CN109270010A - A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm - Google Patents
A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm Download PDFInfo
- Publication number
- CN109270010A CN109270010A CN201811362049.6A CN201811362049A CN109270010A CN 109270010 A CN109270010 A CN 109270010A CN 201811362049 A CN201811362049 A CN 201811362049A CN 109270010 A CN109270010 A CN 109270010A
- Authority
- CN
- China
- Prior art keywords
- dust
- spectral reflectivity
- plant
- laying
- data
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/55—Specular reflectivity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N2021/8466—Investigation of vegetal material, e.g. leaves, plants, fruits
Landscapes
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The plant regulation of absorbing dust capability detection method based on random forests algorithm that the present invention relates to a kind of, comprising: obtain plant in the single order inverse of the spectral reflectivity of each wave band and the amount of the laying the dust data of corresponding characterization plant regulation of absorbing dust capability；Single order inverse and the corresponding amount of laying the dust data according to plant in the spectral reflectivity of each wave band, using the first random forests algorithm, screening obtains wavelength ranges；According to the wavelength ranges of screening, the corresponding spectral reflectivity single order inverse of wavelength ranges and the corresponding amount of laying the dust data are obtained；The inverse model that lays the dust is constructed using grid-search algorithm and the second random forests algorithm；Spectral reflectivity single order inverse to measuring plants in wavelength ranges is input in the inverse model that lays the dust, the amount of laying the dust data are obtained.Wavelength ranges are quickly and accurately filtered out using the realization of the first random forests algorithm, in conjunction with grid-search algorithm and the second random forests algorithm, establish the inverse model that lays the dust, and then realize and accurately estimate the plant amount of laying the dust.
Description
Technical field
The present invention relates to Environmental Monitoring and Assessment technical fields, more particularly to a kind of plant based on random forests algorithm
Regulation of absorbing dust capability detection method.
Background technique
With the rapid development of industrialization and urbanization, urban atmospheric pollution is on the rise, and " gray haze " weather occurs again and again,
The physical condition etc. of normal production and living and human body to the mankind all causes serious influence." gray haze " refers to a large amount of non-
The particulate matters even suspensions such as normal tiny drying grit, soot, salt grain become in air so as to cause stuffy, visibility
The phenomenon of difference, and the exceeded of the particle contents such as dust is one of the main reason for causing " gray haze ".Therefore, how to effectively control
Dust pollution, which has become, improves the main problem that urban air-quality needs to solve.
Plant is a nature " weapon " for alleviating city gray haze, and canopy in great numbers plays the role of reducing wind speed, in air
The biggish dust of particle can with wind speed reduce and be deposited to leaves of plants table or ground, to generate dust retention.In addition,
The various characteristics on plant leaf blade surface can also promote the absorption of dust particles, and energy after the vanes rain drop erosion of dust covering
Again restore dust collection capacity.It can be seen that the dust retention of plant has become an important indicator of screening urban greenery plants.
Therefore, the amount of laying the dust for how assessing plant has become the emphasis of current research.
Summary of the invention
Based on this, the object of the present invention is to provide a kind of plant regulation of absorbing dust capability detection side based on random forests algorithm
Method has the advantages that quick and precisely to estimate the plant amount of laying the dust.
A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm, includes the following steps:
Plant is obtained the single order of the spectral reflectivity of each wave band is reciprocal and corresponding characterization plant regulation of absorbing dust capability
The amount of laying the dust data；
It is random using first according to the single order inverse of the spectral reflectivity of each wave band of plant and the corresponding amount of laying the dust data
Forest algorithm, screening obtain wavelength ranges；
According to the wavelength ranges of screening, it is reciprocal and corresponding lay the dust to obtain the corresponding spectral reflectivity single order of wavelength ranges
Measure data；
According to the corresponding spectral reflectivity single order inverse of wavelength ranges and the corresponding amount of laying the dust data, using grid-
Search algorithm and the second random forests algorithm construct the inverse model that lays the dust；
Spectral reflectivity single order inverse to measuring plants in wavelength ranges is input in the inverse model that lays the dust, is laid the dust
Measure data.
The present invention quickly and accurately filters out wavelength ranges using the realization of the first random forests algorithm, in conjunction with grid-
Search algorithm and the second random forests algorithm establish the inverse model that lays the dust, and then realize and accurately estimate the plant amount of laying the dust.
In one embodiment, the single order inverse of the spectral reflectivity according to plant in each wave band and corresponding stagnant
Dust quantity data screen the step of obtaining wavelength ranges using the first random forests algorithm, comprising:
Using the single order inverse of each band spectrum reflectivity and the corresponding amount of laying the dust data as input, building first is random gloomy
Woods model, and obtain according to first Random Forest model importance score of each wave band；
The importance score of each wave band is sequentially added up according to sequence from big to small, until cumulative importance score
Just stop when AccuFI meets AccuFI >=F cumulative；Wherein, F is preset importance threshold value；
Accumulated wave band is screened as wavelength ranges.
In one embodiment, the single order inverse of the spectral reflectivity with each wave band and the corresponding amount of laying the dust data
As input, the first Random Forest model is constructed, and is obtained according to the importance that first Random Forest model obtains each wave band
The step of dividing, comprising:
It regard the spectral reflectivity first derivative of all wave bands and the corresponding amount of laying the dust as sample set；Wherein, in sample set
Each sample standard deviation include each wave band spectral reflectivity single order is reciprocal and the corresponding amount of laying the dust data；
Sampling is used to have the method put back to randomly select n sample from sample set as n sample set；Wherein, not by
The sample of extraction is known as the outer data of bag；
To each sample set, original classification tree is constructed, every one tree is all an independent Taxonomy and distribution,
The relationship between independent variable and dependent variable is established in each node of every one tree；
Select modeling factors at random at each node and carry out building fractionation, to select optimal case, and then construct with
Machine forest model, and seek the prediction result of each tree；
In the case where keeping the single order of spectral reflectivity of its all band reciprocal constant, upset i-th of wave band in sample
Spectral reflectivity single order inverse sequence, the single order for then calculating the spectral reflectivity of i-th of wave band of certain one tree again falls
Number is disturbed the difference of the prediction error of the outer data of bag of sequence front and back；It calculates each tree and upsets data outside the bag of sequence front and back
Predict that the average value of the difference of error obtains the importance score of each wave band with this as the importance score of i-th of wave band.
In one embodiment, described that the inverting that lays the dust is constructed using grid-search algorithm and the second random forests algorithm
The step of model, comprising:
Using the spectral reflectivity first derivative of wavelength ranges and the corresponding amount of laying the dust data as sample set；Wherein, sample
The each sample concentrated respectively include the spectral reflectivity of one of wavelength ranges single order is reciprocal and the corresponding amount of laying the dust
Data.
Sampling is used to have the method put back to randomly select n sample from sample set as n sample set；
To each sample set, original classification tree is constructed, every one tree is all an independent Taxonomy and distribution,
The relationship between independent variable and dependent variable is established in each node of every one tree；
It is random gloomy to find second using grid-search algorithm for parameter setting range based on the second random forests algorithm
The optimized parameter of woods algorithm, each node obtain the inverse model that lays the dust after carrying out building fractionation according to the optimized parameter.
In one embodiment, the step of the single order inverse of spectral reflectivity of the acquisition plant in each wave band, packet
It includes:
The spectral reflectivity of plant is obtained, and spectral reflectivity is pre-processed；
Spectrum transform is carried out to pretreated spectral reflectivity, the single order for obtaining the spectral reflectivity of each wave band falls
Number.
It is described that pretreated step is carried out to spectral reflectivity, comprising:
The spectral reflectivity of plant is obtained, and rejects the spectral reflectivity that error is more than the first given threshold, will be remained later
Remaining spectral reflectance data is averaging, using average value as the spectral reflectivity of plant；
Actual spectrum of the nearby principle divided by corresponding blank value, as plant is used to the spectral reflectivity of the plant
Reflectivity；
According to the actual spectrum reflectivity of plant, water vapor absorption wave band is obtained, and rejects the spectrum under water vapor absorption wave band
Reflectivity, the curve of spectrum fitting completion further along spectral reflectivity obtain pretreated spectral reflectivity.
The error of DATA REASONING can be reduced by way of rejecting error information and averaging again；By using nearby principle
Divided by the mode of corresponding blank value, the data obtained under different time sections, different experimental conditions can be made to be comparable, and same
When eliminate experimental situation background brought by error；Data processing by rejecting water vapor absorption wave band, after can facilitating.
The plant regulation of absorbing dust capability detection device based on random forests algorithm that the present invention also provides a kind of, comprising:
Data acquisition module, for obtaining plant in the single order inverse of the spectral reflectivity of each wave band and corresponding table
Levy the amount of the laying the dust data of plant regulation of absorbing dust capability；
Wavelength ranges screening module, for according to plant in the single order inverse of the spectral reflectivity of each wave band and corresponding
The amount of laying the dust data, using the first random forests algorithm, screening obtains wavelength ranges；
Spectral reflectivity single order inverse obtains module, and for the wavelength ranges according to screening, it is corresponding to obtain wavelength ranges
Spectral reflectivity single order inverse and the corresponding amount of laying the dust data；
The inverse model that lays the dust constructs module, for reciprocal and corresponding according to the corresponding spectral reflectivity single order of wavelength ranges
The amount of laying the dust data, laid the dust inverse model using grid-search algorithm and the building of the second random forests algorithm；
The amount of laying the dust data acquisition module, for the spectral reflectivity single order inverse to measuring plants in wavelength ranges to be input to
It lays the dust in inverse model, obtains the amount of laying the dust data.
The present invention quickly and accurately filters out wavelength ranges using the realization of the first random forests algorithm, in conjunction with grid-
Search algorithm and the second random forests algorithm establish the inverse model that lays the dust, and then realize and accurately estimate the plant amount of laying the dust.
The present invention also provides a kind of computer readable storage mediums, store computer program thereon, the computer program
The plant regulation of absorbing dust capability detection method based on random forests algorithm as described in above-mentioned any one is realized when being executed by processor
The step of.
The present invention also provides a kind of computer equipment, including reservoir, processor and it is stored in the reservoir simultaneously
The computer program that can be executed by the processor, the processor are realized when executing the computer program as above-mentioned any one
The step of plant regulation of absorbing dust capability detection method based on random forests algorithm described in item.
In order to better understand and implement, the invention will now be described in detail with reference to the accompanying drawings.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the plant regulation of absorbing dust capability detection method of random forests algorithm；
Fig. 2 is that the present invention obtains plant in the flow chart of the single order inverse of the spectral reflectivity of each wave band；
Fig. 3 is that the present invention carries out pretreated flow chart to spectral reflectivity；
Fig. 4 is to compare figure before and after the present invention rejects water vapor absorption wave band；
Fig. 5 is the flow chart of present invention screening wavelength ranges；
Fig. 6 is the flow chart that the present invention obtains each wave band importance score；
Fig. 7 is the flow chart that the present invention constructs the inverse model that lays the dust.
Specific embodiment
Specification of the invention, claims and term " first " in attached drawing, " second ", " third " " the 4th " etc.
(if present) is to be used to distinguish similar objects, without being used to describe a particular order or precedence order.In addition, term
" comprising " and " having " and their any deformation, it is intended that cover not exclusively include, for example, containing a series of steps
Rapid or unit process, method, system, product or equipment those of are not necessarily limited to be clearly listed step or unit, but can
Including other step or units being not clearly listed or intrinsic for these process, methods, product or equipment.
Referring to Fig. 1, it is the present invention is based on the flow charts of the plant regulation of absorbing dust capability detection method of random forests algorithm.Institute
The plant regulation of absorbing dust capability detection method based on random forests algorithm is stated, is included the following steps:
Step S1: it obtains plant and lays the dust in the single order inverse of the spectral reflectivity of each wave band and corresponding characterization plant
The amount of the laying the dust data of ability.
Wherein, the spectral reflectivity of the plant is plant in the luminous flux of the reflection of each wave band and is incident on object
The ratio between luminous flux.The spectral reflectivity of the specific plant described in the present embodiment is that the spectral reflectivity of the plant is the plant
Object is in 350nm-2500nm wave band and is divided into the ratio between the reflected flux of 1nm wave band and the incident flux of the wave band, that is, includes institute
Plant is stated in the ratio between the reflected flux of 350nm and the incident flux of the wave band, the plant in the reflected flux of 351nm and the wave
The plant is in the ratio between the reflected flux of 2500nm and the incident flux of the wave band for the ratio between the incident flux ... of section.
Step S2: it according to the single order inverse of the spectral reflectivity of each wave band of plant and the corresponding amount of laying the dust data, utilizes
First random forests algorithm, screening obtain wavelength ranges.
Step S3: according to the wavelength ranges of screening, it is reciprocal and right to obtain the corresponding spectral reflectivity single order of wavelength ranges
The amount of the laying the dust data answered.
Step S4: it according to the corresponding spectral reflectivity single order inverse of wavelength ranges and the corresponding amount of laying the dust data, uses
Grid-search algorithm and the second random forests algorithm construct the inverse model that lays the dust；
Step S5: the spectral reflectivity single order inverse to measuring plants in wavelength ranges is input in the inverse model that lays the dust,
The acquisition amount of laying the dust data.
The present invention quickly and accurately filters out wavelength ranges using the realization of the first random forests algorithm, in conjunction with grid-
Search algorithm and the second random forests algorithm establish the inverse model that lays the dust, and then realize and accurately estimate the plant amount of laying the dust.
Referring to Fig. 2, it obtains plant in the flow chart of the single order inverse of the spectral reflectivity of each wave band for the present invention.
In one embodiment, in step S1, the single order of spectral reflectivity of the acquisition plant in each wave band is reciprocal
The step of, comprising:
Step S11: the spectral reflectivity of plant is obtained, and spectral reflectivity is pre-processed.
Step S12: spectrum transform is carried out to pretreated spectral reflectivity, obtains the spectral reflectivity of each wave band
Single order is reciprocal.
Wherein, by carrying out spectrum transform to pretreated spectral reflectivity, compressible ambient noise is to target information
Influence, improve follow-up data processing accuracy.
Referring to Fig. 3, it carries out pretreated flow chart to spectral reflectivity for the present invention.
In one embodiment, pretreated step is carried out to spectral reflectivity described in step S11, comprising:
Step S111: obtaining the spectral reflectivity of plant, and rejects the spectral reflectivity that error is more than the first given threshold,
Remaining spectral reflectance data is averaging later, using average value as the spectral reflectivity of plant.
In one embodiment, the blade in the multiple orientation of single plant is chosen for a kind of plant, every leaf is chosen solid
Fixed certain point is as measurement point, 5 groups of spectral reflectivities of each measurement point duplicate measurements, then rejects in this 5 groups of spectral reflectivities
It is averaging after the value in the first given threshold, then by remaining spectral reflectivity, the light of this piece leaf is characterized with average value
Reflectivity is composed, and then reduces the error of DATA REASONING, improves the accuracy of DATA REASONING.Multiple orientation therein can plant to be single
The orientation such as the upper and lower, left and right of strain.First given threshold is 5 groups of spectral reflectivities according to measurement, selected by experience
For rejects with other data with large error data threshold value.
In one embodiment, the spectral reflectivity of plant is obtained using the measurement of 3 spectrometer of ASD Field-Spec, by
It is .asp format in the data format of measurement, the data of this format read and handle, therefore, measuring with being unable to direct convenience
Before carrying out data processing after the spectral reflectivity of acquisition plant, View Spectral Pro software spectral reflectivity is also utilized
Data Format Transform be .txt format.
Step S112: use nearby principle divided by corresponding blank value the spectral reflectivity of the plant, as plant
Actual spectrum reflectivity.
To be comparable the data obtained under different time sections, different experimental conditions, and experimental situation is eliminated simultaneously
Error brought by background, section obtains different blank values to the application in different times, hence in some period
The spectral reflectivity of the plant of acquisition then needs for the spectral reflectivity of the plant to be in the blank value of the period, as place
The spectral reflectivity of plant after reason.
Step S113: it according to the actual spectrum reflectivity of plant, determines water vapor absorption wave band, and rejects water vapor absorption wave pair
The spectral reflectivity answered, the curve of spectrum fitting completion further along spectral reflectivity obtain pretreated spectral reflectivity.
Due to determining instrument of spectral reflectivity itself, the spectral reflectivity determined can be due to the absorption of steam
It is affected to the spectral reflectivity of plant, and the wavelength band studying also without too big meaning, therefore to vegetation spectrum
Data processing by rejecting water vapor absorption wave band, after can facilitating.Referring to Fig. 4, it rejects water vapor absorption wave for the present invention
Compare figure in the front and back of section.Specifically, the actual spectrum reflectivity data of plant is fitted to spectral reflectivity curve, pass through observation
Are there is abnormal wave band as water vapor absorption wave band by spectral reflectivity curve in data in spectral reflectivity curve；Again by the water
Spectral reflectivity in vapour absorption bands is rejected, and the curve of spectrum fitting completion acquisition further along spectral reflectivity is pretreated
Spectral reflectivity.
In one embodiment, spectrum transform is carried out to pretreated spectral reflectivity described in step S22, obtained each
The calculation of the single order inverse of the spectral reflectivity of a wave band are as follows:
Wherein, λi+1, λi, λi-1For adjacent wavelength, dR (λi) it is wavelength XiFirst derivative spectrum, R (λi+1), R (λi), R
(λi-1) be respectively wavelength be λi+1, λi, λi-1The reflectivity at place.
In one embodiment, the calculation of the amount of the laying the dust data of the plant regulation of absorbing dust capability of acquisition characterization described in step S1
Are as follows:
D=Δ W/S
In above-mentioned formula, D is the amount of the laying the dust data for characterizing plant regulation of absorbing dust capability；Δ W is a period of time implants blade
The amount of laying the dust, S are blade area.
Wherein, the amount of the laying the dust data of each blade of plant are consistent, i.e., the spectrum of each wave band is anti-in each blade
The single order corresponding amount of the laying the dust data reciprocal for penetrating rate are identical, it can think the corresponding sample of each blade, each sample
The single order inverse of the spectral reflectivity of each wave band including the blade and the corresponding amount of laying the dust data.
Referring to Fig. 5, its flow chart for present invention screening wavelength ranges.
In one embodiment, the single order inverse of the spectral reflectivity according to plant in each wave band and corresponding stagnant
Dust quantity data screen the step of obtaining wavelength ranges using the first random forests algorithm, comprising:
Step S21: using the single order of the spectral reflectivity of each wave band is reciprocal and the corresponding amount of laying the dust data are as inputting,
The first Random Forest model is constructed, and obtains the importance score of each wave band according to first Random Forest model；
Step S22: the importance score of each wave band is sequentially added up according to sequence from big to small, until cumulative is important
Property score A ccuFI just stop when meeting AccuFI >=F it is cumulative；Wherein, F is preset importance threshold value；
Step S23: accumulated wave band is screened as wavelength ranges.
Please referring to Fig. 6 is the flow chart that the present invention obtains each wave band importance score.
In one embodiment, reciprocal and corresponding stagnant with the single order of the spectral reflectivity of each wave band in step S21
Dust quantity data construct the first Random Forest model as input, and obtain each wave band according to first Random Forest model
The step of importance score, comprising:
Step S211: the spectral reflectivity first derivative of all wave bands and the corresponding amount of laying the dust are regard as sample set；Wherein,
Each sample standard deviation in sample set includes that the single order of the spectral reflectivity of each wave band is reciprocal and the corresponding amount of laying the dust data；
Step S212: sampling is used to have the method put back to randomly select n sample from sample set as n sample set；
Wherein, the sample not being extracted is known as the outer data of bag；
Step S213: to each sample set, original classification tree is constructed, every one tree is all an independent classification
With regression tree, the relationship between independent variable and dependent variable is established in each node of every one tree；
Step S214: modeling factors are selected at random at each node and carry out building fractionation, to select optimal case, in turn
Random Forest model is constructed, and seeks the prediction result of each tree；
Step S215: in the case where keeping the single order of spectral reflectivity of its all band reciprocal constant, upset in sample
The sequence of the single order inverse of the spectral reflectivity of i-th of wave band, then calculates the spectral reflectivity of i-th of wave band of certain one tree again
Single order inverse be disturbed sequence front and back the outer data of bag prediction error difference；Calculate the bag that each tree upsets sequence front and back
The average value of the difference of the prediction error of outer data obtains the important of each wave band as the importance score of i-th of wave band with this
Property score.
Referring to Fig. 7, its flow chart for constructing the inverse model that lays the dust for the present invention.
In one embodiment, described to be constructed using grid-search algorithm and the second random forests algorithm in step S4
Lay the dust inverse model the step of, comprising:
Step S41: using the spectral reflectivity first derivative of wavelength ranges and the corresponding amount of laying the dust data as sample set；Its
In, the single order that each sample in sample set respectively includes the spectral reflectivity of one of wavelength ranges is reciprocal and corresponding
The amount of laying the dust data.
Step S42: sampling is used to have the method put back to randomly select n sample from sample set as n sample set；
Step S43: to each sample set, original classification tree is constructed, every one tree is all an independent classification
With regression tree, the relationship between independent variable and dependent variable is established in each node of every one tree；
Step S44: the parameter setting range based on the second random forests algorithm finds the using grid-search algorithm
The optimized parameter of two random forests algorithms, each node obtain the inverting mould that lays the dust after carrying out building fractionation according to the optimized parameter
Type.
Wherein grid-search algorithm uses the means of exhaustive search, i.e., concentrates in all candidate parameters, by following
Ring traversal, attempts each possibility, the parameter to behave oneself best is exactly final result.It lays the dust instead in the application for rapid build
Model is drilled, according to the optimized parameter for the second random forests algorithm that grid-search algorithm obtains specifically: the number of iterations n_
The candidate parameter collection of estimators are as follows: 20,40,60,80,100,120,140,160,180,200.
For the precision for verifying inverse model, the actual measured value for the amount of laying the dust is compared with predicted value.The precision of model
Verifying is measured using quantitative accuracy evaluation index, specially determines coefficient (the coefficient of
Determination, R2), root-mean-square error (the root-mean-square error, RMSE), mean absolute error
(the mean absolute error, MAE), average deviation (Bias), calculation formula is as follows.
Wherein, xiFor the actual measurement amount of laying the dust of i-th of sample, xi' for i-th of sample model prediction the amount of laying the dust,For institute
There is the average value of the actual measurement amount of laying the dust,For the average value of all model prediction amounts of laying the dust, N is the quantity of sample.
R2Indicate the linearly related degree between two data, MAE and RMSE are used to the integral level of evaluated error, and Bias is anti-
Estimation data and the departure degree of measured data numerically are reflected.
The result shows that the verifying precision RMSE of the amount of laying the dust appraising model is respectively less than 0.3g/m2, efficiency of inverse process is good.
In one embodiment, after the step of amount of laying the dust data are obtained described in step S5, further includes: obtain multiple groups and wait for
The amount of the laying the dust data of measuring plants, and corresponding plant is planted in setting regions according to the amount of the laying the dust data.Specifically, according to stagnant
Dust quantity data are bigger, illustrate that the regulation of absorbing dust capability to measuring plants is stronger, and such as there is the area of sandstorm can plant according to setting regions
Regulation of absorbing dust capability is stronger to measuring plants.
The plant regulation of absorbing dust capability detection device based on random forests algorithm that the present invention also provides a kind of, comprising:
Data acquisition module, for obtaining plant in the single order inverse of the spectral reflectivity of each wave band and corresponding table
Levy the amount of the laying the dust data of plant regulation of absorbing dust capability；
Wavelength ranges screening module, for according to plant in the single order inverse of the spectral reflectivity of each wave band and corresponding
The amount of laying the dust data, using the first random forests algorithm, screening obtains wavelength ranges；
Spectral reflectivity single order inverse obtains module, and for the wavelength ranges according to screening, it is corresponding to obtain wavelength ranges
Spectral reflectivity single order inverse and the corresponding amount of laying the dust data；
The inverse model that lays the dust constructs module, for reciprocal and corresponding according to the corresponding spectral reflectivity single order of wavelength ranges
The amount of laying the dust data, laid the dust inverse model using grid-search algorithm and the building of the second random forests algorithm；
The amount of laying the dust data acquisition module, for the spectral reflectivity single order inverse to measuring plants in wavelength ranges to be input to
It lays the dust in inverse model, obtains the amount of laying the dust data.
The present invention quickly and accurately filters out wavelength ranges using the realization of the first random forests algorithm, in conjunction with grid-
Search algorithm and the second random forests algorithm establish the inverse model that lays the dust, and then realize and accurately estimate the plant amount of laying the dust.
In one embodiment, the data acquisition module includes:
Preprocessing module is pre-processed for obtaining the spectral reflectivity of plant, and to spectral reflectivity.
Spectrum transform module obtains the light of each wave band for carrying out spectrum transform to pretreated spectral reflectivity
The single order for composing reflectivity is reciprocal.
Wherein, the preprocessing module includes:
Mean value calculation module, for obtaining the spectral reflectivity of plant, and rejecting error is more than the first given threshold
Remaining spectral reflectance data is averaging by spectral reflectivity later, using average value as the spectral reflectivity of plant.
Actual spectrum reflectivity computing module, for the spectral reflectivity to the plant using nearby principle divided by correspondence
Blank value, the actual spectrum reflectivity as plant.
Water vapor absorption wave band rejects module, for the actual spectrum reflectivity according to plant, obtains water vapor absorption wave band, and
Reject water vapor absorption wave band under spectral reflectivity, further along spectral reflectivity the curve of spectrum fitting completion pre-processed after
Spectral reflectivity.
In one embodiment, described that spectrum transform is carried out to pretreated spectral reflectivity, obtain each wave band
The calculation of the single order inverse of spectral reflectivity are as follows:
Wherein, λi+1, λi, λi-1For adjacent wavelength, dR (λi) it is wavelength XiFirst derivative spectrum, R (λi+1), R (λi), R
(λi-1) be respectively wavelength be λi+1, λi, λi-1The reflectivity at place.
In one embodiment, the calculation of the amount of the laying the dust data for obtaining characterization plant regulation of absorbing dust capability are as follows:
D=Δ W/S
In above-mentioned formula, D is the amount of the laying the dust data for characterizing plant regulation of absorbing dust capability；Δ W is a period of time implants blade
Regulation of absorbing dust capability, S are blade area.
Wherein, the amount of the laying the dust data of each blade of plant are consistent, i.e., the spectrum of each wave band is anti-in each blade
The single order corresponding amount of the laying the dust data reciprocal for penetrating rate are identical, it can think the corresponding sample of each blade, each sample
The single order inverse of the spectral reflectivity of each wave band including the blade and the corresponding amount of laying the dust data.
The present invention also provides a kind of computer readable storage mediums, store computer program thereon, the computer program
The plant regulation of absorbing dust capability detection method based on random forests algorithm as described in above-mentioned any one is realized when being executed by processor
The step of.
It wherein includes storage medium (the including but not limited to disk of program code that the present invention, which can be used in one or more,
Memory, CD-ROM, optical memory etc.) on the form of computer program product implemented.Computer-readable storage media packet
Permanent and non-permanent, removable and non-removable media is included, can be accomplished by any method or technique information storage.Letter
Breath can be computer readable instructions, data structure, the module of program or other data.The example packet of the storage medium of computer
Include but be not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM),
Other kinds of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory
(EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM), digital versatile disc
(DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices or any other non-biography
Defeated medium, can be used for storage can be accessed by a computing device information.
The present invention also provides a kind of computer equipment, including reservoir, processor and it is stored in the reservoir simultaneously
The computer program that can be executed by the processor, the processor are realized when executing the computer program as above-mentioned any one
The step of plant regulation of absorbing dust capability detection method based on random forests algorithm described in item.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.
Claims (10)
1. a kind of plant regulation of absorbing dust capability detection method based on random forests algorithm, which comprises the steps of:
Plant is obtained in the single order inverse of the spectral reflectivity of each wave band and laying the dust for corresponding characterization plant regulation of absorbing dust capability
Measure data；
According to the single order inverse of the spectral reflectivity of each wave band of plant and the corresponding amount of laying the dust data, the first random forest is utilized
Algorithm, screening obtain wavelength ranges；
According to the wavelength ranges of screening, the corresponding spectral reflectivity single order inverse of wavelength ranges and the corresponding amount of laying the dust number are obtained
According to；
According to the corresponding spectral reflectivity single order inverse of wavelength ranges and the corresponding amount of laying the dust data, using grid-search
Algorithm and the second random forests algorithm construct the inverse model that lays the dust；
Spectral reflectivity single order inverse to measuring plants in wavelength ranges is input in the inverse model that lays the dust, the amount of laying the dust number is obtained
According to.
2. the plant regulation of absorbing dust capability detection method according to claim 1 based on random forests algorithm, which is characterized in that institute
The single order inverse according to plant in the spectral reflectivity of each wave band and the corresponding amount of laying the dust data are stated, the first random forest is utilized
The step of algorithm, screening obtains wavelength ranges, comprising:
Using the single order inverse of each band spectrum reflectivity and the corresponding amount of laying the dust data as input, the first random forest mould is constructed
Type, and obtain according to first Random Forest model importance score of each wave band；
The importance score of each wave band is sequentially added up according to sequence from big to small, until cumulative importance score A ccuFI
Just stop when meeting AccuFI >=F cumulative；Wherein, F is preset importance threshold value；
Accumulated wave band is screened as wavelength ranges.
3. the plant regulation of absorbing dust capability detection method according to claim 2 based on random forests algorithm, which is characterized in that institute
It states using the single order inverse of the spectral reflectivity of each wave band and the corresponding amount of laying the dust data as input, constructs the first random forest
Model, and the step of importance score of each wave band is obtained according to first Random Forest model, comprising:
It regard the spectral reflectivity first derivative of all wave bands and the corresponding amount of laying the dust as sample set；Wherein, every in sample set
A sample standard deviation includes that the single order of the spectral reflectivity of each wave band is reciprocal and the corresponding amount of laying the dust data；
Sampling is used to have the method put back to randomly select n sample from sample set as n sample set；Wherein, it is not extracted
Sample be known as the outer data of bag；
To each sample set, original classification tree is constructed, every one tree is all an independent Taxonomy and distribution, every
The relationship between independent variable and dependent variable is established in each node of one tree；
It selects modeling factors at random at each node and carries out building fractionation, to select optimal case, and then construct random gloomy
Woods model, and seek the prediction result of each tree；
In the case where keeping the single order of spectral reflectivity of its all band reciprocal constant, upset the light of i-th of wave band in sample
The sequence for composing the single order inverse of reflectivity, then calculates the single order inverse quilt of the spectral reflectivity of i-th of wave band of certain one tree again
Upset the difference of the prediction error of the outer data of bag of sequence front and back；Calculate the prediction that each tree upsets the outer data of bag of sequence front and back
The average value of the difference of error obtains the importance score of each wave band with this as the importance score of i-th of wave band.
4. the plant regulation of absorbing dust capability detection method according to claim 1 based on random forests algorithm, which is characterized in that institute
It states and the step of laying the dust inverse model is constructed using grid-search algorithm and the second random forests algorithm, comprising:
Using the spectral reflectivity first derivative of wavelength ranges and the corresponding amount of laying the dust data as sample set；Wherein, in sample set
Each sample spectral reflectivity for respectively including one of wavelength ranges single order is reciprocal and the corresponding amount of laying the dust data.
Sampling is used to have the method put back to randomly select n sample from sample set as n sample set；
To each sample set, original classification tree is constructed, every one tree is all an independent Taxonomy and distribution, every
The relationship between independent variable and dependent variable is established in each node of one tree；
Parameter setting range based on the second random forests algorithm is found the second random forest using grid-search algorithm and is calculated
The optimized parameter of method, each node obtain the inverse model that lays the dust after carrying out building fractionation according to the optimized parameter.
5. the plant regulation of absorbing dust capability detection method according to claim 1 based on random forests algorithm, which is characterized in that institute
It states and obtains plant in the step of the single order inverse of the spectral reflectivity of each wave band, comprising:
The spectral reflectivity of plant is obtained, and spectral reflectivity is pre-processed；
Spectrum transform is carried out to pretreated spectral reflectivity, the single order for obtaining the spectral reflectivity of each wave band is reciprocal；
It is wherein, described that pretreated step is carried out to spectral reflectivity, comprising:
The spectral reflectivity of plant is obtained, and rejects the spectral reflectivity that error is more than the first given threshold, it later will be remaining
Spectral reflectance data is averaging, using average value as the spectral reflectivity of plant；
Use nearby principle divided by corresponding blank value the spectral reflectivity of the plant, the actual spectrum as plant reflects
Rate；
According to the actual spectrum reflectivity of plant, water vapor absorption wave band is obtained, and rejects the spectral reflectance under water vapor absorption wave band
Rate, the curve of spectrum fitting completion further along spectral reflectivity obtain pretreated spectral reflectivity.
6. the plant regulation of absorbing dust capability detection method according to claim 1 based on random forests algorithm, which is characterized in that
The calculation of the single order inverse of the spectral reflectivity of each wave band are as follows:
Wherein, λi+1, λi, λi-1For adjacent wavelength, dR (λi) it is wavelength XiFirst derivative spectrum, R (λi+1), R (λi), R (λi-1)
Be respectively wavelength be λi+1, λi, λi-1The reflectivity at place.
The calculation of the amount of the laying the dust data for obtaining characterization plant regulation of absorbing dust capability are as follows:
D=Δ W/S
Wherein, D is the amount of the laying the dust data for characterizing plant regulation of absorbing dust capability；Δ W is the amount of laying the dust of a period of time implants blade；S is
Blade area.
7. the plant regulation of absorbing dust capability detection method according to claim 1 based on random forests algorithm, which is characterized in that institute
After the step of stating the acquisition amount of laying the dust data, further includes: obtain multiple groups and wait for the amount of the laying the dust data of measuring plants, and laid the dust according to described
It measures data and plants corresponding plant in setting regions.
8. a kind of plant regulation of absorbing dust capability detection device based on random forests algorithm characterized by comprising
Data acquisition module is planted for obtaining plant in the single order inverse of the spectral reflectivity of each wave band and corresponding characterization
The amount of the laying the dust data of object regulation of absorbing dust capability；
Wavelength ranges screening module, in the single order inverse of the spectral reflectivity of each wave band and corresponding being laid the dust according to plant
Data are measured, using the first random forests algorithm, screening obtains wavelength ranges；
Spectral reflectivity single order inverse obtains module, for the wavelength ranges according to screening, obtains the corresponding spectrum of wavelength ranges
Reflectivity single order inverse and the corresponding amount of laying the dust data；
The inverse model that lays the dust constructs module, for reciprocal and corresponding stagnant according to the corresponding spectral reflectivity single order of wavelength ranges
Dust quantity data construct the inverse model that lays the dust using grid-search algorithm and the second random forests algorithm；
The amount of laying the dust data acquisition module lays the dust for will be input to measuring plants in the spectral reflectivity single order inverse of wavelength ranges
In inverse model, the amount of laying the dust data are obtained.
9. a kind of computer readable storage medium, stores computer program thereon, which is characterized in that the computer program is located
It manages and realizes the plant regulation of absorbing dust capability inspection based on random forests algorithm as claimed in any of claims 1 to 7 in one of claims when device executes
The step of survey method.
10. a kind of computer equipment, which is characterized in that including reservoir, processor and be stored in the reservoir and can
The computer program executed by the processor, the processor realize such as claim 1 to 7 when executing the computer program
Any one of described in the plant regulation of absorbing dust capability detection method based on random forests algorithm the step of.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811362049.6A CN109270010A (en) | 2018-11-15 | 2018-11-15 | A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811362049.6A CN109270010A (en) | 2018-11-15 | 2018-11-15 | A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109270010A true CN109270010A (en) | 2019-01-25 |
Family
ID=65189372
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811362049.6A Pending CN109270010A (en) | 2018-11-15 | 2018-11-15 | A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109270010A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110823190A (en) * | 2019-09-30 | 2020-02-21 | 广州地理研究所 | Island reef shallow sea water depth prediction method based on random forest |
| WO2021090324A1 (en) * | 2019-11-06 | 2021-05-14 | Migal Applied Research Ltd. | Remote measurement of crop stress |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104697964A (en) * | 2015-03-10 | 2015-06-10 | 西北大学 | Method for quantitative analysis of sulfur and phosphorus in steel and iron by combining random forest algorithm with laser induced breakdown spectroscopy |
| CN104897592A (en) * | 2015-06-11 | 2015-09-09 | 石河子大学 | Monitoring method of salt ion content in saline soil based on hyperspectral technology |
-
2018
- 2018-11-15 CN CN201811362049.6A patent/CN109270010A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104697964A (en) * | 2015-03-10 | 2015-06-10 | 西北大学 | Method for quantitative analysis of sulfur and phosphorus in steel and iron by combining random forest algorithm with laser induced breakdown spectroscopy |
| CN104897592A (en) * | 2015-06-11 | 2015-09-09 | 石河子大学 | Monitoring method of salt ion content in saline soil based on hyperspectral technology |
Non-Patent Citations (2)
| Title |
|---|
| WENLONG JING等: "Machine Learning for Estimating Leaf Dust Retention Based on Hyperspectral Measurements", 《HINDAWI JOURNAL OF SENSORS》 * |
| 陈元鹏等: "基于网格搜索随机森林算法的工矿复垦区土地利用分类", 《农业工程学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110823190A (en) * | 2019-09-30 | 2020-02-21 | 广州地理研究所 | Island reef shallow sea water depth prediction method based on random forest |
| WO2021090324A1 (en) * | 2019-11-06 | 2021-05-14 | Migal Applied Research Ltd. | Remote measurement of crop stress |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fan et al. | Empirical and machine learning models for predicting daily global solar radiation from sunshine duration: A review and case study in China | |
| Maltamo et al. | Predicting tree attributes and quality characteristics of Scots pine using airborne laser scanning data | |
| Zhou et al. | Maximum nighttime urban heat island (UHI) intensity simulation by integrating remotely sensed data and meteorological observations | |
| Yang | Double fence intercomparison reference (DFIR) vs. bush gauge for “true” snowfall measurement | |
| CN103983584B (en) | The inversion method of a kind of inland case �� waters chlorophyll-a concentration and device | |
| CN103868860A (en) | Method for monitoring nitrogen concentration of vegetation canopies in wetland based on hyperspectral vegetation index | |
| CN110378925B (en) | Ecological water reserve estimation method of airborne L iDAR and multispectral remote sensing technology | |
| CN109270010A (en) | A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm | |
| Henry et al. | Sugarcane land classification with satellite imagery using logistic regression model | |
| Yu et al. | Secchi depth inversion and its temporal and spatial variation analysis—A case study of nine plateau lakes in Yunnan Province of China | |
| CN114997514B (en) | Evaluation and prediction method for development degree of fracture disease of rammed earth site | |
| CN109270012A (en) | A kind of plant regulation of absorbing dust capability detection method based on relative coefficient | |
| CN116384591A (en) | Drought prediction method, system and medium based on big data | |
| Suk et al. | Creation of the snow avalanche susceptibility map of the Krkonoše Mountains using GIS | |
| Fukushima et al. | Semi-analytical prediction of Secchi depth using remote-sensing reflectance for lakes with a wide range of turbidity | |
| CN117010274B (en) | Intelligent early warning method for harmful elements in underground water based on integrated incremental learning | |
| Hamadeh et al. | Studying the factors affecting the risk of forest fire occurrence and applying neural networks for prediction | |
| CN107479097A (en) | A kind of fuzzy guarantor side filtering method based on efficient frontier structural scan | |
| CN109270009A (en) | A kind of plant regulation of absorbing dust capability detection method based on Taxonomy and distribution algorithm | |
| Çınaroğlu et al. | A data mining application of local weather forecast for Kayseri Erkilet Airport | |
| Shiranvand et al. | An analysis of dieback areas of Zagros oak forests using remote sensing data case study: Lorestan oak forest, Iran | |
| CN109490229A (en) | A kind of wavelength ranges screening technique, device, storage medium and equipment | |
| CN112380984A (en) | Remote sensing-based saline marsh vegetation slow flow capacity space evaluation method | |
| Beyene | Estimation of forest variable and aboveground biomass using terrestrial laser scanning in the tropical rainforest | |
| Asghari Saraskanrood et al. | Land surface temperature assessment in relation to land-use/land-cover (A case study: Isfahan City, Central Iran) |
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: 20190125 |