CN108830325A - A kind of vibration information classification of landform recognition methods based on study - Google Patents
A kind of vibration information classification of landform recognition methods based on study Download PDFInfo
- Publication number
- CN108830325A CN108830325A CN201810639327.1A CN201810639327A CN108830325A CN 108830325 A CN108830325 A CN 108830325A CN 201810639327 A CN201810639327 A CN 201810639327A CN 108830325 A CN108830325 A CN 108830325A
- Authority
- CN
- China
- Prior art keywords
- neural network
- vector
- multilayer feedforward
- feedforward neural
- vibration information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Abstract
A kind of vibration information classification of landform recognition methods based on study, the present invention relates to the vibration information classification of landform recognition methods based on study.The present invention is low in order to solve the problems, such as existing classifying identification method accuracy rate.The present invention includes：One：Acquire the initial data of sensor x-axis, y-axis and z-axis direction vibration information under coordinate system；Two：Original data division processing is formed to the vector of n duration 1s；Three：Terrain type label is carried out to n vector；Four：N vector after segmentation is transformed into frequency domain；Five：N vector after will transition to frequency domain carries out off-line learning training using multilayer feedforward neural network, the multilayer feedforward neural network after being trained；Six：Real-time online obtains vibration data, executes step 2 to step 4, carries out online classification and identification using the multilayer feedforward neural network after step 5 training, obtain terrain type.Present invention sorting technique field for identification.
Description
Technical field
The present invention relates to identification sorting technique fields, and in particular to the vibration information classification of landform identification side based on study
Method.
Background technique
1) Fast Fourier Transform (FFT)
FFT (Fast Fourier Transformation) i.e. Fast Fourier Transform (FFT), is DFT (Discrete
Fourier Transformation) accelerating algorithm.It is the characteristics such as odd, even, empty, real according to discrete fourier transform, right
The algorithm of discrete Fourier transform improves acquisition.
Its basic thought is that original N point sequence is successively resolved into a series of short sequence.DFT is made full use of to calculate
Symmetric property and periodic property possessed by exponential factor in formula, so find out the corresponding DFT of these short sequences and carry out it is appropriate
Combination, reaches deletion and computes repeatedly, and reduces multiplying and simplifies the purpose of structure.Hereafter, by developing on this idea basis
The fast algorithms such as high base and split-radix.
2) neural network
Neural network is the complex networks system being widely interconnected to form by a large amount of, simple processing unit,
It reflects many essential characteristics of human brain function, is a highly complex non-linear dynamic learning system.Neural network mould
Type is described based on the mathematical model of neuron.Mainly have Serial Distribution Processing, height robustness and fault-tolerant energy
Power, distribution storage and learning ability can sufficiently approach the features such as complicated non-linear relation.Typically neural network model includes
BP neural network, Hopfield network, ART network and Kohonen network.
Summary of the invention
The purpose of the present invention is to solve the low disadvantages of existing classifying identification method accuracy rate, and propose a kind of based on
The vibration information classification of landform recognition methods of habit.
A kind of vibration information classification of landform recognition methods based on study includes the following steps：
Step 1：Acquire the initial data of sensor x-axis, y-axis and z-axis direction vibration information under body coordinate system；
Step 2：When the original data division processing of the collected vibration information of step 1 is formed n a length of t to
Amount；
Step 3：Carrying out terrain type label to n vector after step 2 dividing processing, (each vector corresponds to a kind ofly
Shape)；
Step 4：N vector after segmentation is transformed into frequency domain；
Step 5：N vector after will transition to frequency domain carries out off-line learning training using multilayer feedforward neural network, obtains
Multilayer feedforward neural network after to training；
Step 6：Real-time online obtains vibration data, executes step 2 to step 4, utilizes the multilayer after step 5 training
Feedforward neural network carries out online classification and identification, obtains terrain type.
The present invention acquires multi-dimensional vibration data and is handled not merely with the vibration data of vertical direction, first will
It is split the vector for foring duration 1s, and is standardized to it, and each moments of vibration is normalized to mean value
It is 1 for 0 and standard deviation, then utilizes Fast Fourier Transform (FFT), obtain the feature vector under frequency domain, then utilizes multilayer feedforword net
Network carries out off-line learning training, and trained network model is finally used for online detection and is classified.
The present invention collects experimental data using ground robot.It has harder rubber wheel, can generate and clearly vibrate
Signal.The algorithm improves the classification accuracy of landform using the vibration signal of multiple directions.There are two ranks for classification of landform algorithm
Section：Training and classification.Training has relatively high requirement, therefore a usually off-line step to calculating.Sorting phase right and wrong
It is often fast, it is called directly using trained model.
Beneficial effects of the present invention are：
The purpose of the present invention is to provide it is a kind of not by environment, material limit based on the identification of vibration information classification of landform
Method is made of two submodules, i.e. off-line learning training and online classification and identification, and nicety of grading is high, and generalization ability is strong, and
Have online real-time resolving ability less to computing resource consumption, improves to varying environment difference situation to landform classification and Detection
Feasibility and accuracy.
Can be accurately realized under varying environment difference situation to landform classification and Detection, at the same algorithm calculate that consumption is low can be
Line is realized, in addition to this, has reply emergency situations, as relied only on holding for vibration information under vision/laser radar failure environment
Continuous detectivity.
The present invention tests the multiple groups signal of 5 class landform of acquisition, and final experiment obtains：First kind accuracy 94.74%, second
Class accuracy 94.44%, third class accuracy are 100%, and the 4th class accuracy is 100%, and the 5th class accuracy is 100%.
Detailed description of the invention
Fig. 1 is the vibration information classification of landform recognition methods algorithm flow schematic diagram based on study.
Fig. 2 is multilayer feedforward neural network schematic diagram.
Fig. 3 is verification platform and five kinds of landform material schematic diagrames.
Fig. 4 is vibration original signal schematic diagram.
Fig. 5 is FFT transform treated signal schematic representation.
Fig. 6 is five class terrain result contrast schematic diagrams.
Fig. 7 is classification results error schematic diagram.
Specific embodiment
Specific embodiment one：As shown in Figure 1, a kind of vibration information classification of landform recognition methods based on study include with
Lower step：
Step 1：Acquire the initial data of sensor x-axis, y-axis and z-axis direction vibration information under body coordinate system；
Step 2：When the original data division processing of the collected vibration information of step 1 is formed n a length of t to
Amount；
Step 3：Carrying out terrain type label to n vector after step 2 dividing processing, (each vector corresponds to a kind ofly
Shape)；
Step 4：N vector after segmentation is transformed into frequency domain；
Step 5：N vector after will transition to frequency domain carries out off-line learning training using multilayer feedforward neural network, obtains
Multilayer feedforward neural network after to training；
Step 6：Real-time online obtains vibration data, executes step 2 to step 4, real-time online is obtained vibration data
After being transformed into frequency domain, online classification and identification is carried out using the multilayer feedforward neural network after step 5 training, obtains terrain type.
Characteristic vector pickup：
In the training stage, learn the vibration signal characteristics of known terrain type.For this purpose, robot needs to pass through different tables
Face and acquisition vibration signal.In order to collect vibration, it is based on accelerometer using one, is worked at 100 hertz.In next step,
Acceleration signal is divided into section, each section corresponds to 1 second of robot stroke, so as to generate 1 × 100 size
Vector.Then, each vector is labeled as its corresponding terrain type.
Next, original acceleration signal is transformed into frequency domain.Standardization is done to original data, by each acceleration
Degree vector is normalized to that mean value is 0 and standard deviation is 1, then does one 100 points of Fast Fourier Transform (FFT) (FFT), experiment hair
The original time-domain data of unmanned vehicle is now transformed under frequency domain to using FFT the difference that can more characterize between different terrain.By FFT
After being applied to each vector, each vector is normalized into [0,1] section.Standardization prevents high-magnitude data in training below
It is dominant.For initial data after data normalization is handled, each index is in the same order of magnitude, is appropriate for Comprehensive Correlation
Evaluation.
Currently based on the classification of landform method of vibration, acceleration (z) direction for usually indicating measurement or more is vibrated.The reason is that
Landform is jolted may significant impact in above-below direction.However, the acceleration measured in the other direction, such as after forward direction (x)
Or left and right (y), it can also be used to capture vibration.Experiment show these acceleration ratios the direction z measurement data possibly even more
It is suitble to do classification of landform.
In addition, many acceleration transducers can measure the acceleration along three axis simultaneously.For being equipped with this sensor
Robot, and the method that effectively improves classification simple the invention proposes one.For each landform section, three axis are collected
Acceleration signal, i.e. front and back (x1:100), left and right (y1:100), and (z up and down1:100).Then, with FFT respectively to the letter after standardization
It number is converted, obtains front and back (F (x)1:100), left and right (F (y)1:100), and (F (z) up and down1:100), then each signal is returned
One changes to [0,1] section.In next step, transformed three axis signal is connected as feature vector.At this moment the dimension of feature vector is 1
× 300, then these feature vectors are carried out with the training of neural network.
Specific embodiment two：The present embodiment is different from the first embodiment in that：The value of t in the step 2
For 1s~5s.
Other steps and parameter are same as the specific embodiment one.
Specific embodiment three：The present embodiment is different from the first and the second embodiment in that：When in the step 2
The vector magnitude of long 1s is 1 × 100.
Other steps and parameter are the same as one or two specific embodiments.
Specific embodiment four：Unlike one of present embodiment and specific embodiment one to three：The step 3
Mesorelief type includes：Level land, meadow, sandy soil, stone and meadow stone.
Other steps and parameter are identical as one of specific embodiment one to three.
Specific embodiment five：Unlike one of present embodiment and specific embodiment one to four：The step 4
The detailed process that the middle n vector by after segmentation is transformed into frequency domain is：
N vector after segmentation is done into standardization, i.e., each vector is normalized to mean value is 0 and standard deviation is 1；
N vector after standardization is done into Fast Fourier Transform (FFT), each vector after Fast Fourier Transform (FFT) is normalized to
[0,1] section.
Other steps and parameter are identical as one of specific embodiment one to four.
Specific embodiment six：Unlike one of present embodiment and specific embodiment one to five：The step 5
In will transition to n vector after frequency domain and carry out off-line learning training using multilayer feedforward neural network, it is more after being trained
Layer feedforward neural network detailed process be：
Multilayer feedforward neural network training
Multi-layered Feedforward Networks are mainly characterized by before signal to transmitting, Feedback error.In propagated forward, input letter
It number is handled from input layer through hidden layer composition, until output layer.Each layer of neuron state can only under the influence of one layer of neuron
State.If output layer cannot export, it is transferred to backpropagation, according to prediction error transfer factor network weight and threshold value, thus
Multilayer feedforward neural network prediction output is set constantly to approach desired output.The topological structure of multilayer feedforward neural network such as Fig. 2 institute
Show.
In attached drawing 2, I1,I2,…,InThe input value of multilayer feedforward neural network, herein for by pretreated feature to
Amount.Y1,Y2,…,YmIt is the predicted value of multilayer feedforward neural network.From the graph as can be seen that multilayer feedforward neural network can be seen
At a nonlinear function, network inputs value and predicted value are respectively the argument of function and dependent variable.Work as input number of nodes
When for n, output node number being m, multilayer feedforward neural network just expresses the Function Mapping from n independent variable to m dependent variable
Relationship.
Multilayer feedforward neural network first has to train network before doing online classification, so that network is had association's note by training
Recall and predictive ability.The training process of multilayer feedforward neural network includes following steps：
Step 5 one：Carry out multilayer feedforward neural network initialization；
According to multilayer feedforward neural network list entries (I, Y) determine network input layer number of nodes n, node in hidden layer l,
Output layer number of nodes m initializes the connection weight ω between input layer and hidden layerij, between hidden layer and output layer neuron
Connection weight ωjk, hidden layer threshold value a, output layer threshold value b are initialized, learning rate and neuron excitation function are given；
Wherein I=[I1,I2,…,In]TIt is the input value of multilayer feedforward neural network, the n after being as transformed into frequency domain
Feature vector；Y=[Y1,Y2,…,Ym]TIt is the predicted value of multilayer feedforward neural network, the terrain type as predicted；
Step 5 two：According to input variable I, input layer and implicit interlayer connection weight ωijAnd hidden layer threshold value a, meter
It calculates hidden layer and exports H；
H in formulajFor j-th of the output of H, IiFor i-th of value of input variable I, ajFor j-th of hidden layer threshold value, f is hidden
Excitation function containing layer selects following form excitation function：
Step 5 three：H, connection weight ω are exported according to hidden layerjkWith output layer threshold value b, multilayer feedforward nerve net is calculated
Network prediction output O；
Wherein OkFor k-th of value for predicting output vector O, bkFor k-th of output layer threshold value；
Step 5 four：Output O and desired output Y is predicted according to multilayer feedforward neural network, calculates neural network forecast error e；
ek=Yk-OkK=1,2 ..., m (4)
Wherein ekFor k-th of neural network forecast error, YkFor the predicted value of k-th of multilayer feedforward neural network；
Step 5 five：Network connection weight ω is updated according to neural network forecast error eij、ωjk；
η is learning rate in formula；
Step 5 six：Network node threshold value a, b are updated according to neural network forecast error e；
Step 5 seven：Judge whether neural network forecast error e is less than setting value, terminates if being less than, otherwise iteration executes step
Rapid 52 to step 5 seven.
After completing training, neural network network can enter the online classification stage, and robot passes through unknown landform and adopts
Collect vibration signal.Once per second, the acceleration signal for merging three directions generate one 1 × 300 vector (x1:100,y1:100,
z1:100), standardize to each component, then with three components of the FFT transform vector, obtains (F (x)1:100,F(y)1:100,
F(z)1:100).In addition it is also necessary to by it, normalizationization finally obtains test vector respectively.Then, trained neural network pair
Test vector is classified and returns to the estimation to terrain type.
Other steps and parameter are identical as one of specific embodiment one to five.
Beneficial effects of the present invention are verified using following embodiment：
Embodiment one：
1) experimental situation
To verify correctness and reasonability of the invention, present invention experiment uses the Classification and Identification of 5 class unlike material landform,
Specific environment is as shown in Fig. 3.
2) experimental result and analysis
The present invention tests the multiple groups signal of 5 class landform of acquisition, and signal element is divided into 1s, then carries out a series of above-mentioned behaviour
Make, original signal and finally obtained feature vector are as shown in attached drawing 4 and Fig. 5.It is not difficult to find out that the original signal in time domain is distinguished
Less, feature becomes relatively obvious to degree after being transformed to frequency domain, this is beneficial to the training of subsequent neural network, Ke Yi great
It is big to improve classification accuracy.
5 class landform (level land, meadow, sandy soil, stone, meadow stone mix) are indicated with number 1,2,3,4,5 respectively, it is attached
Fig. 6 is the comparison that neural network predicts terrain category and actual landform classification on test set.Error in classification is as shown in Fig. 7.
Final experiment obtains：First kind accuracy 94.74%, the second class accuracy 94.44%, third class accuracy are
100%, the 4th class accuracy is 100%, and the 5th class accuracy is 100%.
15 experiments are carried out, it is as follows to obtain 5 class classification of landform accuracy：
Accuracy
0.9474 0.9444 1.0000 1.0000 1.0000
Accuracy
1.0000 0.9500 0.9130 1.0000 1.0000
Accuracy
0.9545 0.9500 0.7368 0.9091 1.0000
Accuracy
1.0000 1.0000 0.9545 0.9474 0.9545
Accuracy
1.0000 1.0000 0.9375 1.0000 1.0000
Accuracy
0.9583 1.0000 0.7778 1.0000 0.8824
Accuracy
1.0000 0.9565 0.9444 1.0000 0.9474
Accuracy
0.9091 1.0000 0.9333 1.0000 0.9167
Accuracy
1.0000 0.8667 0.8667 1.0000 0.8750
Accuracy
1.0000 0.9474 1.0000 0.9259 1.0000
Accuracy
1.0000 1.0000 0.8421 1.0000 0.9048
Accuracy
1.0000 1.0000 1.0000 0.9524 1.0000
Accuracy
1.0000 0.9600 0.7273 0.9615 0.9500
Accuracy
1.0000 1.0000 0.9375 0.9600 0.9231
Accuracy
1.0000 1.0000 0.8571 0.9474 1.0000
Classification method according to the present invention simple and clear can carry out Classification and Identification to varying environment landform based on vibration information,
New thinking is provided for terrain detection sort research.
The present invention can also have other various embodiments, without deviating from the spirit and substance of the present invention, this field
Technical staff makes various corresponding changes and modifications in accordance with the present invention, but these corresponding changes and modifications all should belong to
The protection scope of the appended claims of the present invention.
Claims (6)
1. a kind of vibration information classification of landform recognition methods based on study, it is characterised in that：The vibration letter based on study
Breath classification of landform recognition methods includes the following steps：
Step 1：Acquire the initial data of sensor x-axis, y-axis and z-axis direction vibration information under body coordinate system；
Step 2：The vector of a length of t when the original data division processing of the collected vibration information of step 1 is formed n；
Step 3：Terrain type label is carried out to n vector after step 2 dividing processing；
Step 4：N vector after segmentation is transformed into frequency domain；
Step 5：N vector after will transition to frequency domain carries out off-line learning training using multilayer feedforward neural network, is instructed
Multilayer feedforward neural network after white silk；
Step 6：Real-time online obtains vibration data, executes step 2 to step 4, utilizes the multilayer feedforward after step 5 training
Neural network carries out online classification and identification, obtains terrain type.
2. a kind of vibration information classification of landform recognition methods based on study according to claim 1, it is characterised in that：It is described
The value of t is 1s~5s in step 2.
3. a kind of vibration information classification of landform recognition methods based on study according to claim 1 or claim 2, it is characterised in that：
The vector magnitude of a length of t is 1 × 100 when in the step 2.
4. a kind of vibration information classification of landform recognition methods based on study according to claim 3, it is characterised in that：It is described
Step 3 mesorelief type includes：Level land, meadow, sandy soil, stone and meadow stone.
5. a kind of vibration information classification of landform recognition methods based on study according to claim 4, it is characterised in that：It is described
It is by the detailed process that n vector after segmentation is transformed into frequency domain in step 4：
N vector after segmentation is done into standardization, i.e., each vector is normalized to mean value is 0 and standard deviation is 1；It will mark
Treated that n vector does Fast Fourier Transform (FFT) for standardization, and each vector after Fast Fourier Transform (FFT) is normalized to [0,1]
Section.
6. a kind of vibration information classification of landform recognition methods based on study according to claim 5, it is characterised in that：It is described
N vector after will transition to frequency domain in step 5 carries out off-line learning training using multilayer feedforward neural network, is trained
The detailed process of multilayer feedforward neural network afterwards is：
Step 5 one：Carry out multilayer feedforward neural network initialization；
Network input layer number of nodes n, node in hidden layer l, output are determined according to multilayer feedforward neural network list entries (I, Y)
Node layer number m initializes the connection weight ω between input layer and hidden layerij, company between hidden layer and output layer neuron
Meet weight ωjk, hidden layer threshold value a, output layer threshold value b are initialized, learning rate and neuron excitation function are given；
Wherein I=[I1,I2,…,In]TThe input value of multilayer feedforward neural network, n feature after being as transformed into frequency domain to
Amount；Y=[Y1,Y2,…,Ym]TIt is the predicted value of multilayer feedforward neural network, the terrain type as predicted；
Step 5 two：According to input variable I, input layer and implicit interlayer connection weight ωijAnd hidden layer threshold value a, it calculates hidden
H is exported containing layer；
H in formulajFor j-th of the output of H, IiFor i-th of value of input variable I, ajFor j-th of hidden layer threshold value, f is hidden layer
Excitation function selects following form excitation function：
Step 5 three：H, connection weight ω are exported according to hidden layerjkWith output layer threshold value b, it is pre- to calculate multilayer feedforward neural network
Survey output O；
Wherein OkFor k-th of value for predicting output vector O, bkFor k-th of output layer threshold value；
Step 5 four：Output O and desired output Y is predicted according to multilayer feedforward neural network, calculates neural network forecast error e；
ek=Yk-OkK=1,2 ..., m (4)
Wherein ekFor k-th of neural network forecast error, YkFor the predicted value of k-th of multilayer feedforward neural network；
Step 5 five：Network connection weight ω is updated according to neural network forecast error eij、ωjk；
η is learning rate in formula；
Step 5 six：Network node threshold value a, b are updated according to neural network forecast error e；
Step 5 seven：Judge whether neural network forecast error e is less than setting value, terminate if being less than, otherwise iteration executes step 5
Two to step 5 seven.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810639327.1A CN108830325A (en) | 2018-06-20 | 2018-06-20 | A kind of vibration information classification of landform recognition methods based on study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810639327.1A CN108830325A (en) | 2018-06-20 | 2018-06-20 | A kind of vibration information classification of landform recognition methods based on study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108830325A true CN108830325A (en) | 2018-11-16 |
Family
ID=64141712
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810639327.1A Pending CN108830325A (en) | 2018-06-20 | 2018-06-20 | A kind of vibration information classification of landform recognition methods based on study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830325A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109948556A (en) * | 2019-03-21 | 2019-06-28 | 中国农业科学院农业资源与农业区划研究所 | Crops Classification recognition methods and system |
| CN110051292A (en) * | 2019-05-29 | 2019-07-26 | 尚科宁家(中国)科技有限公司 | A kind of sweeping robot control method |
| CN110956651A (en) * | 2019-12-16 | 2020-04-03 | 哈尔滨工业大学 | Terrain semantic perception method based on fusion of vision and vibrotactile sense |
| CN110956154A (en) * | 2019-12-11 | 2020-04-03 | 哈尔滨高斯触控科技有限公司 | Vibration information terrain classification and identification method based on CNN-LSTM |
| CN111027627A (en) * | 2019-12-11 | 2020-04-17 | 哈尔滨高斯触控科技有限公司 | Vibration information terrain classification and identification method based on multilayer perceptron |
| CN111539132A (en) * | 2020-07-09 | 2020-08-14 | 南京航空航天大学 | Dynamic load time domain identification method based on convolutional neural network |
| CN111680642A (en) * | 2020-06-11 | 2020-09-18 | 石家庄铁道大学 | Terrain classification method and device |
| CN112257817A (en) * | 2020-12-18 | 2021-01-22 | 之江实验室 | Geological geology online semantic recognition method and device and electronic equipment |
| CN113065388A (en) * | 2021-02-03 | 2021-07-02 | 湖南大学 | Real-time soil category identification method and system and excavator |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101275900A (en) * | 2008-05-08 | 2008-10-01 | 江汉大学 | Method for recognizing road surface types based on vehicle wheel vibration |
| JP2010095135A (en) * | 2008-10-16 | 2010-04-30 | Toyota Motor Corp | Wheel vibration extraction device and road surface state estimation device |
| CN102289674A (en) * | 2011-07-15 | 2011-12-21 | 王世峰 | Pavement type recognition method and device based on vertical acceleration and pavement image |
| CN105426858A (en) * | 2015-11-26 | 2016-03-23 | 哈尔滨工业大学 | Vision and vibration information fusion based ground type identification method |
| CN105844211A (en) * | 2015-01-29 | 2016-08-10 | 通用汽车环球科技运作有限责任公司 | System and method for classifying a road surface |
| WO2017012978A1 (en) * | 2015-07-17 | 2017-01-26 | Jaguar Land Rover Limited | Acoustic sensor for use in a vehicle |
| WO2018007079A1 (en) * | 2016-07-08 | 2018-01-11 | Jaguar Land Rover Limited | Improvements in vehicle speed control |
| CN107644456A (en) * | 2017-09-22 | 2018-01-30 | 南京理工大学 | Sea-floor relief reconstructing system and method based on polarization characteristics of lasers |
| CN107977641A (en) * | 2017-12-14 | 2018-05-01 | 东软集团股份有限公司 | A kind of method, apparatus, car-mounted terminal and the vehicle of intelligent recognition landform |
| CN108027427A (en) * | 2015-07-17 | 2018-05-11 | 捷豹路虎有限公司 | The ultrasonic transducer system for being used for landform identification in vehicle |
-
2018
- 2018-06-20 CN CN201810639327.1A patent/CN108830325A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101275900A (en) * | 2008-05-08 | 2008-10-01 | 江汉大学 | Method for recognizing road surface types based on vehicle wheel vibration |
| JP2010095135A (en) * | 2008-10-16 | 2010-04-30 | Toyota Motor Corp | Wheel vibration extraction device and road surface state estimation device |
| CN102289674A (en) * | 2011-07-15 | 2011-12-21 | 王世峰 | Pavement type recognition method and device based on vertical acceleration and pavement image |
| CN105844211A (en) * | 2015-01-29 | 2016-08-10 | 通用汽车环球科技运作有限责任公司 | System and method for classifying a road surface |
| WO2017012978A1 (en) * | 2015-07-17 | 2017-01-26 | Jaguar Land Rover Limited | Acoustic sensor for use in a vehicle |
| CN108027427A (en) * | 2015-07-17 | 2018-05-11 | 捷豹路虎有限公司 | The ultrasonic transducer system for being used for landform identification in vehicle |
| CN105426858A (en) * | 2015-11-26 | 2016-03-23 | 哈尔滨工业大学 | Vision and vibration information fusion based ground type identification method |
| WO2018007079A1 (en) * | 2016-07-08 | 2018-01-11 | Jaguar Land Rover Limited | Improvements in vehicle speed control |
| CN107644456A (en) * | 2017-09-22 | 2018-01-30 | 南京理工大学 | Sea-floor relief reconstructing system and method based on polarization characteristics of lasers |
| CN107977641A (en) * | 2017-12-14 | 2018-05-01 | 东软集团股份有限公司 | A kind of method, apparatus, car-mounted terminal and the vehicle of intelligent recognition landform |
Non-Patent Citations (5)
| Title |
|---|
| CHRISTIAN WEISS等: "SVMs for vibration-based terrain classification", 《AUTONOME MOBILE SYSTEME 2007》 * |
| CHRISTIAN WEISS等: "Vibration-based Terrain Classification Using Support Vector Machines", 《2006 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS》 * |
| RUBKWAN JITPAKDEE等: "Neural Networks Terrain Classification using Inertial Measurement Unit for an Autonomous Vehicle", 《2008 SICE ANNUAL CONFERENCE》 * |
| 韩明等: "《数学建模案例》", 30 June 2012, 同济大学出版社 * |
| 高亮等: "《类电磁机制算法的研究与应用》", 30 November 2017, 华中科技大学出版社 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109948556A (en) * | 2019-03-21 | 2019-06-28 | 中国农业科学院农业资源与农业区划研究所 | Crops Classification recognition methods and system |
| CN110051292A (en) * | 2019-05-29 | 2019-07-26 | 尚科宁家(中国)科技有限公司 | A kind of sweeping robot control method |
| CN110051292B (en) * | 2019-05-29 | 2021-11-02 | 尚科宁家(中国)科技有限公司 | Control method of floor sweeping robot |
| CN110956154A (en) * | 2019-12-11 | 2020-04-03 | 哈尔滨高斯触控科技有限公司 | Vibration information terrain classification and identification method based on CNN-LSTM |
| CN111027627A (en) * | 2019-12-11 | 2020-04-17 | 哈尔滨高斯触控科技有限公司 | Vibration information terrain classification and identification method based on multilayer perceptron |
| CN110956651A (en) * | 2019-12-16 | 2020-04-03 | 哈尔滨工业大学 | Terrain semantic perception method based on fusion of vision and vibrotactile sense |
| CN110956651B (en) * | 2019-12-16 | 2021-02-19 | 哈尔滨工业大学 | Terrain semantic perception method based on fusion of vision and vibrotactile sense |
| CN111680642A (en) * | 2020-06-11 | 2020-09-18 | 石家庄铁道大学 | Terrain classification method and device |
| CN111680642B (en) * | 2020-06-11 | 2023-06-23 | 石家庄铁道大学 | Terrain classification method and device |
| CN111539132A (en) * | 2020-07-09 | 2020-08-14 | 南京航空航天大学 | Dynamic load time domain identification method based on convolutional neural network |
| CN112257817A (en) * | 2020-12-18 | 2021-01-22 | 之江实验室 | Geological geology online semantic recognition method and device and electronic equipment |
| CN113065388A (en) * | 2021-02-03 | 2021-07-02 | 湖南大学 | Real-time soil category identification method and system and excavator |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830325A (en) | A kind of vibration information classification of landform recognition methods based on study | |
| CN108960426B (en) | Road slope comprehensive estimation system based on BP neural network | |
| CN103489009B (en) | Mode identification method based on adaptive correction neutral net | |
| CN105260786B (en) | A kind of simulation credibility of electric propulsion system assessment models comprehensive optimization method | |
| CN101404071B (en) | Electronic circuit fault diagnosis neural network method based on grouping particle swarm algorithm | |
| CN108053052B (en) | A kind of oil truck oil and gas leakage speed intelligent monitor system | |
| CN111680875B (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN108362510A (en) | A kind of engineering goods method of fault pattern recognition based on evidence neural network model | |
| CN112560079A (en) | Hidden false data injection attack detection method based on deep belief network and transfer learning | |
| JP7464236B2 (en) | Complex environment model, perception system and perception method for autonomous vehicles using complex networks | |
| Siregar | Neural Network Analysis With Backpropogation In Predicting Human Development Index (HDI) Component by Regency/City In North Sumatera | |
| Göktepe et al. | Comparison of multilayer perceptron and adaptive neuro-fuzzy system on backcalculating the mechanical properties of flexible pavements | |
| Liu et al. | Road roughness acquisition and classification using improved restricted Boltzmann machine deep learning algorithm | |
| De Luca et al. | Artificial Neural Networks for forecasting user flows in transportation networks: literature review, limits, potentialities and open challenges | |
| CN111967308A (en) | Online road surface unevenness identification method and system | |
| CN107976934A (en) | A kind of oil truck oil and gas leakage speed intelligent early-warning system based on wireless sensor network | |
| Wang et al. | A hybrid fuzzy method for performance evaluation of fusion algorithms for integrated navigation system | |
| Zhengxiang et al. | Transport volume forecast based on GRNN network | |
| CN113884903B (en) | Battery fault diagnosis method based on multi-layer perceptron neural network | |
| CN115359197A (en) | Geological curved surface reconstruction method based on spatial autocorrelation neural network | |
| Pandit et al. | Identification and ranking of weak buses using modified counterpropagation neural network | |
| Zhaojing et al. | Weigh-in-motion based on multi-sensor and RBF neural network | |
| Tanyu et al. | Parameter identification by deep learning of a material model for granular media | |
| Ye et al. | Driving Cycle Condition Identification Model Based on Long Short-Term Memory Algorithm [J] | |
| ZHAO et al. | Impact Monitoring of Large and Complex Structures Based on Transfer Learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181116 |