculating the probability of landslide
Figure GDA0002298092740000082
Initializing q to 2;
step S9: non-leaf Node for generating probability R-treepqCalculating the spatial range Node of the non-leaf NodeMBRpqCalculating the probability of landslide of non-leaf nodes
Figure GDA0002298092740000083
FIG. 7 is a schematic diagram of the spatial range of the local detection node in FIG. 6;
step S10: in the display module according to
Figure GDA0002298092740000084
The value of (a) displays different warning colors in the maximum boundary rectangle;
step S11: if q is more than or equal to 1 and less than or equal to C, go to step S9, otherwise go to step S12;
step S12: when the detection process at the current moment is finished, a schematic diagram of generation of a probability R-tree of the local detection node in fig. 6 is shown in fig. 8, and a detection interface diagram of a landslide area in a hilly area is shown in fig. 9.
Example 2: easy-to-slide area of residential area
The map of the test node layout interface for the entire area is shown in FIG. 10.
The training process is as follows:
the process S1: detection point data set DP of detection region { DP ═ DP1,dp2,...,dpn,...dpNH, wherein the number of the detection points dpnN is more than or equal to 1 and less than or equal to N, N is 40 which is the total number of detection points in the detection area, each detection point comprises a plurality of sensors, and each detection point sensor data set D is { D ═ D1,d2,...,ds,...dSS is not less than 1 and not more than S, S is 6 and is the total number of the nodes of the detection point sensor, a data set of the detection point sensor is selected as { soil density, lithology, temperature, humidity, rainfall and displacement }, and a data set DN is not more than { D } of the detection point sensor in the detection area1,D2,...,DND for each inspection point sensor datasn(ii) a Initializing s to be 0;
the process S2: s +1, calculating the significant correlation value rho of the sensor data of the detection point,
Figure GDA0002298092740000085
the process S3: judging whether the data set is in a set safety threshold range, wherein E is a lower threshold limit, F is an upper threshold limit, selecting E as 0.01, F as 0.91, and whether E is equal to or less than rho and equal to F, if so, turning to a process S4, and if not, turning to a process S5;
the process S4: discarding D in checkpoint sensor dataset DsGo to process S6;
the process S5: retention of D in checkpoint sensor dataset DsGo to process S6;
the process S6: if S is greater than or equal to 1 and less than S, go to step S2, otherwise go to procedure S7;
the process S7: obtaining a simplified detection point sensor data set DN', wherein the detection point sensor data dmn，1≤m≤M，1≤M≤S；
The process S8: extracting a data training set T from a detection point sensor data set DN', wherein T is { g (x)i,ti)|xi∈DN',tiE.g. DN',1 ≦ i ≦ K }, and data [ x ≦1,x2,...xi...,xK]TFor neuronal input, data
[t1,t2,...ti,...,tK]TThe method is used for hidden layer neuron input, wherein K is the total number of samples, and K is 4;
the process S9: neuron activation function: a ═ g (w)ixi+bi)，biFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]TWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
the process S10: for K training samples (x)i,ti) The entire training matrix is represented as:
Figure GDA0002298092740000091
the process S11: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) is expressed as:
Figure GDA0002298092740000092
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]TThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]T；
The process S12: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000093
Pij TThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
the process S13: and finishing the training.
The detection process is as follows:
step S1: sensor data set DC ═ D for each detection point in the detection area at the present time1,D2,...,DND for each inspection point sensor datamnExtracting C ═ g (x)i,ti)|xi∈DC,tiE.g. DC, i is more than or equal to 1 and less than or equal to K is used as a data detection set, x1,x2,...xi...,xK]CFor input to a neuron; [ t ] of1,t2,...ti,...,tK]CThe method is used for hidden layer neurons, K is the total detection number, and K is 4;
step S2: neuron activation function: a ═ g (w)ixi+bi) Wherein b isiFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]CWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
step S3: for K test samples (x)i,ti) The entire detection matrix is represented as:
Figure GDA0002298092740000101
step S4: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) can be expressed as:
Figure GDA0002298092740000102
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]CThe output vector of the single hidden layer feedforward neural network is [ o ] when the output vector of the single hidden layer feedforward neural network at the detection point dpn is the output vector of the single hidden layer feedforward neural network1n,o2n,...,omn]C；
Step S5: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000103
Pij CThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
step S6: using a logistic regression model: z is a radical ofj＝a0+a1P1j T+a2P2j T+...,ai-1P(i-1)j T,aiPij C,ai+ 1P(i+1)j T,...+anPnj T
Figure GDA0002298092740000104
zjAs an intermediate variable parameter, a0Is a regression constant, aiCalculating the regression coefficient of j variable, i is more than or equal to 1 and less than or equal to M, j is more than or equal to 1 and less than or equal to N to obtain Pj CThe regression prediction value is the landslide occurrence probability of the area where the jth detection point is located;
step S7: prediction of key landslide area, local area RemAnd M is more than or equal to 1 and less than or equal to M, classifying all detection points of logistic regression layer by adopting a probability R-tree to obtain a local detection point region with high landslide probability, wherein the landslide probability of leaf nodes of the probability R-tree is Pp CP is a detection point number, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to C, C is the maximum number of layers of the R-tree, and C is selected to be 4;
step S8: generating probability R-tree bottom layer leaf Noden1Calculating the probability of landslide
Figure GDA0002298092740000105
Initializing q to 2;
step S9: non-leaf Node for generating probability R-treepqCalculating the spatial range Node of the non-leaf NodeMBRpqCalculating the probability of landslide of non-leaf nodes
Figure GDA0002298092740000111
Step S10: in the display module according to
Figure GDA0002298092740000112
The value of (a) displays different warning colors in the maximum boundary rectangle;
step S11: if q is more than or equal to 1 and less than or equal to C, go to step S9, otherwise go to step S12;
step S12: when the current time detection process is finished, as shown in fig. 11, a detection interface diagram of the landslide area of the residential area is shown;
example 3: area easy to slide along river
The map of the test node layout interface for the entire area is shown in FIG. 12.
The training process is as follows:
the process S1: detection point data set DP of detection region { DP ═ DP1,dp2,...,dpn,...dpNH, wherein the number of the detection points dpnN is more than or equal to 1 and less than or equal to N, N is 45 and is the total number of detection points in the detection area, each detection point comprises a plurality of sensors, and each detection point sensor data set D is { D ═ D1,d2,...,ds,...dSAnd S is more than or equal to 1 and less than or equal to S, the total number of the nodes of the detection point sensors is S-9, the data set of the detection point sensors is selected as { soil density, vegetation coverage rate, distance from a water system, drainage, temperature, humidity, rainfall, displacement and inclination }, and the data set DN of the detection point sensors in the detection area is D-D1,D2,...,DND for each inspection point sensor datasn(ii) a Initializing s to be 0;
the process S2: s +1, calculating the significant correlation value rho of the sensor data of the detection point,
the process S3: judging whether the data set is in a set safety threshold range, wherein E is a lower threshold limit, F is an upper threshold limit, selecting E as 0.04 and F as 0.71, and judging whether E is equal to or less than rho and is equal to or less than F, if so, turning to a process S4, and if not, turning to a process S5;
the process S4: discarding D in checkpoint sensor dataset DsGo to process S6;
the process S5: retention of D in checkpoint sensor dataset DsGo to process S6;
the process S6: if S is greater than or equal to 1 and less than S, go to step S2, otherwise go to procedure S7;
the process S7: obtaining a simplified detection point sensor data set DN', wherein the detection point sensor data dmn，1≤m≤M，1≤M≤S；
The process S8: extracting a data training set T from a detection point sensor data set DN', wherein T is { g (x)i,ti)|xi∈DN',tiE.g. DN',1 ≦ i ≦ K }, and data [ x ≦1,x2,...xi...,xK]TFor neuronal input, data [ t ]1,t2,...ti,...,tK]TThe method is used for hidden layer neuron input, wherein K is the total number of samples, and K is 6;
the process S9: neuron activation function: a ═ g (w)ixi+bi)，biFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]TWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
the process S10: for K training samples (x)i,ti) The entire training matrix is represented as:
Figure GDA0002298092740000121
the process S11: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) is expressed as:
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]TThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]T；
The process S12: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000123
Pij TThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
the process S13: and finishing the training.
The detection process is as follows:
step S1: sensor data set DC ═ D for each detection point in the detection area at the present time1,D2,...,DND for each inspection point sensor datamnExtracting C ═ g (x)i,ti)|xi∈DC,tiE.g. DC, i is more than or equal to 1 and less than or equal to K is used as a data detection set, x1,x2,...xi...,xK]CFor input to a neuron; [ t ] of1,t2,...ti,...,tK]CThe method is used for hidden layer neurons, K is the total detection number, and K is 6;
step S2: neuron activation function: a ═ g (w)ixi+bi) Wherein b isiFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]CWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
step S3: for K test samples (x)i,ti) The entire detection matrix is represented as:
Figure GDA0002298092740000124
step S4: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) can be expressed as:
Figure GDA0002298092740000131
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]CThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]C；
Step S5: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure GDA0002298092740000132
Pij CThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
step S6: using a logistic regression model: z is a radical ofj＝a0+a1P1j T+a2P2j T+...,ai-1P(i-1)j T,aiPij C,ai+ 1P(i+1)j T,...+anPnj T
Figure GDA0002298092740000133
zjAs an intermediate variable parameter, a0Is a regression constant, aiCalculating the regression coefficient of j variable, i is more than or equal to 1 and less than or equal to M, j is more than or equal to 1 and less than or equal to N to obtain Pj CThe regression prediction value is the landslide occurrence probability of the area where the jth detection point is located;
step S7: prediction of key landslide area, local area RemAnd M is more than or equal to 1 and less than or equal to M, classifying all detection points of logistic regression layer by adopting a probability R-tree to obtain a local detection point region with high landslide probability, wherein the landslide probability of leaf nodes of the probability R-tree is Pp CP is a detection point number, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to C, C is the maximum number of layers of the R-tree, and C is selected to be 7;
step S8: generating probability R-tree bottom layer leaf Noden1Calculating the probability of landslide
Figure GDA0002298092740000134
Initializing q to 2;
step S9: non-leaf Node for generating probability R-treepqCalculating is notSpatial extent Node of leaf NodeMBRpqCalculating the probability of landslide of non-leaf nodes
Figure GDA0002298092740000135
Step S10: in the display module according to
Figure GDA0002298092740000136
The value of (a) displays different warning colors in the maximum boundary rectangle;
step S11: if q is more than or equal to 1 and less than or equal to C, go to step S9, otherwise go to step S12;
step S12: and (5) ending the detection process at the current moment, and displaying a detection interface diagram of the landslide area in the river region as shown in fig. 13.

Claims (2)

1. A landslide area detection method, comprising: the method comprises a training process and a detection process;
the training process is as follows:
the process S1: detection point data set DP ═ DP1,dp2,...,dpn,...dpNH, wherein the number of the detection points dpnN is more than or equal to 1 and less than or equal to N, N is the total number of detection points of the detection area, each detection point comprises a plurality of sensors, and a sensor data set D is { D ═ at each detection point1,d2,...,ds,...dSS is more than or equal to 1 and less than or equal to S, S is the total number of the nodes of the detection point sensor, and DN is equal to { D ═ D ] of the detection point sensor data set of the detection area1,D2,...,DND for each inspection point sensor datasn(ii) a Initializing s to be 0;
the process S2: s +1, calculating the significant correlation value rho of the sensor data of the detection point,
Figure FDA0002207856810000011
the process S3: judging whether the data set is in a set safety threshold range, wherein E is a lower threshold, F is an upper threshold, whether E is more than or equal to rho and less than or equal to F is met, if so, turning to a process S4, and if not, turning to a process S5;
the process S4: discarding D in checkpoint sensor dataset DsGo to process S6;
the process S5: retention of D in checkpoint sensor dataset DsGo to process S6;
the process S6: if S is greater than or equal to 1 and less than S, go to step S2, otherwise go to procedure S7;
the process S7: obtaining a simplified detection point sensor data set DN', wherein the detection point sensor data dmn，1≤m≤M，1≤M≤S；
The process S8: extracting a data training set T from a simplified detection point sensor data set DN', wherein T is { g (x)i,ti)|xi∈DN',tiE.g. DN',1 ≦ i ≦ K }, and data [ x ≦1,x2,...xi...,xK]TFor neuronal input, data [ x ]1,x2,...xi...,xK]TT at the upper right corner represents a training set of data, and the following matrix is also defined according to the rule, and the data [ T1,t2,...ti,...,tK]TFor hidden neuron input, K is the total number of samples;
the process S9: neuron activation function: a ═ g (w)ixi+bi)，biFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]TWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
the process S10: for K training samples (x)i,ti) The entire training matrix is represented as:
Figure FDA0002207856810000012
the process S11: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) is expressed as:
Figure FDA0002207856810000013
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]TThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]T；
The process S12: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure FDA0002207856810000021
Pij TThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
the process S13: finishing the training;
the detection process comprises the following steps:
step S1: sensor data set DC ═ D at current time of detection point1,D2,...,DND for each inspection point sensor datamnExtracting C ═ g (x)i,ti)|xi∈DC,tiE.g. DC, i is more than or equal to 1 and less than or equal to K is used as a data detection set, x1,x2,...xi...,xK]CFor input to a neuron; [ t ] of1,t2,...ti,...,tK]CFor hidden neurons, K is the total number of detections;
step S2: neuron activation function: a ═ g (w)ixi+bi) Wherein b isiFor the ith hidden layer neuron tiBias value of [ w ]1,w2,...wi,...wK]CWeight, w, for connecting neuron input and ith hidden neuronixiRepresenting the inner product of the two, a is the neuron output;
step S3: for K test samples (x)i,ti) Whole moment of detectionThe matrix is represented as:
Figure FDA0002207856810000022
step S4: the output of a standard single hidden layer feed forward neural network with K neurons and an activation function g (x) can be expressed as:
Figure FDA0002207856810000023
{β1,β2,...βj...,βKis the weight vector connecting hidden layer neuron and neuron output, [ o ]1,o2,...,om]CThe output vector of the single hidden layer feedforward neural network is used as the detection point dpnThe output vector of the single hidden layer feedforward neural network is [ o ]1n,o2n,...,omn]C；
Step S5: performing logistic regression to calculate the landslide occurrence probability of the detection point, wherein the condition probability of the landslide occurrence in the case of the change of the variable set data is
Figure FDA0002207856810000024
Pij CThe landslide occurrence probability value of the ith sensor node in the jth detection point is obtained;
step S6: using a logistic regression model:
zj＝a0+a1P1j T+a2P2j T+...,ai-1P(i-1)j T,aiPij C,ai+1P(i+1)j T,...+anPnj T
Figure FDA0002207856810000031
zjas an intermediate variable parameter, a0Is a regression constant, aiIs the jth variableThe regression coefficients of i is more than or equal to 1 and less than or equal to M, j is more than or equal to 1 and less than or equal to N, and P is obtained by calculationj CThe regression prediction value is the landslide occurrence probability of the area where the jth detection point is located;
step S7: prediction of key landslide area, local area RemAnd M is more than or equal to 1 and less than or equal to M, classifying all detection points of logistic regression layer by adopting a probability R-tree to obtain a local detection point region with high landslide probability, wherein the landslide probability of leaf nodes of the probability R-tree is Pp CP is the number of detection points, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to Com, and Com is the maximum number of layers of the R-tree;
step S8: generating probability R-tree bottom layer leaf Noden1Calculating the probability of landslide
Figure FDA0002207856810000032
Initializing q to 2;
step S9: non-leaf Node for generating probability R-treepqCalculating the spatial range Node of the non-leaf NodeMBRpqCalculating the probability of landslide of non-leaf nodes
Figure FDA0002207856810000033
q＝q+1；
Step S10: in the display module according to
Figure FDA0002207856810000034
The value of (a) displays different warning colors in the maximum boundary rectangle;
step S11: if q is more than or equal to 1 and less than or equal to Com, go to step S9, otherwise go to step S12;
step S12: the current moment detection process ends.
2. The landslide region detection method of claim 1, wherein said probability R-tree is expressed as follows: the nodes in the probability R-tree are divided into 2 types: leaf nodes and non-leaf nodes, leaf nodesn1Storing the coordinate positions dp of all detection pointsnLo, probability value P of occurrence of landslide at detection pointn C，dpnThe data items of Lo correspond to respective coordinate positions of a spatial coordinate system, where Lo represents a coordinate position and comprises 2 data items dpn.Lo.x,dpnY, wherein x represents an x coordinate and y represents a y coordinate; non-leaf NodepqThe storage is the maximum boundary rectangle Node including the space range of all the sensor nodesMBRpqDetection of regional landslide occurrence probability valuesp is the number of detection points, p is more than or equal to 1 and less than or equal to N, q is the number of layers of a probability R-tree, q is more than or equal to 1 and less than or equal to Com, and Com is the maximum number of layers of the R-tree; nodepqLo represents the coordinate position of a non-leaf Node, where Node11X denotes the x coordinate of the coordinate position of a non-leaf Node, Node11Y represents the y coordinate of the coordinate position of the non-leaf node;
NodeMBRpqfor detecting the maximum bounding rectangle of a leaf Node in a region, NodeMBRpqComprises 2 data items NodeMBRpq.st,NodeMBRpqEd, the calculation method is as follows:
NodeMBRp2.st.x＝max(Node11.Lo.x,Node21.Lo.x....Noden1.Lo.x)
NodeMBRp2.ed.x＝min(Node11.Lo.x,Node21.Lo.x....Noden1.Lo.x)
NodeMBRp2.st.y＝max(Node11.Lo.y,Node21.Lo.y....Noden1.Lo.y)
NodeMBRp2.ed.y＝min(Node11.Lo.y,Node21.Lo.y....Noden1.Lo.y)
NodeMBRpq.st.x＝max(NodeMBR1(q-1).st.x,NodeMBR2(q-1).st.x....NodeMBRn(q-1).st.x)
NodeMBRpq.ed.x＝min(NodeMBR1(q-1).ed.x,NodeMBR2(q-1).ed.x....NodeMBRn(q-1).ed.x)
NodeMBRpq.st.y＝max(NodeMBR1(q-1).st.y,NodeMBR2(q-1).st.y....NodeMBRn(q-1).st.y)
NodeMBRpq.ed.y＝min(NodeMBR1(q-1).ed.y,NodeMBR2(q-1).ed.y....NodeMBRn(q-1).ed.y)。
CN201610821272.7A 2016-09-13 2016-09-13 Landslide area detection device and method Active CN106372352B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610821272.7A CN106372352B (en) 2016-09-13 2016-09-13 Landslide area detection device and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610821272.7A CN106372352B (en) 2016-09-13 2016-09-13 Landslide area detection device and method

Publications (2)

Publication Number Publication Date
CN106372352A CN106372352A (en) 2017-02-01
CN106372352B true CN106372352B (en) 2020-01-24

Family

ID=57897497

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610821272.7A Active CN106372352B (en) 2016-09-13 2016-09-13 Landslide area detection device and method

Country Status (1)

Country Link
CN (1) CN106372352B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111639813A (en) * 2020-06-01 2020-09-08 福州大学 Deep learning-based slag disposal site risk early warning method and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101634459A (en) * 2009-08-24 2010-01-27 陶晓鹏 Thermal power generation boiler intelligent combustion optimizing system and realizing method thereof
CN102542295A (en) * 2012-01-08 2012-07-04 西北工业大学 Method for detecting landslip from remotely sensed image by adopting image classification technology
CN102721370A (en) * 2012-06-18 2012-10-10 南昌航空大学 Real-time mountain landslide monitoring method based on computer vision
CN102819023A (en) * 2012-07-27 2012-12-12 中国地质大学（武汉） Method and system of landslide recognition of complicated geological background area based on LiDAR

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101964135A (en) * 2010-10-13 2011-02-02 南京泰通科技有限公司 Device for monitoring landslide
CN105225046A (en) * 2015-09-30 2016-01-06 武汉工程大学 A kind of Regional Landslide sensitivity assessment data sampling method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101634459A (en) * 2009-08-24 2010-01-27 陶晓鹏 Thermal power generation boiler intelligent combustion optimizing system and realizing method thereof
CN102542295A (en) * 2012-01-08 2012-07-04 西北工业大学 Method for detecting landslip from remotely sensed image by adopting image classification technology
CN102721370A (en) * 2012-06-18 2012-10-10 南昌航空大学 Real-time mountain landslide monitoring method based on computer vision
CN102819023A (en) * 2012-07-27 2012-12-12 中国地质大学（武汉） Method and system of landslide recognition of complicated geological background area based on LiDAR

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
GIS支持下滑坡灾害空间预测方法研究;胡德勇 等;《遥感学报》;20071115;第11卷(第6期);第852-859页 *
Landslide prediction from machine learning;Oliver Korup 等;《Geology Today》;20140228;第30卷(第1期);第26-33页 *

Also Published As

Publication number Publication date
CN106372352A (en) 2017-02-01

Similar Documents

Publication Publication Date Title
Han et al. Convolutional neural network for convective storm nowcasting using 3-D Doppler weather radar data
CN107169426B (en) Crowd emotion abnormality detection and positioning method based on deep neural network
CN107545577B (en) Sedimentary facies image segmentation method based on neural network
CN110119687A (en) Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine
Zhao et al. Automatic recognition of loess landforms using Random Forest method
Etzelmüller et al. Automatic regional classification of topography in Norway.
CN109614973A (en) Rice seedling and Weeds at seedling image, semantic dividing method, system, equipment and medium
CN106408030A (en) SAR image classification method based on middle lamella semantic attribute and convolution neural network
CN118095972B (en) Marine ecological environment prediction evaluation system based on artificial intelligence
CN108596108A (en) Method for detecting change of remote sensing image of taking photo by plane based on the study of triple semantic relation
CN110738388A (en) Method, device, equipment and storage medium for risk conduction of associated map evaluation
Nascimento et al. Automated computational delimitation of SST upwelling areas using fuzzy clustering
CN104634265B (en) A kind of mineral floating froth bed soft measurement method of thickness based on multiplex images Fusion Features
CN112685950A (en) Method, system and equipment for detecting abnormality of ocean time sequence observation data
Hatim et al. Addressing challenges and demands of intelligent seasonal rainfall forecasting using artificial intelligence approach
CN105809119A (en) Sparse low-rank structure based multi-task learning behavior identification method
CN111815075A (en) Prediction method for transportation travel demand under major public health incident
CN106372352B (en) Landslide area detection device and method
CN110363675A (en) A kind of wheat scab monitoring method based on satellite remote sensing
CN117830381A (en) Lake area change space analysis model, construction and area change prediction method and device
Miao et al. Crop weed identification system based on convolutional neural network
CN104899464B (en) A kind of sampling study machine remote sensing quantitative inversion method under adaptation noise conditions
CN117172547A (en) Highway tunnel construction safety evaluation method based on unknown measure theory
Kachwala et al. Predicting Rainfall from Historical Data Trends
CN112380984A (en) Remote sensing-based saline marsh vegetation slow flow capacity space evaluation method

Legal Events

Date Code Title Description
C06 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