CN108896996A - A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest - Google Patents
A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest Download PDFInfo
- Publication number
- CN108896996A CN108896996A CN201810453235.4A CN201810453235A CN108896996A CN 108896996 A CN108896996 A CN 108896996A CN 201810453235 A CN201810453235 A CN 201810453235A CN 108896996 A CN108896996 A CN 108896996A
- Authority
- CN
- China
- Prior art keywords
- signal
- echo
- threshold
- decision tree
- sample
- 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
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
- G01S15/02—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems using reflection of acoustic waves
- G01S15/06—Systems determining the position data of a target
- G01S15/08—Systems for measuring distance only
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
- G01S15/88—Sonar systems specially adapted for specific applications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/52—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00
- G01S7/539—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention discloses absorbing well, absorption well water sludge interface ultrasonic echo signal classification methods under a kind of Pb-Zn deposits based on random forest, include the following steps：Ultrasonic echo signal under different operating conditions is collected using being mounted on the underwater ultrasonic transducer of absorbing well, absorption well first；The signal being collected into is subjected to wavelet decomposition, calculates threshold value using Hesusure Research on threshold selection, carries out coefficient processing using soft-threshold function, then reconstruction signal completes denoising；To the signal extraction modulus maximum feature after denoising；Have put back to randomly select modulus maximum feature and part sample establishes decision tree base learner, form random forest grader for Modulation recognition by more decision trees.The present invention is high to the classification accuracy of echo-signal, and operation cost is low, has greatly value to the echo parameter estimation under different mathematics.
Description
Technical field
The invention belongs to ultrasound examination fields, and in particular to a kind of Pb-Zn deposits underground absorbing well, absorption well mud ultrasonic echo point
Class method.
Background technique
Mine drainage is an important factor for influencing mining area safety production safely, and drainage underground is to underground work personnel and equipment
Safety is extremely important, is the most basic condition of underground work.Due to underground work exploitation etc., the water of underground is all more muddy
Turbid, water is easy to generate mud in absorbing well, absorption well after absorbing well, absorption well aggregation.The measurement of water suction Well Water Level is supersonic liquid level at present
Meter provides liquid level by measurement liquid level surface distance, and this measurement method has ignored the silt depth of shaft bottom precipitating, with
Dampening assemble index is longer, and error is increasing.Water color in absorbing well, absorption well is deeper, and the mud of precipitated sludge can not be estimated with eye
Position height causes water pump to take out and is extracted into mud less than water or water pump, cause water pump when mud height reaches Pump Suction water level
Damage.With the further development of computer automation technology, unattended pump house etc. is a development trend, if water
Position measurement is inaccurate, or cannot accurately know the thickness of mud, may cause serious consequence.Therefore the accurate of mud is obtained
Thickness is to realize mine water pump house etc. unattended reliable guarantee.
When absorbing well, absorption well SEA LEVEL VARIATION is slower, solid particle in water precipitating sufficiently, shaft bottom formed one it is relatively clear
Water sludge interface all reflected since the acoustic reactance of water differs larger with the acoustic reactance of mud when ultrasonic wave travels to water sludge interface, surpass
Sonic sensor receives the ultrasound echo signal of single echo, is easier to estimate TOF (ultrasonic propagation time).When solid
When grain precipitating is insufficient, ultrasonic signal may generate multiple echo-signals, echo quantity it is different for mud position estimate
Method is different, it is therefore desirable to which the corresponding echo-signal type of different operating conditions is classified.
Ultrasonic distance measurement is based on generating reflection signal when ultrasonic signal encounters measured object, can be estimated according to echo-signal
The distance between ultrasonic probe and measured object out, when Most current ultrasonic distance measurement is propagated using threshold method estimation ultrasonic wave
Between, under Pb-Zn deposits in absorbing well, absorption well, there are impurity under heavy metal particles and other mines to interfere, and is difficult to obtain using conventional threshold values method
Accurate information.Therefore it designs and a kind of estimates that ultrasonic propagation time can pole based on the estimation method of Gauss echo signal model
Big raising measurement accuracy, and the corresponding echo mathematical model of different echo-signals is different, and the echo under different models is believed
It number sorts out to be the premise accurately measured.
Summary of the invention
The ultrasonic echo signal point that absorbing well, absorption well sludge level generates under mine is directed to the object of the present invention is to provide a kind of
The method of class.Noise reduction is carried out to signal first, reduces the interference that ambient noise and sensor self-noise extract subsequent characteristics,
Then modulus maximum feature is extracted, is instructed in conjunction with random forests algorithm using the signal being collected into based on extracted feature
Practice, establishes the classifier based on random forests algorithm and classify to signal.
Specific step is as follows for technical solutions according to the invention：
S1：Ultrasonic transducer probe is mounted on to the position of distance water suction bottom fixed height, uses ultrasonic waves
Energy device and signal acquiring system obtain echo-signal of the ultrasonic wave after water suction bottom water sludge interface reflection, in different operating conditions
Lower collection p group echo-signal, p meet 50≤p≤500；
S2：9 layers of wavelet decomposition are carried out for ultrasonic echo signal, each layer wavelet coefficient is obtained, uses Hesusure threshold
It is worth choosing method and carries out noise reduction process, carries out coefficient processing using soft-threshold function, wavelet coefficient carries out letter using treated
Number reconstruct, improve Signal-to-Noise；
S3：The dyadic wavelet decomposition algorithm that Mallat is carried out to the signal after noise reduction is decomposed into 6 grades of scales, selects 22、23、
24、25Scale space is as feature extraction space, respectively in every layer of 4 modulus maximum of extraction and the modulus maximum corresponding time
Shaft position forms 32 dimensional feature vector F=[V1, T1, V2, T2...,V16, T16]；
S4：Using the feature vector obtained in S3 alternatively attribute, 4 dimensional features are therefrom randomly selected as single base
The feature input for practising device, randomly selects p from sample of signal obtained in S1xA sample is defeated as the sample of single decision tree
Enter, with Gini impurity level come select divide attribute, establish CART decision tree；
S5：Step S4 is repeated into ncIt is secondary, establish ncThe RandomForest classifier of decision tree composition；
S6：For the data being newly collected into, it is input in the RandomForest classifier of above step foundation, each
Prediction classification is carried out in CART decision tree, final classification result is selected in a manner of ballot.
In the S1, by ultrasonic transducer probe installation, fixed height position, the selected probe angle of departure are 6 ° under water,
Probe is h apart from bottom of pond height, and historical high mud position is hm, h meets h-hm>=30cm needs to exclude to produce when installation is popped one's head in
The barrier of raw interference echo, the point centered on probe, probe face is horizontal plane, and to down toward bottom of pond, radius is h*tan3 ° of+h '
Cylindrical space in for measurement space, solid obstacle there may be interference echo should be excluded by measuring in space, h ' is superfluous
Surplus meets
9 layers of wavelet decomposition are carried out to original signal using wavelet function in the S2, obtain each layer wavelet coefficient, are used
Hesusure threshold value selection rule, in conjunction with fixed threshold rule and unbiased possibility predication adaptive threshold rule, specific steps
It is as follows：
S21：Threshold value as defined in fixed threshold rule is calculated firstN is signal length, and σ is noise
Standard deviation；
S22：Then unbiased possibility predication adaptive threshold Rigrsure threshold value, wavelet coefficient square vector X=are calculated
[x1, x2..., xn], wherein x1≤x2≤…≤xn, n is the number of this layer of wavelet coefficient, if a risk vector is：S=[s1,
s2..., si..., sn], each element is in risk vector：From risk vector
In find out minimum value sminAs value-at-risk, the corresponding x of minimum risk value subscript coefficient is found outmin, then adaptive threshold be：
S23：In conjunction with above two threshold value selection rule, if K is certain layer of wavelet coefficient quadratic sum after wavelet decomposition, definition ginseng
NumberParameterMixed type threshold value λhValue rule isWhen
After threshold value, soft-threshold function is selected to handle wavelet coefficient, soft-threshold function is：
Wherein xnFor through soft-threshold function treated new wavelet systems；
S24：After carrying out threshold process to wavelet coefficient, signal reconstruction is carried out, the signal after obtaining noise reduction.
Signal decomposition after noise reduction is 6 grades of scales, chosen by the dyadic wavelet decomposition algorithm that Mallat is used in the S3
22、23、24、25As feature extraction space, 4 modulus maximums of extraction are used as in each feature extraction space is somebody's turn to do scale space
Layer feature, extracts the corresponding time shaft location parameter of each feature extreme value, forms 4*4*2=32 dimensional feature vector F=[V1,
T1, V2, T2...,V16, T16] as classifier division attribute.
It selects Gini impurity level as attribute selection rule is divided in the S4, CART decision tree is established, from sample data
In put back to randomly select pxA sample data is inputted as single decision tree data, have from 32 dimensional feature vectors put back to
Machine extracts 4 dimensions as decision tree and divides attribute, and Gini impurity level is defined as：
Wherein y is sample class quantity, pkFor kth class sample proportion, wherein the Gini impurity level definition of attribute Z
For：
For the 4 dimensional feature values extracted, every dimension includes pxA sample data, there is px+ 1 can division points, definition is candidate
Division points areDefinitionWherein [z1, z2..., zi..., zj] be
The j different values on attribute Z, j is single decision tree sample size, according to formula
Optimal dividing attribute and two optimal dividing points are selected, so that the data Gini impurity level after dividing accordingly is most
It is small.
Pb-Zn deposits absorbing well, absorption well water sludge interface echo-signal is divided into three kinds of classifications, single echo signal, double echo letter by the S5
Number and three echo-signals, respectively correspond three kinds of different water sludge interface situations, can preferably estimate water sludge interface position and heavy
Starch constituent.
The present invention proposes a kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification side based on random forest
Method.It is continual in absorbing well, absorption well under Pb-Zn deposits to there is the underground water generated in mining process to flow into and extract out, by what is flowed into
In underground water contain a large amount of mud and heavy metal particles impurity, therefore can water suction bottom generate precipitating, silt depth with
Time persistently thickens.Ingredient difference contained in precipitating can then generate different echo-signal, for different echo-signals,
Corresponding model is different, is not available identical parameter model and carries out parameter Estimation to it, thus identifies echo-signal classification
It is estimation parameter only way.Collect echo-signal first and be used as input signal, include in input signal largely noise at
Point, ambient noise and sensor itself are mostly come from, therefore using the method for Wavelet Denoising Method, combine unbiased possibility predication certainly
Threshold rule and fixed threshold rule are adapted to, soft-threshold function is chosen and wavelet coefficient is handled, experiments have shown that working as small wavelength-division
It is best to solve denoising effect when the number of plies is 9 layers.Contain reflecting interface structure and shape etc. in wavelet modulus maxima largely to believe
Breath, by the scale parameter of modulus maximum, range parameter and corresponding time shaft parameter can return target as target characteristic amount
Wave is classified, and echo signal is carried out to the dyadic wavelet decomposition algorithm of Mallat herein, is decomposed into 6 grades of scales, selects 22、23、
24、25Scale space is as feature extraction space, respectively in every layer of 4 modulus maximum of extraction and the modulus maximum corresponding time
Shaft position forms 32 dimensional feature vectors.Echo-signal is divided into three kinds of classifications herein, since sample characteristics are continuous variable, institute
To need to select two division points, will divide attribute rule definition herein is：
It is regular accordingly, it selects and divides attribute and division points, when division points include g0Or gj+1In one when, indicate
Only there are two types of signal classifications in the sample extracted, when selected division points are g0And gj+1When, it indicates only one in institute's sample drawn
Kind signal classification.Random forest avoids asking for single decision tree over-fitting or mistake classification as a kind of integrated study mode
Topic, every decision tree is trained by the way of randomly selecting feature and sample, constitutes forest, base by the decision tree of certain scale
In the classification results of every decision tree, carries out ballot mode finally to determine final result, greatly improve nicety of grading.
Detailed description of the invention
Fig. 1 is flow diagram of the present invention；
Fig. 2 is to carry out the comparison diagram after unbiased adaptive threshold noise reduction and heuristic threshold deniosing respectively to original signal.
Specific embodiment
In order in further detail, specific description technical solution of the present invention and advantage, it is with reference to the accompanying drawings and examples, right
The present invention is further described in detail.
Fig. 1 is flow diagram of the present invention, gives basic procedure sequence of the invention.Detailed process includes following step
Suddenly：
S1：In order to be collected into echo-signal, needs for ultrasonic transducer probe to be mounted on distance water suction bottom and fix
The position of height, the ultrasonic transducer angle of departure used in the present embodiment is 6 °, in order to make in the signal being collected into without dry
Signal, selected distance pool wall 0.5m are disturbed, sluggish flow precipitates sufficient region, pops one's head in using irony stock fixing of energy converter.Iron
Matter stock upper end is fixed on above absorbing well, absorption well on baffle, and lower end connects transducer probe, is fixed on the height and position chosen.This
The water suction of Pb-Zn deposits selected by embodiment well depth 3.5m, historical high mud position 0.9m, water level is generally kept in 1.5m height or more, therefore selects
It selects away from the fixed probe of bottom of pond 1.4m height.Run signal acquisition system after ultrasonic transducer is installed, obtains ultrasonic wave through inhaling
Echo-signal after the reflection of well bottom water sludge interface, collects 50 groups of echo-signals under different operating conditions.
S2：9 layers of wavelet decomposition are carried out for ultrasonic echo signal, each layer wavelet coefficient is obtained, uses Hesusure threshold
It is worth choosing method and carries out noise reduction process, carries out coefficient processing using soft-threshold function, wavelet coefficient carries out letter using treated
Number reconstruct, improve Signal-to-Noise, in conjunction with fixed threshold rule and unbiased possibility predication adaptive threshold rule, calculate first
Threshold value as defined in fixed threshold ruleN is signal length, and σ is the standard deviation of noise, then calculate it is unbiased seemingly
Right estimation self-adaptive threshold value Rigrsure threshold value, vector X=[x1, x2..., xn], wherein x1≤x2≤…≤xn, n is this layer small
The number of wave system number, if a risk vector is：S=[s1, s2..., si..., sn], each element is in risk vector：Minimum value s is found out from risk vectorminAs value-at-risk, minimum is found out
The corresponding x of value-at-risk subscript coefficientmin, then adaptive threshold be：In conjunction with above two threshold value selection rule,
If K is certain layer of wavelet coefficient quadratic sum after wavelet decomposition, defined parametersParameterMixed type
Threshold value λhValue rule isAfter threshold value, select soft-threshold function to wavelet coefficient
It is handled, soft-threshold function is：
Wherein xnFor through soft-threshold function treated new wavelet coefficient, the present embodiment uses gradient thresholdEach layer of wavelet coefficient all corresponds to different threshold values, and Fig. 2 is to use heuristic threshold rule and unbiased respectively
Adaptive threshold rule carries out the comparison after noise reduction, it can be seen that the denoising effect under heuristic threshold rule is substantially better than unbiased
Adaptive threshold rule.
S3：The dyadic wavelet decomposition algorithm that Mallat is carried out to the echo-signal after noise reduction, is decomposed into 6 grades of scales, selects
22、23、24、25Scale space is corresponding in every layer of 4 modulus maximum of extraction and the modulus maximum respectively as feature extraction space
Time shaft position, form 32 dimensional feature vector F=[V1, T1, V2, T2...,V16, T16]；
S4：Using the feature vector obtained in S3 alternatively attribute, 32 dimensional features are shared, formula d=log is used2K is calculated
The characteristic extracted out is 4 in the present embodiment, and the feature for therefrom randomly selecting 4 dimensional features as single base learner inputs, from
It randomly selects sample input of 10 samples as single decision tree in sample of signal obtained in S1, is selected with Gini impurity level
Division attribute is selected, CART decision tree is established, Gini impurity level is defined as：
Wherein y is sample class quantity, pkFor kth class sample proportion, wherein the Gini impurity level definition of attribute Z
For：
For the 4 dimensional feature values extracted, every dimension includes pxA sample data, have 11 can division points, fixed definition waits
The division points are selected to beDefinitionWherein [z1, z2..., zi..., zj]
For the j different values on attribute Z, j is single decision tree sample size, according to formula
Optimal dividing attribute and two optimal dividing points are selected, so that the data Gini impurity level after dividing accordingly is most
It is small.Decision Tree algorithms are often trapped in over-fitting state, divide that number is excessive, may training set data it is some itself
Feature as input data general features and there is error in classification, therefore need to carry out beta pruning to tree establishing single decision tree
Processing and termination condition are arranged to prevent over-fitting.Every decision tree uses 10 sample datas in the present embodiment, randomly selects 4
Dimensional feature vector, single decision tree data volume is smaller, and random forests algorithm inherently can be to avoid the beta pruning of single decision tree
Operation, therefore need to only be arranged termination condition, provide in the present embodiment each child node only have a type of signal or
Gini impurity level after all division Attribute transpositions is mutually simultaneously stopped division.
S5：Step S4 is repeated 100 times, the RandomForest classifier of 100 decision trees composition is established；
S6：For the data being newly collected into, it is input in the RandomForest classifier of above step foundation, each
Prediction classification is carried out in CART decision tree, selects final classification in a manner of ballot as a result, the present embodiment uses relative majority
Ballot method, defining random forest grader prediction output described herein is 100 dimensional vectorsI is indicated
The corresponding classification of prediction result, then voting rule can be described asI.e. prediction result is that who gets the most votes is defeated
Classification out.
Embodiment in being described above is only a part of the embodiments of the present invention, the claimed range of the present invention and not only
It is limited only to above-mentioned specific embodiment, without creative efforts, obtains the side substantially identical with the present invention
Case also belongs to the scope of the present invention.
Claims (5)
1. a kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest, it is characterised in that
It comprises the steps of：
S1：Ultrasonic transducer probe is mounted on to the position of distance water suction bottom fixed height, uses ultrasonic transducer
And signal acquiring system, echo-signal of the ultrasonic wave after water suction bottom water sludge interface reflection is obtained, is received under different operating conditions
Collect p group echo-signal, p meets 50≤p≤500；
S2：9 layers of wavelet decomposition are carried out for ultrasonic echo signal, each layer wavelet coefficient is obtained, is selected using Hesusure threshold value
It takes method to carry out noise reduction process, carries out coefficient processing using soft-threshold function, wavelet coefficient carries out signal weight using treated
Structure improves Signal-to-Noise；
S3：The dyadic wavelet for carrying out Mallat to the signal after noise reduction decomposes, and is decomposed into 6 grades of scales, selects 22、23、24、25Scale
Space is as feature extraction space, respectively in every layer of 4 modulus maximum of extraction and the corresponding time shaft position of the modulus maximum, group
At 32 dimensional feature vector F=[V1, T1, V2, T2...,V16, T16]；
S4：Using the feature vector obtained in S3 alternatively attribute, 4 dimensional features are therefrom randomly selected as single base learner
Feature input, the sample input of sample as single decision tree is randomly selected from sample of signal obtained in S1, with
Gini impurity level come select divide attribute, establish CART decision tree；
S5：Step S4 is repeated into ncIt is secondary, establish ncThe RandomForest classifier of decision tree composition；
S6：For the data being newly collected into, it is input in the RandomForest classifier of above step foundation, at each
Prediction classification is carried out in CART decision tree, and final classification result is selected in a manner of ballot.
2. the Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification according to claim 1 based on random forest
Method, it is characterised in that：By ultrasonic transducer probe installation, fixed height position, the selected probe angle of departure are S1 under water
6 °, probe is h apart from bottom of pond height, and historical high mud position is hm, h meets h-hm>=30cm, the point centered on probe, probe face
For horizontal plane, to being measurement space down toward bottom of pond, in the cylindrical space that radius is h*tan3 ° of+h ', h ' is amount of redundancy, is met
3. the Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification according to claim 1 based on random forest
Method, it is characterised in that：9 layers of wavelet decomposition are carried out to original signal using wavelet function in S2, each layer wavelet coefficient is obtained, makes
It is calculated first with Hesusure threshold value selection rule in conjunction with fixed threshold rule and unbiased possibility predication adaptive threshold rule
Threshold value as defined in fixed threshold rule outσ is the standard deviation of noise, and N is the length of signal, then calculates nothing
Partial likelihood estimation self-adaptive threshold value Rigrsure threshold value, wavelet coefficient square vector X=[x1, x2..., xn], wherein x1≤x2
≤…≤xn, n is the number of this layer of wavelet coefficient, if a risk vector is：S=[s1, s2..., si..., sn], in risk vector
Each element is：Minimum value s is found out from risk vectorminAs value-at-risk,
Find out the corresponding x of minimum risk value subscript coefficientmin, then adaptive threshold be：It is selected in conjunction with above two threshold value
Rule is taken, if K is certain layer of wavelet coefficient quadratic sum after wavelet decomposition, defined parametersParameter
Mixed type threshold value λhValue rule isAfter threshold value, select soft-threshold function to small
Wave system number is handled, and soft-threshold function is：
Wherein xnFor through soft-threshold function treated new wavelet coefficient.
4. the Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification according to claim 1 based on random forest
Method, it is characterised in that：It selects Gini impurity level as attribute selection rule is divided in S4, CART decision tree is established, from sample
That puts back in data randomly selects pxA sample data is inputted as single decision tree data, is put back to from 32 dimensional feature vectors
4 dimensions of randomly selecting divide attribute as decision tree, Gini impurity level is defined as：
Wherein y is sample class quantity, pkFor kth class sample proportion, wherein the Gini impurity level of attribute Z is defined as：
For the 4 dimensional feature values extracted, every dimension includes pxA sample data, there is px+ 1 can division points, define candidate divide
It puts and isDefinitionWherein [z1, z2..., zi..., zj] it is in attribute
J different values on Z, j is single decision tree sample size, according to formula
Optimal dividing attribute and two optimal dividing points are selected, so that the data Gini impurity level after dividing accordingly is minimum.
5. the Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification according to claim 1 based on random forest
Method, it is characterised in that：Pb-Zn deposits absorbing well, absorption well water sludge interface echo-signal is divided into three kinds of classifications, respectively correspond three kinds it is different
Water sludge interface situation, respectively single echo type, double echo type and three echo types.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810453235.4A CN108896996B (en) | 2018-05-11 | 2018-05-11 | A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810453235.4A CN108896996B (en) | 2018-05-11 | 2018-05-11 | A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108896996A true CN108896996A (en) | 2018-11-27 |
| CN108896996B CN108896996B (en) | 2019-09-20 |
Family
ID=64343227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810453235.4A Active CN108896996B (en) | 2018-05-11 | 2018-05-11 | A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108896996B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126622A (en) * | 2019-12-19 | 2020-05-08 | 中国银联股份有限公司 | Data anomaly detection method and device |
| CN112697887A (en) * | 2020-12-08 | 2021-04-23 | 江苏科技大学 | Ultrasonic detection defect qualitative identification method based on neural network |
| CN113378473A (en) * | 2021-06-23 | 2021-09-10 | 中国地质科学院水文地质环境地质研究所 | Underground water arsenic risk prediction method based on machine learning model |
| CN116660389A (en) * | 2023-07-21 | 2023-08-29 | 山东大禹水务建设集团有限公司 | River sediment detection and repair system based on artificial intelligence |
| CN118410415A (en) * | 2024-06-18 | 2024-07-30 | 安徽大学 | Power system fault diagnosis method based on MP-Convformer parallel network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7177808B2 (en) * | 2000-11-29 | 2007-02-13 | The United States Of America As Represented By The Secretary Of The Air Force | Method for improving speaker identification by determining usable speech |
| CN104215935A (en) * | 2014-08-12 | 2014-12-17 | 电子科技大学 | Weighted decision fusion based radar cannonball target recognition method |
| CN104348741A (en) * | 2013-08-06 | 2015-02-11 | 南京理工大学常熟研究院有限公司 | Method and system for detecting P2P (peer-to-peer) traffic based on multi-dimensional analysis and decision tree |
| CN106529416A (en) * | 2016-10-18 | 2017-03-22 | 国网山东省电力公司电力科学研究院 | Electric-power line detection method and system based on millimeter wave radar decision tree classification |
| CN106990018A (en) * | 2017-02-28 | 2017-07-28 | 河海大学 | A kind of prestressed concrete beam Grouted density intelligent identification Method |
| CN107180140A (en) * | 2017-06-08 | 2017-09-19 | 中南大学 | Shafting fault recognition method based on dual-tree complex wavelet and AdaBoost |
| CN107292335A (en) * | 2017-06-06 | 2017-10-24 | 云南电网有限责任公司信息中心 | A kind of transmission line of electricity cloud data automatic classification method based on Random Forest model |
-
2018
- 2018-05-11 CN CN201810453235.4A patent/CN108896996B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7177808B2 (en) * | 2000-11-29 | 2007-02-13 | The United States Of America As Represented By The Secretary Of The Air Force | Method for improving speaker identification by determining usable speech |
| CN104348741A (en) * | 2013-08-06 | 2015-02-11 | 南京理工大学常熟研究院有限公司 | Method and system for detecting P2P (peer-to-peer) traffic based on multi-dimensional analysis and decision tree |
| CN104215935A (en) * | 2014-08-12 | 2014-12-17 | 电子科技大学 | Weighted decision fusion based radar cannonball target recognition method |
| CN106529416A (en) * | 2016-10-18 | 2017-03-22 | 国网山东省电力公司电力科学研究院 | Electric-power line detection method and system based on millimeter wave radar decision tree classification |
| CN106990018A (en) * | 2017-02-28 | 2017-07-28 | 河海大学 | A kind of prestressed concrete beam Grouted density intelligent identification Method |
| CN107292335A (en) * | 2017-06-06 | 2017-10-24 | 云南电网有限责任公司信息中心 | A kind of transmission line of electricity cloud data automatic classification method based on Random Forest model |
| CN107180140A (en) * | 2017-06-08 | 2017-09-19 | 中南大学 | Shafting fault recognition method based on dual-tree complex wavelet and AdaBoost |
Non-Patent Citations (1)
| Title |
|---|
| 李璐等: "基于随机森林的铝铸件内部缺陷类型识别研究", 《研究与开发》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126622A (en) * | 2019-12-19 | 2020-05-08 | 中国银联股份有限公司 | Data anomaly detection method and device |
| CN111126622B (en) * | 2019-12-19 | 2023-11-03 | 中国银联股份有限公司 | Data anomaly detection method and device |
| CN112697887A (en) * | 2020-12-08 | 2021-04-23 | 江苏科技大学 | Ultrasonic detection defect qualitative identification method based on neural network |
| CN113378473A (en) * | 2021-06-23 | 2021-09-10 | 中国地质科学院水文地质环境地质研究所 | Underground water arsenic risk prediction method based on machine learning model |
| CN113378473B (en) * | 2021-06-23 | 2024-01-12 | 中国地质科学院水文地质环境地质研究所 | Groundwater arsenic risk prediction method based on machine learning model |
| CN116660389A (en) * | 2023-07-21 | 2023-08-29 | 山东大禹水务建设集团有限公司 | River sediment detection and repair system based on artificial intelligence |
| CN116660389B (en) * | 2023-07-21 | 2023-10-13 | 山东大禹水务建设集团有限公司 | River sediment detection and repair system based on artificial intelligence |
| CN118410415A (en) * | 2024-06-18 | 2024-07-30 | 安徽大学 | Power system fault diagnosis method based on MP-Convformer parallel network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108896996B (en) | 2019-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108896996B (en) | A kind of Pb-Zn deposits absorbing well, absorption well water sludge interface ultrasonic echo signal classification method based on random forest | |
| CN108613645B (en) | A kind of Pb-Zn deposits absorbing well, absorption well surveying on sludge thickness method based on parameter Estimation | |
| Baidai et al. | Machine learning for characterizing tropical tuna aggregations under Drifting Fish Aggregating Devices (DFADs) from commercial echosounder buoys data | |
| CN108985304B (en) | Automatic sedimentary layer structure extraction method based on shallow profile data | |
| CN117198330B (en) | Sound source identification method and system and electronic equipment | |
| CN107064894A (en) | A kind of clutter suppression method based on deep learning | |
| CN111738332A (en) | Underwater multi-source acoustic image substrate classification method and system based on feature level fusion | |
| CN115774248A (en) | Sea surface small target detection method and system based on multiple characteristics of graph | |
| CN115169733A (en) | Deep learning-based method for predicting resuspension quantity of deep sea sediments by using internal solitary waves | |
| CN111738278B (en) | Underwater multi-source acoustic image feature extraction method and system | |
| CN103886606A (en) | SAR image segmentation method based on joint generalized gamma distribution parameters | |
| CN116187168A (en) | Method for improving water depth inversion accuracy based on neural network-gravity information wavelet decomposition | |
| Sun et al. | Probabilistic neural network based seabed sediment recognition method for side-scan sonar imagery | |
| CN113064133B (en) | Sea surface small target feature detection method based on time-frequency domain depth network | |
| CN104268397A (en) | Underwater acoustic target identifying method based on ship-radiated noise optimization features | |
| Baidai et al. | Recent advances on the use of supervised learning algorithms for detecting tuna aggregations under fads from echosounder buoys data | |
| CN113705605B (en) | Automatic cleaning method for abnormal values of multi-beam sounding data through partial manual intervention | |
| CN109269480A (en) | The multibeam bathymetric data processing method of robust curved surface is assumed based on optimal more depth of waters | |
| TANG et al. | Application of LVQ neural network combined with the genetic algorithm in acoustic seafloor classification | |
| CN113221651A (en) | Seafloor sediment classification method using acoustic propagation data and unsupervised machine learning | |
| Preston et al. | Acoustic classification of submerged aquatic vegetation | |
| CN107942313A (en) | Sea clutter optimal soft survey instrument and method based on TSP question drosophila optimization algorithm optimization wavelet neural network | |
| CN104268394A (en) | Underwater acoustic target identifying method based on ship-radiated noise time domain features | |
| Sobradillo et al. | Target strength measurements of yellowfin tuna (Thunnus albacares) and acoustic discrimination of three tropical tuna species | |
| CN116819540B (en) | Method for intelligently calculating type and depth of fishing group |
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 |