CN110109995A - A kind of multi-source heterogeneous data fusion method of fully-mechanized mining working based on deep learning - Google Patents
A kind of multi-source heterogeneous data fusion method of fully-mechanized mining working based on deep learning Download PDFInfo
- Publication number
- CN110109995A CN110109995A CN201910397221.XA CN201910397221A CN110109995A CN 110109995 A CN110109995 A CN 110109995A CN 201910397221 A CN201910397221 A CN 201910397221A CN 110109995 A CN110109995 A CN 110109995A
- Authority
- CN
- China
- Prior art keywords
- source heterogeneous
- fully
- deep learning
- heterogeneous data
- mechanized mining
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of multi-source heterogeneous data fusion methods of the fully-mechanized mining working based on deep learning, are suitable for intelligent coal mine production technique field.According to fully-mechanized mining working in recovery process multi-source heterogeneous data information, establish the deep learning Fusion Model of the multi-source heterogeneous data of fully-mechanized mining working: establishing the multi-source heterogeneous data model of fully-mechanized mining working;Unitized processing is carried out to the space reference, data attribute and element relationship of multi-source heterogeneous data model;Generate the irregular triangle network of the multi-source heterogeneous data model of fully-mechanized mining working;The characteristic parameter of irregular triangle network is extracted, whether verification meets the iterated conditional of deep learning algorithm, establish the deep learning Fusion Model of the multi-source heterogeneous data of fully-mechanized mining working.Energy rapid fusion fully-mechanized mining working of the present invention various mode isomeric datas in recovery process, and effectively promote the efficiency and accuracy of data fusion.
Description
Technical field
The present invention relates to a kind of multi-source heterogeneous data fusion methods of fully-mechanized mining working, are particularly suitable for intelligent coal mine exploitation
The multi-source heterogeneous data fusion method of the fully-mechanized mining working based on deep learning that technical field uses.
Background technique
Intelligent coal mine exploitation is the cutting edge technology that international field of coal mining is pursued jointly, is to reduce casualties, protect
The important means of barrier safety in production is also that China's coal-mine realizes safe and efficient, lasting exploit effective way.In order to realize coal mine
Intelligentized mining needs to acquire a large amount of multi-source heterogeneous data in fully-mechanized mining working recovery process, such as slot wave detection data, laser
Actual measurement creation data of scanner data, the mine ground in measured data, Coherent Noise in GPR Record, 3D seismic data and recovery process
Deng.Due to the approach of these data acquisitions, time and use sensor different, in space reference, data attribute and element
There is skimble-scamble problems for relationship etc..And existing multi-source heterogeneous data processing method mostly uses at " fighting separately "
Reason mode is established the database of respective sensor measurement respectively, and is not associated between each database, causing cannot be accurately anti-
Reflect the corresponding relationship in recovery process between manipulating object and mining environment.
Summary of the invention
Technical problem: shortcoming in view of the above technology, it is simple to provide a kind of step, and execution efficiency is high, can be quickly
It is efficient to provide accurately multi-source heterogeneous data fusion information, while labor-saving fusion fully mechanized mining work for intelligent coal mine exploitation
Make the method for face various mode isomeric datas in recovery process
Technical solution: to achieve the goals above, the multi-source heterogeneous number of the fully-mechanized mining working of the invention based on deep learning
According to fusion method, according to fully-mechanized mining working in recovery process multi-source heterogeneous data information, it is multi-source heterogeneous to establish fully-mechanized mining working
The deep learning Fusion Model of data, specific steps include:
According to fully-mechanized mining working in recovery process multi-source heterogeneous data information, establish the multi-source heterogeneous data of fully-mechanized mining working
Deep learning Fusion Model;Specific steps include:
A acquires multi-source heterogeneous data information in fully-mechanized mining working recovery process, comprising: working face slot wave detection data,
Laser scanner data, mine ground measured data, Coherent Noise in GPR Record, 3D seismic data, and to above-mentioned all multi-source heterogeneous numbers
It is believed that breath optimizes processing, comprising: carry out smothing filtering, three using action difference method removal noise spot, mean filter method
Secondary spline interpolation carries out data modification, establishes the multi-source heterogeneous data model of fully-mechanized mining working;
It is more that b using seven parameter Coordinate Transformation Models of Mo Luojinsiji converts fully-mechanized mining working for multi-source heterogeneous data information
Source Feature Between Heterogeneous Data Model, and by space reference, data attribute and the element relationship in the multi-source heterogeneous data model of fully-mechanized mining working
Carry out unitized processing;
C generates the irregular triangle network of the multi-source heterogeneous data model of fully-mechanized mining working using incremental algorithm;
D extracts the characteristic parameter of irregular triangle network, the number of plies, number of nodes and training algebra including the triangulation network, verification three
The absolute value linear convergence of the number of plies, number of nodes and training algebra that angle is netted is to judge whether the iteration for meeting deep learning algorithm
Condition is such as unsatisfactory for iterated conditional, by adjusting the triangulation network the number of plies is how many, the size of quantity of node and training algebra come it is real
The existing multi-source heterogeneous data model of fully-mechanized mining working meets the iterated conditional requirement of deep learning algorithm;
E such as meets the iterated conditional of deep learning algorithm, then carries out feature training using deep learning training sample, later
The multi-source heterogeneous data model performance of fully-mechanized mining working is verified using deep learning verifying sample,
F completes feature training in deep learning training sample in step e and deep learning verifying sample completes model
Performance verification merges mould using the deep learning that general convolution neural network method establishes the multi-source heterogeneous data of fully-mechanized mining working
Type.
The unitized treatment process of space reference includes: using seven parameter Coordinate Transformation Models of Mo Luojinsiji acquisition
Multi-source heterogeneous data are transformed under Chinese earth coordinates (CGCS2000),
In formula, XCGCS, YCGCS, ZCGCSIt is coordinate value under Chinese earth coordinates, XM, YM, ZMIt is multi-source heterogeneous DATA REASONING
Coordinate value under sensor coordinate system, Δ X, Δ Y, Δ Z are the translation parameters between two coordinate origins, RX, RY, RZIt is 3 coordinates
Axis rotation parameter, ε are scale parameters;
The unitized treatment process of the data attribute include: using multistage description method to multi-source heterogeneous data attribute into
The unitized processing of row, if the data attribute of isomeric data A and B are respectively as follows: (μA,sA) and (μB,sB), judge the data category of A and B
Whether property meets following formula:
In formula, sim () is data attribute compatibility function, and μ and s respectively indicate coordinate similarity and direction similarity, and N is
The sum of isomeric data, ω1And ω2It is weight parameter;
The unitized treatment process of the element relationship includes: using 9- intersection model and to calculate isomeric data A and B as follows
Topological relation equation:
In formula, R () is data topology relation function, A0,And A-Data A inner topology relationship is respectively indicated, boundary is opened up
Flutter relationship and external topological relation, B0,And B-Data A inner topology relationship is respectively indicated, boundary topology relationship and outside are opened up
Flutter relationship.
The irregular triangle network specific steps of the multi-source heterogeneous data model of fully-mechanized mining working are generated using incremental algorithm are as follows:
C1 defines the initial polygon comprising the multi-source heterogeneous data point set of all fully-mechanized mining workings;
C2 arbitrarily selects a point from the multi-source heterogeneous data point set of fully-mechanized mining working, is inserted into initial polygon focus point
Place establishes initial triangulation;
C3 is iterated calculating, until all data points are processed, that is, is firstly inserted into a discrete sampling point initial
Inside the triangulation network, the triangle comprising the point is found out in initial triangulation, and point is connect with Atria vertex, generates three
A new small triangle;Then optimize the triangulation network from the inside to surface with LOP (Local Optimization Procedure) algorithm,
Ensure that new small triangle meets the requirement of sealene triangle;
C4 deletes the triangle comprising initial polygon point, forms final irregular triangle network.
The iterated conditional of deep learning algorithm will meet following equation in step d:
In formula, P indicates the joint probability distribution function of deep learning hidden layer and input layer, is the layer of irregular triangle network
Number h, v are the number of plies number of nodes of irregular triangle network, and sigm is deep learning conditional probability excitation function, wijIndicate depth
Practise i-th of neuron of input layer and j-th of hidden layer interneuronal weight matrix, ajFor the inclined of j-th of neuron of input layer
It sets, biFor the biasing of i-th of neuron of hidden layer, nv nhRespectively indicate the quantity of input layer and hidden layer.
It verifies sample and verifies model performance step are as follows:
A training set S={ x is established according to the multi-source heterogeneous data model of fully-mechanized mining working(1),x(2),…,x(N), it carries out
Deep learning training: s(k)=f (x(k))=f (σ (wx(k)+ b)), in formula: N is expressed as the sum of multi-source heterogeneous data, and x is expressed as
The multi-source heterogeneous data to be trained, σ indicate that deep learning activation primitive, w indicate input layer to the weight matrix of hidden layer, and b is hidden
The biasing of layer is hidden, s is the training characteristics expression formula of x;
Utilize formula: y(k)=g (s(k))=g (σ (wTs(k)+bT)) deep learning verifying is carried out, in formula: wTIndicate turning for w
Set matrix, bTIndicate that the transposed matrix of b, y are the verifying feature expression of s;
Guarantee that the minimum error values ξ of multi-source heterogeneous data fusion meets in deep learning training process:
The utility model has the advantages that the invention has the following advantages over the prior art:
1) the multi-source heterogeneous data fusion method of a kind of fully-mechanized mining working based on deep learning provided by the invention, can be quickly
Fusion fully-mechanized mining working a variety of isomeric datas in recovery process, including slot wave detection data, laser scanner data, mine
Actual measurement creation data in measured data, Coherent Noise in GPR Record, 3D seismic data and recovery process, and effectively promote data fusion
Efficiency and accuracy.
2) fusion method provided by the invention, can effectively by the multi-source heterogeneous data fusion of fully-mechanized mining working together, and
The accuracy and reliability of established model can be effectively ensured, while providing accurate multi-source for the unmanned exploitation of fully-mechanized mining working
Isomeric data.
Detailed description of the invention
Fig. 1 is the multi-source heterogeneous data fusion method flow chart of the fully-mechanized mining working of the invention based on deep learning.
Fig. 2 (a) is the definition of the invention that multi-source heterogeneous data model irregular triangle network is generated using incremental algorithm
Initial polygon schematic diagram.
Fig. 2 (b) is of the invention to generate the initial of multi-source heterogeneous data model irregular triangle network using incremental algorithm
Triangulation diagram.
Fig. 2 (c) is of the invention to generate multi-source heterogeneous data model irregular triangle network insertion the using incremental algorithm
Schematic diagram after two points.
Fig. 2 (d) is of the invention to generate multi-source heterogeneous data model irregular triangle network insertion the using incremental algorithm
Schematic diagram after three points.
Fig. 2 (e) is the iteration of the invention that multi-source heterogeneous data model irregular triangle network is generated using incremental algorithm
Schematic diagram after the completion.
Fig. 2 (f) is that the multi-source heterogeneous data model irregular triangle network of utilization incremental algorithm generation of the invention is most lifelong
At irregular triangle network schematic diagram.
Specific embodiment
The invention will be further described for embodiment in reference to the accompanying drawing:
As shown in Figure 1, a kind of multi-source heterogeneous data fusion method of fully-mechanized mining working based on deep learning of the invention, root
According to fully-mechanized mining working in recovery process multi-source heterogeneous data information, establish the deep learning of the multi-source heterogeneous data of fully-mechanized mining working
Fusion Model, specific steps include:
A acquires multi-source heterogeneous data information in fully-mechanized mining working recovery process, comprising: working face slot wave detection data,
Laser scanner data, mine ground measured data, Coherent Noise in GPR Record, 3D seismic data, and to above-mentioned all multi-source heterogeneous numbers
It is believed that breath optimizes processing, comprising: carry out smothing filtering, three using action difference method removal noise spot, mean filter method
Secondary spline interpolation carries out data modification, establishes the multi-source heterogeneous data model of fully-mechanized mining working;
It is more that b using seven parameter Coordinate Transformation Models of Mo Luojinsiji converts fully-mechanized mining working for multi-source heterogeneous data information
Source Feature Between Heterogeneous Data Model, and by space reference, data attribute and the element relationship in the multi-source heterogeneous data model of fully-mechanized mining working
Carry out unitized processing;
The unitized treatment process of the space reference includes: to be obtained using seven parameter Coordinate Transformation Models handle of Mo Luojinsiji
The multi-source heterogeneous data taken are transformed under Chinese earth coordinates (CGCS2000),
In formula, XCGCS, YCGCS, ZCGCSIt is coordinate value under Chinese earth coordinates, XM, YM, ZMIt is multi-source heterogeneous DATA REASONING
Coordinate value under sensor coordinate system, Δ X, Δ Y, Δ Z are the translation parameters between two coordinate origins, RX, RY, RZIt is 3 coordinates
Axis rotation parameter, ε are scale parameters;
The unitized treatment process of the data attribute include: using multistage description method to multi-source heterogeneous data attribute into
The unitized processing of row, if the data attribute of isomeric data A and B are respectively as follows: (μA,sA) and (μB,sB), judge the data category of A and B
Whether property meets following formula:
In formula, sim () is data attribute compatibility function, and μ and s respectively indicate coordinate similarity and direction similarity, and N is
The sum of isomeric data, ω1And ω2It is weight parameter;
The unitized treatment process of the element relationship includes: using 9- intersection model and to calculate isomeric data A and B as follows
Topological relation equation:
In formula, R () is data topology relation function, A0,And A-Data A inner topology relationship is respectively indicated, boundary is opened up
Flutter relationship and external topological relation, B0,And B-Data A inner topology relationship is respectively indicated, boundary topology relationship and outside are opened up
Flutter relationship;
Shown in c such as Fig. 2 (a), irregular the three of the multi-source heterogeneous data model of fully-mechanized mining working are generated using incremental algorithm
Angle net, specific steps are as follows:
Shown in c1 such as Fig. 2 (b), defining one includes the initial polygon of the multi-source heterogeneous data point set of all fully-mechanized mining workings
Shape;
C2 arbitrarily selects a point from the data point set that c1 is defined, and is inserted at initial polygon focus point and establishes initially
The triangulation network;
C3 is iterated calculating, until all data points are processed, after as shown in Fig. 2 (c) being second point of insertion
Schematic diagram, Fig. 2 (d) are the schematic diagrames being inserted into after third point, that is, are firstly inserted into a discrete sampling point in initial triangulation
Portion finds out the triangle comprising the point in initial triangulation, and point is connect with Atria vertex, newly small of generation three
Triangle;Then optimize the triangulation network from the inside to surface with LOP (Local Optimization Procedure) algorithm, it is ensured that new
Small triangle meets the requirement of sealene triangle, wherein
C4 deletes the triangle comprising initial polygon point, final irregular triangle network is formed, as shown in Fig. 2 (f);
D extracts the characteristic parameter of irregular triangle network, and the number of plies, number of nodes and training algebra, verification including the triangulation network are
The absolute value linear convergence of the no iterated conditional for meeting deep learning algorithm, the i.e. number of plies of the triangulation network, number of nodes and training algebra,
Such as be unsatisfactory for iterated conditional, by adjusting the triangulation network the number of plies is how many, the size of quantity of node and training algebra is comprehensive to realize
The multi-source heterogeneous data model of mining face meets the iterated conditional requirement of deep learning algorithm;
The iterated conditional of the deep learning algorithm will meet following equation:
In formula, P indicates the joint probability distribution function of deep learning hidden layer and input layer, is the layer of irregular triangle network
Number h, v are the number of plies number of nodes of irregular triangle network, and sigm is deep learning conditional probability excitation function, wijIndicate depth
Practise i-th of neuron of input layer and j-th of hidden layer interneuronal weight matrix, ajFor the inclined of j-th of neuron of input layer
It sets, biFor the biasing of i-th of neuron of hidden layer, nv nhThe quantity for respectively indicating input layer and hidden layer, if Fig. 2 (e) is repeatedly
Triangulation diagram after the completion of generation;
E such as meets the iterated conditional of deep learning algorithm, then carries out feature training using deep learning training sample, later
The multi-source heterogeneous data model performance of fully-mechanized mining working, specific steps are verified using deep learning verifying sample are as follows:
A training set S={ x is established according to the multi-source heterogeneous data model of fully-mechanized mining working(1),x(2),…,x(N), it carries out
Deep learning training: s(k)=f (x(k))=f (σ (wx(k)+ b)), in formula: N is expressed as the sum of multi-source heterogeneous data, and x is expressed as
The multi-source heterogeneous data to be trained, σ indicate that deep learning activation primitive, w indicate input layer to the weight matrix of hidden layer, and b is hidden
The biasing of layer is hidden, s is the training characteristics expression formula of x;
Utilize formula: y(k)=g (s(k))=g (σ (wTs(k)+bT)) deep learning verifying is carried out, in formula: wTIndicate turning for w
Set matrix, bTIndicate that the transposed matrix of b, y are the verifying feature expression of s;
Guarantee that the minimum error values ξ of multi-source heterogeneous data fusion meets in deep learning training process:
F completes feature training in deep learning training sample in step e and deep learning verifying sample completes model
Performance verification merges mould using the deep learning that general convolution neural network method establishes the multi-source heterogeneous data of fully-mechanized mining working
Type.
Claims (5)
1. a kind of multi-source heterogeneous data fusion method of fully-mechanized mining working based on deep learning, it is characterised in that: according to fully mechanized mining work
Make face multi-source heterogeneous data information in recovery process, establishes the deep learning fusion mould of the multi-source heterogeneous data of fully-mechanized mining working
Type;Specific steps include:
A acquires multi-source heterogeneous data information in fully-mechanized mining working recovery process, comprising: working face slot wave detection data, laser
Scanner data, mine ground measured data, Coherent Noise in GPR Record, 3D seismic data, and above-mentioned all multi-source heterogeneous data are believed
Breath optimizes processing, comprising: carries out smothing filtering, three times sample using action difference method removal noise spot, mean filter method
Interpolation carries out data modification, establishes the multi-source heterogeneous data model of fully-mechanized mining working;
It is different that b using seven parameter Coordinate Transformation Models of Mo Luojinsiji converts fully-mechanized mining working multi-source for multi-source heterogeneous data information
Structure data model, and space reference, data attribute and the element relationship in the multi-source heterogeneous data model of fully-mechanized mining working are carried out
Unitized processing;
C generates the irregular triangle network of the multi-source heterogeneous data model of fully-mechanized mining working using incremental algorithm;
D extracts the characteristic parameter of irregular triangle network, and the number of plies, number of nodes and training algebra including the triangulation network verify the triangulation network
The number of plies, number of nodes and training algebra absolute value linear convergence to judging whether the iteration item for meeting deep learning algorithm
Part is such as unsatisfactory for iterated conditional, by adjusting the triangulation network the number of plies is how many, the size of quantity of node and training algebra is realized
The multi-source heterogeneous data model of fully-mechanized mining working meets the iterated conditional requirement of deep learning algorithm;
E such as meets the iterated conditional of deep learning algorithm, then carries out feature training using deep learning training sample, utilize later
Deep learning verifies sample and verifies the multi-source heterogeneous data model performance of fully-mechanized mining working,
F completes feature training in deep learning training sample in step e and deep learning verifying sample completes model performance
Verifying, the deep learning Fusion Model of the multi-source heterogeneous data of fully-mechanized mining working is established using general convolution neural network method.
2. the multi-source heterogeneous data fusion method of the fully-mechanized mining working according to claim 1 based on deep learning, special
Sign is that the unitized treatment process of space reference includes: using seven parameter Coordinate Transformation Models of Mo Luojinsiji the more of acquisition
Under source M IS to Chinese earth coordinates (CGCS2000),
In formula, XCGCS, YCGCS, ZCGCSIt is coordinate value under Chinese earth coordinates, XM, YM, ZMIt is multi-source heterogeneous DATA REASONING sensing
Coordinate value under device coordinate system, Δ X, Δ Y, Δ Z are the translation parameters between two coordinate origins, RX, RY, RZIt is 3 reference axis rotations
Turn parameter, ε is scale parameter;
The unitized treatment process of the data attribute includes: to be united using multistage description method to multi-source heterogeneous data attribute
One change processing, if the data attribute of isomeric data A and B are respectively as follows: (μA,sA) and (μB,sB), judge that the data attribute of A and B are
It is no to meet following formula:
In formula, sim () is data attribute compatibility function, and μ and s respectively indicate coordinate similarity and direction similarity, and N is isomery
The sum of data, ω1And ω2It is weight parameter;
The unitized treatment process of the element relationship includes: using 9- intersection model and to calculate isomeric data A and B topological as follows
Relation equation:
In formula, R () is data topology relation function, A0,And A-Data A inner topology relationship is respectively indicated, boundary topology closes
System and external topological relation, B0,And B-Data A inner topology relationship is respectively indicated, boundary topology relationship and external topology are closed
System.
3. the multi-source heterogeneous data fusion method of a kind of fully-mechanized mining working based on deep learning according to claim 1,
It is characterized in that generating the irregular triangle network specific steps of the multi-source heterogeneous data model of fully-mechanized mining working using incremental algorithm
Are as follows:
C1 defines the initial polygon comprising the multi-source heterogeneous data point set of all fully-mechanized mining workings;
C2 arbitrarily selects a point from the multi-source heterogeneous data point set of fully-mechanized mining working, is inserted at initial polygon focus point and builds
Vertical initial triangulation;
C3 is iterated calculating, until all data points are processed, that is, is firstly inserted into a discrete sampling point in initial triangle
Net is internal, and the triangle comprising the point is found out in initial triangulation, and point is connect with Atria vertex, generates three newly
Small triangle;Then optimize the triangulation network from the inside to surface with LOP (Local Optimization Procedure) algorithm, it is ensured that
New small triangle meets the requirement of sealene triangle;
C4 deletes the triangle comprising initial polygon point, forms final irregular triangle network.
4. the multi-source heterogeneous data fusion method of the fully-mechanized mining working according to claim 1 based on deep learning, special
Sign is that the iterated conditional of deep learning algorithm in step d will meet following equation:
In formula, P indicates the joint probability distribution function of deep learning hidden layer and input layer, is the number of plies h of irregular triangle network,
V is the number of plies number of nodes of irregular triangle network, and sigm is deep learning conditional probability excitation function, wijIndicate deep learning input
I-th of neuron of layer and j-th of hidden layer interneuronal weight matrix, ajFor the biasing of j-th of neuron of input layer, biFor
The biasing of i-th of neuron of hidden layer, nv nhRespectively indicate the quantity of input layer and hidden layer.
5. the multi-source heterogeneous data fusion method of the fully-mechanized mining working according to claim 1 based on deep learning, feature
It is to verify sample verifying model performance step are as follows:
A training set S={ x is established according to the multi-source heterogeneous data model of fully-mechanized mining working(1),x(2),…,x(N), carry out depth
Learning training: s(k)=f (x(k))=f (σ (wx(k)+ b)), in formula: N is expressed as the sum of multi-source heterogeneous data, and x is expressed as instructing
Experienced multi-source heterogeneous data, σ indicate that deep learning activation primitive, w indicate input layer to the weight matrix of hidden layer, and b is hidden layer
Biasing, s be x training characteristics expression formula;
Utilize formula: y(k)=g (s(k))=g (σ (wTs(k)+bT)) deep learning verifying is carried out, in formula: wTIndicate the transposition square of w
Battle array, bTIndicate that the transposed matrix of b, y are the verifying feature expression of s;
Guarantee that the minimum error values ξ of multi-source heterogeneous data fusion meets in deep learning training process:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910397221.XA CN110109995B (en) | 2019-05-14 | 2019-05-14 | Fully mechanized mining face multi-source heterogeneous data fusion method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910397221.XA CN110109995B (en) | 2019-05-14 | 2019-05-14 | Fully mechanized mining face multi-source heterogeneous data fusion method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110109995A true CN110109995A (en) | 2019-08-09 |
CN110109995B CN110109995B (en) | 2021-12-17 |
Family
ID=67489853
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910397221.XA Active CN110109995B (en) | 2019-05-14 | 2019-05-14 | Fully mechanized mining face multi-source heterogeneous data fusion method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110109995B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111736234A (en) * | 2020-07-27 | 2020-10-02 | 中国矿业大学 | Coal thickness prediction method based on ground penetrating radar and earthquake transmission channel waves |
CN112329302A (en) * | 2020-11-02 | 2021-02-05 | 中煤科工集团重庆研究院有限公司 | Simulation and three-dimensional visualization analysis system for underground dust production environment |
CN113746855A (en) * | 2021-09-09 | 2021-12-03 | 国网电子商务有限公司 | Data access method of energy industry cloud network and related equipment |
CN115774861A (en) * | 2022-12-22 | 2023-03-10 | 广东五度空间科技有限公司 | Natural resource multi-source heterogeneous data convergence and fusion service system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101906965A (en) * | 2010-07-14 | 2010-12-08 | 中国矿业大学(北京) | Technology for identifying and analyzing stratums in VR (Virtual Reality) geological environment |
CN102609986A (en) * | 2012-03-14 | 2012-07-25 | 中国矿业大学(北京) | Ore body modeling system in digital mine and modeling and detecting method of ore body modeling system |
WO2018184195A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Joint training of neural networks using multi-scale hard example mining |
CN108647607A (en) * | 2018-04-28 | 2018-10-12 | 国网湖南省电力有限公司 | Objects recognition method for project of transmitting and converting electricity |
CN109063753A (en) * | 2018-07-18 | 2018-12-21 | 北方民族大学 | A kind of three-dimensional point cloud model classification method based on convolutional neural networks |
CN109547032A (en) * | 2018-10-12 | 2019-03-29 | 华南理工大学 | A kind of confidence spread LDPC interpretation method based on deep learning |
CN109543828A (en) * | 2018-12-28 | 2019-03-29 | 中国石油大学(华东) | A kind of intake profile prediction technique based under condition of small sample |
-
2019
- 2019-05-14 CN CN201910397221.XA patent/CN110109995B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101906965A (en) * | 2010-07-14 | 2010-12-08 | 中国矿业大学(北京) | Technology for identifying and analyzing stratums in VR (Virtual Reality) geological environment |
CN102609986A (en) * | 2012-03-14 | 2012-07-25 | 中国矿业大学(北京) | Ore body modeling system in digital mine and modeling and detecting method of ore body modeling system |
WO2018184195A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Joint training of neural networks using multi-scale hard example mining |
CN108647607A (en) * | 2018-04-28 | 2018-10-12 | 国网湖南省电力有限公司 | Objects recognition method for project of transmitting and converting electricity |
CN109063753A (en) * | 2018-07-18 | 2018-12-21 | 北方民族大学 | A kind of three-dimensional point cloud model classification method based on convolutional neural networks |
CN109547032A (en) * | 2018-10-12 | 2019-03-29 | 华南理工大学 | A kind of confidence spread LDPC interpretation method based on deep learning |
CN109543828A (en) * | 2018-12-28 | 2019-03-29 | 中国石油大学(华东) | A kind of intake profile prediction technique based under condition of small sample |
Non-Patent Citations (6)
Title |
---|
张伟等: "基于神经网络的三角形网格智能重建", 《工程图学学报》 * |
张佑生等: "一种新颖的网格综合优化算法", 《***仿真学报》 * |
朱蕊: "多源空间矢量数据一致性处理技术研究", 《中国博士学位论文全文数据库 基础科学辑》 * |
杨金中等: "《矿山遥感监测理论方法与实践》", 30 September 2011, 测绘出版社 * |
陈雯柏等: "《人工神经网络原理与实践》", 31 January 2016, 西安电子科技大学出版社 * |
韩峰等: "《基于点云信息的既有铁路轨道状态检测与评估技术研究》", 31 December 2017, 武汉大学出版社 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111736234A (en) * | 2020-07-27 | 2020-10-02 | 中国矿业大学 | Coal thickness prediction method based on ground penetrating radar and earthquake transmission channel waves |
CN111736234B (en) * | 2020-07-27 | 2021-06-11 | 中国矿业大学 | Coal thickness prediction method based on ground penetrating radar and earthquake transmission channel waves |
CN112329302A (en) * | 2020-11-02 | 2021-02-05 | 中煤科工集团重庆研究院有限公司 | Simulation and three-dimensional visualization analysis system for underground dust production environment |
CN113746855A (en) * | 2021-09-09 | 2021-12-03 | 国网电子商务有限公司 | Data access method of energy industry cloud network and related equipment |
CN115774861A (en) * | 2022-12-22 | 2023-03-10 | 广东五度空间科技有限公司 | Natural resource multi-source heterogeneous data convergence and fusion service system |
CN115774861B (en) * | 2022-12-22 | 2023-07-21 | 广东五度空间科技有限公司 | Natural resource multi-source heterogeneous data convergence fusion service system |
Also Published As
Publication number | Publication date |
---|---|
CN110109995B (en) | 2021-12-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110109995A (en) | A kind of multi-source heterogeneous data fusion method of fully-mechanized mining working based on deep learning | |
Gibbs et al. | Approaches to three-dimensional reconstruction of plant shoot topology and geometry | |
CN107562833B (en) | Contour line-based numerical model modeling method for complex terrain three-dimensional discrete unit method | |
CN110992473B (en) | Tree branch modeling method and system based on vehicle-mounted laser scanning point cloud | |
Liu et al. | Construction of iso-contours, bisectors, and Voronoi diagrams on triangulated surfaces | |
CN110223314A (en) | A kind of single wooden dividing method based on the distribution of tree crown three-dimensional point cloud | |
CN103247041B (en) | A kind of dividing method of the cloud data of the many geometric properties based on local sampling | |
CN107146280A (en) | A kind of point cloud building method for reconstructing based on cutting | |
CN108171780A (en) | A kind of method that indoor true three-dimension map is built based on laser radar | |
CN107016725A (en) | A kind of vegetation three-dimensional live modeling method for taking LiDAR point cloud data distribution difference into account | |
Zhang et al. | Data-driven synthetic modeling of trees | |
CN103559705A (en) | Computer method for comparing similarity of different plant forms | |
CN114004938B (en) | Urban scene reconstruction method and device based on mass data | |
CN106091923A (en) | The central point rapid assay methods of industrial bolt circular hole based on three-dimensional laser scanning technique | |
CN104036550A (en) | Laser radar point-cloud interpretation and reconstruction method for building elevations on basis of shape semantics | |
CN102799763A (en) | Point cloud posture standardization-based method for extracting linear characteristic of point cloud | |
CN109829353A (en) | A kind of facial image stylizing method based on space constraint | |
CN101783016A (en) | Crown appearance extract method based on shape analysis | |
Yandun et al. | Visual 3d reconstruction and dynamic simulation of fruit trees for robotic manipulation | |
Sreevalsan-Nair et al. | Contour extraction in buildings in airborne lidar point clouds using multiscale local geometric descriptors and visual analytics | |
CN111310811A (en) | Large-scene three-dimensional point cloud classification method based on multi-dimensional feature optimal combination | |
CN103839289A (en) | Method for real tree interactive modeling based on sparse images | |
CN110992375A (en) | Chord angle discrimination clustering single tree segmentation method for layer-by-layer LiDAR point cloud | |
Kim et al. | Occlusion reasoning for skeleton extraction of self-occluded tree canopies | |
Cao et al. | Understanding 3D point cloud deep neural networks by visualization techniques |
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 |