CN107895136B - Coal mine area identification method and system - Google Patents
Coal mine area identification method and system Download PDFInfo
- Publication number
- CN107895136B CN107895136B CN201710698775.4A CN201710698775A CN107895136B CN 107895136 B CN107895136 B CN 107895136B CN 201710698775 A CN201710698775 A CN 201710698775A CN 107895136 B CN107895136 B CN 107895136B
- Authority
- CN
- China
- Prior art keywords
- data
- image data
- remote sensing
- training
- sensing image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a coal mine area identification method and a system, which are used for identifying a coal mine area, and the method comprises the following steps: acquiring remote sensing image data of a target area and actually measured spectrum data of coal in the target area; screening the actually measured spectrum data to obtain spectrum data which is consistent with the wave band of the remote sensing image data in the actually measured spectrum data and serves as sample spectrum data, wherein the sample spectrum data comprises training data and testing data; training data by using a network training set containing coal and non-coal spectral data and adopting a preset extreme learning machine to obtain an optimal ELM feature classification model set with the optimal classification recognition rate aiming at the training data; and classifying the remote sensing image data by using the optimal ELM characteristic classification model set, and acquiring the remote sensing image data which is classified and identified as having coal characteristic data by the optimal ELM characteristic classification model set in the remote sensing image data as target image data, wherein the region corresponding to the target image data is a coal mine region.
Description
Technical Field
The invention relates to the technical field of coal mine detection, in particular to a coal mine area identification method and a coal mine area identification system.
Background
Coal is widely applied to various fields as a traditional energy source, and the consumption of coal is increased day by day along with the increasing of the population. Therefore, the discovery of new coal mines and the rational planning and maximum exploitation of coal resources in a mining area become important means for solving the increase in coal demand.
In the prior art, the identification of coal mine areas is mainly carried out by manual field reconnaissance or actual measurement on the basis of collected data. The method not only consumes a large amount of manpower and material resources in the practical process, but also cannot ensure the identification accuracy for the coal mine areas with complex terrain features.
Therefore, a coal mine area identification method and a coal mine area identification system are needed to effectively and accurately identify a coal mine area.
Disclosure of Invention
Technical problem to be solved
The invention discloses a coal mine area identification method, which aims to solve the problems that in the prior art, a large amount of manpower and material resources are consumed and the accuracy is poor when a coal mine is found through methods of data collection, field exploration and the like.
The invention discloses a coal mine area identification system which is matched with the coal mine area identification method for use.
(II) technical scheme
In order to achieve the above purpose, on one hand, the invention adopts the following technical scheme:
a coal mine area identification method comprises the following steps:
acquiring remote sensing image data of a target area and actually measured spectrum data of coal in the target area;
screening the measured spectrum data to obtain spectrum data which is consistent with the wave band of the remote sensing image data in the measured spectrum data and serves as sample spectrum data, wherein the sample spectrum data comprises training data and test data;
training the training data by using a network training set containing coal and non-coal spectral data and adopting a preset extreme learning machine to obtain an optimal ELM feature classification model set with the optimal classification recognition rate aiming at the training data;
and classifying the remote sensing image data by using the optimal ELM feature classification model set, and acquiring the remote sensing image data which is classified and identified as having coal feature data by the optimal ELM feature classification model set in the remote sensing image data as target image data, wherein the region corresponding to the target image data is a coal mine region.
Further, performing N rounds of training on the training data by adopting a preset extreme learning machine, performing T times of training on the training data in each round of training, and obtaining a classification recognition rate corresponding to each training after each training is completed;
after each round of training is finished, selecting the highest classification recognition rate from the T classification recognition rates as the optimal classification recognition rate, and taking the classification model corresponding to the optimal classification recognition rate as the optimal classification model;
and after N rounds of training are finished, forming a set by the N optimal classification models to serve as the optimal ELM characteristic classification model set.
Further, an excitation function between an input layer and a hidden layer in the preset extreme learning machine is an Ln function;
the formula of the Ln function is as follows:
Wherein, x is input data;
ai-the ith neuron inputs a weight;
bi-the ith neuron deviation value.
Further, the value range of the number N of rounds of training performed by the training data is 5 to 101, and N is an odd number.
Further, in the process of classifying the remote sensing image data by using the optimal ELM feature classification model set, N optimal classification models are used for classifying the remote sensing image data respectively;
when the remote sensing image data is more than or equal to
And if the optimal classification model is classified and identified as the target image data, the region corresponding to the target image data is a coal mine region.
Further, after obtaining the remote sensing image data, the remote sensing image data is corrected before classifying the remote sensing image data.
Further, the training data is trained for T times in each round of training, and the value range of T is 100-1000.
In order to achieve the above purpose, on the other hand, the invention adopts the following technical scheme:
a coal mine area identification system for performing coal mine area identification in combination with the coal mine area identification method as described above, comprising:
the acquisition unit is used for acquiring remote sensing image data of a target area and actually measured spectrum data of coal in the target area;
the screening unit is used for screening the actually measured spectrum data to obtain sample spectrum data;
the training unit is used for training data in the sample spectrum data to obtain an optimal ELM characteristic classification model set;
the classification unit is used for classifying the remote sensing image data to obtain target image data;
the acquisition unit, the screening unit, the training unit and the classification unit are connected in sequence.
The remote sensing image data acquisition unit is used for acquiring remote sensing image data, and the remote sensing image data acquisition unit is used for acquiring remote sensing image data;
the deviation rectifying unit comprises a radiation deviation rectifying subunit and an atmosphere deviation rectifying subunit.
Further, the acquisition unit comprises a remote sensing image acquisition subunit and an actual measurement spectrum acquisition subunit.
(III) advantageous effects
The coal mine area identification method and the system thereof disclosed by the invention are used for acquiring remote sensing image data and actual measurement spectrum data of a target area, training by utilizing the actual measurement spectrum data to acquire an optimal ELM characteristic classification model set and classifying the remote sensing image data by using the set so as to identify the coal mine area in the target area. By using the method, an operator does not need to go to the ground for surveying, so that the limitation of manual surveying is avoided, and the manpower, financial resources and material resources are saved. Meanwhile, the method can effectively improve the coal mine area identification accuracy and efficiently and quickly identify whether coal mine distribution exists in the target area.
Drawings
FIG. 1 is a flow chart of a coal mine area identification method according to an embodiment of the invention;
FIG. 2 is a coal mine area identification system according to a second embodiment of the present invention;
FIG. 3 is a flow chart of a coal mine area identification method disclosed in the third embodiment of the invention;
fig. 4 is a coal mine area identification system disclosed in the fourth embodiment of the present invention.
[ description of reference ]
1. An acquisition unit; 11. a remote sensing image acquisition subunit; 12. an actual measurement spectrum acquisition subunit;
2. a screening unit;
3. a training unit;
4. a classification unit;
5. a deviation rectifying unit; 51. a radiation deviation rectifying subunit; 52. and the atmosphere deviation rectifying subunit.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. The embodiments and features of the embodiments of the present invention may be arbitrarily combined with each other without conflict. Also, while a logical order is shown in the flow diagrams, in some cases, the steps shown or described may be performed in an order different than here.
In addition, the term "and/or" herein is only one kind of association relationship describing an associated object, and means that there may be three kinds of relationships, for example, a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" in this document generally indicates that the preceding and following related objects are in an "or" relationship unless otherwise specified.
For better understanding of the above technical solutions, the following detailed descriptions will be provided in conjunction with the drawings and the detailed description of the embodiments.
Example one
As shown in fig. 1, the coal mine area identification method disclosed in this embodiment specifically includes the following steps when the coal mine area identification is performed by using the method:
and S10, acquiring remote sensing image data of the target area and actually measured spectrum data of the coal in the target area.
The target area in this step may be a documented coal mine area, or an area in which a coal mine is suspected to exist, or an area in which a coal mine has been found, or an area in which a coal mine area overlaps with other mine areas. After the target area is determined, remote sensing image data of the target area needs to be acquired, preferably, remote sensing image data with high resolution is acquired, for example, remote sensing image data generated by using Landsat 8OLI satellites, the spatial resolution of the remote sensing image data is preferably 30 meters, and may be higher or lower, and there is no excessive limitation on the resolution requirement of the remote sensing image data. Besides the method of obtaining the remote sensing image data through satellite generation, the remote sensing image data corresponding to the target area can be downloaded from the USGS remote sensing image database, and the remote sensing image data of the target area can be obtained from the remote sensing image data.
In the prior art, methods for identifying coal mine areas by using remote sensing image data exist, although the types of the remote sensing image data are continuously increased, the algorithm for analyzing and processing the remote sensing image data is relatively lagged, so that the accuracy of the result obtained by simply identifying the coal mine areas by using the remote sensing image data is low. Therefore, in the embodiment, the remote sensing image data and the actually measured spectrum data of the target area are obtained simultaneously for subsequent analysis and calculation, and the identification accuracy is improved.
After the target area is determined, spectral data of the coal in the target area is collected using a device, preferably a spectrometer. Spectral data is collected for the target area, for example using SVC HR-1024, to obtain measured spectral data of the coal in the target area. For example, to determine whether a coal mine exists in shenyang, spectral data collection needs to be performed on coal existing in the shenyang range, and if it is desired to identify whether a coal mine exists in the qinhuang island, spectral data collection needs to be performed on coal existing in the qinhuang island.
And step S20, screening the actually measured spectrum data to obtain spectrum data which is consistent with the wave band of the remote sensing image data in the actually measured spectrum data and is used as sample spectrum data, and dividing the sample spectrum data into training data and test data.
Because the spectrometer is used for collecting the spectral data, the actually measured spectral data totally comprises 1024 wave bands, and the range of the spectral data reaches 350 nm-2500 nm. The actually measured spectrum data has a wide band range, but the bands of the remote sensing image data collected in step S10 are 7, and the data in each band may interfere with each other, so the actually measured spectrum data needs to be screened to reduce the band range of the actually measured spectrum data. And obtaining spectral data consistent with the wave band of the remote sensing image data according to the corresponding relation of the wave band of the remote sensing image data and the actually measured spectral data, and using the spectral data as sample spectral data.
As shown in table 1, the comparison table of the wavelength band positions of the remote sensing image data collected by using Landsat 8OLI and the actual measurement spectrum data collected by using SVCHR-1024 is disclosed in this embodiment. The measured spectrum data that is identical to the 7 bands included in the remote sensing image data can be screened out from the measured spectrum data by the correspondence between the bands of the measured spectrum data and the bands of the remote sensing image data indicated in table 1.
After screening, sample spectrum data is obtained and is divided into training data and testing data for subsequent processes.
TABLE 1 Landsat 8OLI and SVC HR-1024 band position comparison table
And step S30, training the training data by using a network training set containing the coal and non-coal spectral data and adopting a preset extreme learning machine to obtain an optimal ELM feature classification model set with the optimal classification recognition rate aiming at the training data.
The network training set containing the coal and non-coal spectral data can be downloaded from a database already available in the prior art, and the network training set containing the coal and non-coal data can be considered to actually contain the coal characteristic data and the non-coal characteristic data. The main purpose of the step is to adopt a preset extreme learning machine to train training data in sample spectrum data for multiple times, and an optimal ELM characteristic classification model set is established in the training process. The essence of the method is to obtain a calculation model, and whether the data contains coal characteristic data and non-coal characteristic data recorded in a network training set can be rapidly and accurately identified.
The remote sensing blackbody spectrum data in the remote sensing image data within the wave band range of 1 to 5 in the table 1 come from various different substances on the earth surface, and are greatly different from the actually measured spectrum data of coal. If the remote sensing image data is used for identifying and sampling coal mine areas, a large amount of error data can be sampled, and the data is used for classification, so that the result is inaccurate. Therefore, the optimal ELM feature classification model set established in the embodiment is obtained by utilizing the actually measured spectral data after the extreme learning machine training and screening, so that the accuracy of model establishment is ensured, and the accuracy of coal mine area identification is improved.
After the optimal ELM classification model set is obtained, the test data is needed to be used for testing each optimal classification model in the set, and the classification recognition rate of each optimal classification model for the test data is obtained, so that whether each optimal classification model in the set is optimal or not is verified. The test data is used for testing, so that the functions of verification and insurance are achieved, and the condition that the set contains the non-optimal classification model due to errors or errors is prevented.
And S40, classifying the remote sensing image data by using the obtained optimal ELM feature classification model set, and obtaining the remote sensing image data which is classified and identified as having coal feature data by the optimal ELM feature classification model set in the remote sensing image data as target image data, wherein the region corresponding to the target image data is a coal mine region.
The essence of the step is that the remote sensing image data is judged by utilizing an optimal ELM characteristic classification identification model set, whether the remote sensing image data contains coal characteristic data or not is judged, and if the remote sensing image data contains the coal characteristic data, the coal exists in a region of the remote sensing image with the coal characteristic data; if the remote sensing image data contains non-coal characteristic data, the remote sensing image is proved to have no coal, and the aim of identifying the coal mine area is fulfilled.
Example two
As shown in fig. 2, the coal mine area recognition system disclosed in this embodiment is a coal mine area recognition system that can be applied to the recognition system in this embodiment, and the system includes an acquisition unit 1, a screening unit 2, a training unit 3, and a classification unit 4, where the acquisition unit 1, the screening unit 2, the training unit 3, and the classification unit 4 are connected in sequence.
The acquiring unit 1 is used for matching with step S10 in the first embodiment, and is divided into a remote sensing image acquiring subunit 11 and an actual measurement spectrum acquiring subunit 12. The remote sensing image obtaining subunit 11 is configured to obtain remote sensing image data, and specifically, is any device capable of obtaining remote sensing image data of a target area. The measured spectrum acquiring subunit 12 is configured to acquire measured spectrum data of a target area, specifically, any device capable of acquiring measured spectrum data of a target area. The above-mentioned devices for acquiring remote sensing image data and measured spectral data are not described in detail herein, and any device capable of achieving the above-mentioned purpose is included in the acquisition unit 1.
The screening unit 2 is configured to implement step S20 in a coordinated manner, so that the waveband of the screened measured spectral data is consistent with the waveband included in the remote sensing image data, thereby reducing the waveband range of the measured spectral data and avoiding mutual interference between the wavebands.
The training unit 3 is configured to implement step S30 in a matching manner, where the training unit 3 needs to use a computer, and an extreme learning machine is used to establish an optimal ELM feature classification model set corresponding to training data by using the training data in the measured spectral data.
The classification unit 4 is configured to implement step S40 in a coordinated manner, and the classification unit 4 needs to use a computer to classify the remote sensing image data by using the established optimal ELM feature classification model set, and identify a region containing coal feature data in the remote sensing image data, where the region is a coal mine region.
EXAMPLE III
As shown in fig. 3, it is a specific flow of the coal mine area identification method disclosed in this embodiment, and specifically includes the following steps:
and step S10, obtaining remote sensing image data of the target area and actually measured spectrum data of the coal in the target area.
And step S20, screening the actually measured spectrum data to obtain spectrum data which is consistent with the wave band of the remote sensing image data in the actually measured spectrum data and is used as sample spectrum data, wherein the sample spectrum data comprises training data and test data.
And step S30, carrying out correction processing on the remote sensing image data.
Because the interference of atmosphere and illumination on the reflection of the ground object is generally received in the process of acquiring the remote sensing image data, the reflection spectrum data can be distorted, in order to improve the accuracy of the remote sensing image data, the remote sensing image data needs to be corrected so as to eliminate the influence of factors such as atmosphere and illumination on the reflection of the ground object, and more real reflectivity and radiance of the ground object are obtained, so that the corrected remote sensing image data can be closer to the actual spectrum characteristic of the ground object.
Specifically, in this embodiment, the correction processing performed on the remote sensing image data includes radiometric calibration and atmospheric correction, but is not limited to the two correction processing, and any correction method capable of obtaining better remote sensing image data can be applied to the system, such as geometric correction of ground control points. Remote sensing image data representing the actual characteristics of the substance included in the target region with higher accuracy can be obtained by the correction.
Radiometric calibration is the process of converting the voltage or digital quantized value (DN) recorded by a sensor into an absolute radiance value (radiance). The remote sensing image data to be corrected can be selected by meeting parameters within a certain threshold range, for example, a calibration tool for Landsat8 satellite images is provided through software ENVI, the remote sensing image data to be corrected is input, the correction parameters are selected, and the obtained output is the corrected remote sensing image data.
After radiometric calibration, atmospheric corrections may be made, such as by FLAASH. Since the parameters have been set to those required for atmospheric correction at the time of irradiation calibration. Therefore, atmospheric correction can be performed according to the LandsatFLAASH method of the MODTRAN model, data for opening radiometric calibration is input, atmospheric correction is set according to relevant parameters of the region, the relevant parameters such as satellite types and image acquisition dates are output, and remote sensing image data after atmospheric correction is output.
And step S40, training the training data by using a network training set containing the coal and non-coal spectral data and adopting a preset extreme learning machine to obtain an optimal ELM feature classification model set with the optimal classification recognition rate aiming at the training data.
An extreme learning machine, namely ELM, is an algorithm for replacing iteration by solving a linear equation set for traditional neural network parameter optimization. Compared with the traditional learning algorithm, the ELM overcomes the defects of repeated iteration of parameters, so that the ELM has higher learning speed and good generalization capability, and is explained in principle below.
In the extreme learning algorithm, for any given A different samples (x)i,ti)，xi＝[xi1,xi2,…,xin]T∈Rn，ti＝[ti1,ti2,…,tim]T∈RmThen, the single hidden layer feedforward neural network representing L hidden layer nodes is as follows:
wherein, x ∈ R, ai∈Rn,βi∈RmAnd the relation between the ith hidden node and x is shown.
From g (x): r → R, yield:
G(ai,bi,x)＝g(ai·x+bi),bi∈R
wherein, ai＝[ai1,ai2,…,ain]T∈RnThe input weight vector from the input layer to the ith hidden layer node is obtained;
bia threshold value representing the ith hidden layer node;
βi＝[βi1,βi2,…βim]Trepresenting the output weight vector from the ith hidden layer node to the output layer;
ai，bithe values are randomly taken during the training of the model.
B samples (x) are selectedi,ti)∈Rn×Rm，xi∈Rn，ti∈RmThen, equation (1) can be simplified to:
Hβ＝T
in the above equation, H is the hidden layer output matrix.
Therefore, the extreme learning machine can obtain a better algorithm, and therefore, the extreme learning machine is used for establishing an algorithm model to classify the remote sensing image data.
Step S40 specifically includes the following steps:
and step S41, performing N rounds of training on the training data by adopting a preset extreme learning machine, performing T times of training on the training data in each round of training, and obtaining the classification recognition rate corresponding to the training after each time of training is completed.
The screened actually measured spectrum data is used as sample spectrum data, training data in the sample spectrum data is trained, a limit learning machine needs to be preset before training, and the limit learning machine comprises two parameters which are respectively an activation function and a hidden layer node number.
The activation functions of the prior art ELMs are typically the following: sigmoid function, ReLU function, Softplus function. The Sigmoid function is a good threshold function, and has great advantages in processing the neural network problem, and as the ReLU function is closer to a biological activation model and has a simple form, the modified linear ReLU function is widely applied, and gradually replaces the Sigmoid activation function to become the mainstream. Because power operation and division operation are not used, the ReLU operation speed is higher, and the generalization performance is good. The Softplus activation function is approximately smooth and closer to a biological activation model than the ReLU function, and the Softplus function can enable the average performance of the whole network model to be better than the ReLU function, but the Softplus function uses power operation.
In some cases, a large amount of data needs to be processed, and the data value is large, and if the data is not properly preprocessed, the algorithm can reach infinite results in the calculation process, and the whole system is crashed. Therefore, selecting a superior algorithm, i.e., selecting a superior function, is of crucial importance for data processing and recognition.
In the embodiment, when the limit learning machine is preset, an excitation function between the input layer and the hidden layer is selected as an Ln function;
the formula of the Ln function is as follows:
Wherein, x is input data;
ai-the ith neuron inputs a weight;
bi-the ith neuron deviation value.
The Ln function is a nonlinear continuous differentiable function, overcomes the defects of the Softplus activation function, and has good effect in a large number of experiments, so that better training effect can be obtained by using the Ln function.
TABLE 2 comparison of accuracy of various activation function models
Table 2 is an accuracy comparison table of various excitation function models, and from the comparison result in table 2, the accuracy of the Ln function selected in this embodiment reaches 98.6012%, which is much higher than that of other functions, so that the effect of using the Ln function to identify remote sensing image data is better, and the identification accuracy is higher.
The number of hidden layer nodes can greatly affect the learning and information processing capabilities of the extreme learning machine, and the number of hidden layer nodes can increase the complexity of the extreme learning machine network, so that the learning time is prolonged, and the over-fitting phenomenon is easy to occur. If the number of hidden layer nodes is small, a certain limit is generated on the learning and processing capacity of the limit learning machine network. In the embodiment of the invention, the number of hidden layer nodes is selected in the range of 10-500 during setting, the number of hidden layer nodes in the range enables the learning time to be short, the overfitting phenomenon cannot occur, the range is generally obtained by adopting an empirical formula, and preferably, the number of hidden layer nodes is 45, so that the best training effect can be realized.
The traditional ELM algorithm model has an unobvious effect on remote sensing image data classification, and because ELM input weights and hidden layer thresholds are randomly assigned, the output of ELM is unstable, and the ELM is easy to fall into a local minimum value, so that the accuracy is low.
Therefore, in this embodiment, N rounds of training are performed on the training data, and T times of training are performed on the training data in each round of training, where a value range of N is 5 to 101, N is an odd number, and a value range of T is 100 to 1000, so as to obtain a plurality of optimal classification models. In the specific implementation process of this embodiment, N is 11, that is, 11 rounds of training are performed, and T is 200, that is, 200 times of training are performed in each round, so that the output result can be more accurate.
And step S42, after each round of training is finished, selecting the highest classification recognition rate from the T classification recognition rates as the optimal classification recognition rate, and taking the classification model corresponding to the optimal classification recognition rate as the optimal classification model.
According to the description in step S41, after 1 round of training is completed, 200 classification recognition accuracies are obtained, and then the highest one of the 200 classification recognition accuracies is taken as the optimal classification recognition accuracy, the corresponding classification model is the optimal classification model, and the essence of the optimal classification model here is the input weight and the hidden layer threshold between the corresponding hidden layers in the extreme learning machine in the process of obtaining the optimal classification recognition rate.
And step S43, after N rounds of training are finished, forming a set by the N optimal classification models, and using the set as an optimal ELM feature classification model set.
The optimal classification models obtained in each round in step S42 include parameters such as optimal input weights and optimal hidden layer thresholds, and these parameters are collected together to form a set, so as to obtain an optimal ELM classification model feature set, where the optimal ELM classification model feature set in this embodiment includes 11 optimal classification models in total.
The optimal ELM feature classification model set in the embodiment is obviously improved compared with the traditional ELM, each optimal classification model corresponds to one group of parameters, the results of the same sample discrimination may be different, and the stability is not high. In order to improve the stability of the model effect, an optimal ELM feature classification model set is provided, and the prediction precision of the model can be further improved.
And S50, classifying the remote sensing image data by using the optimal ELM feature classification model set, and acquiring the remote sensing image data which is classified and identified as having coal feature data by the optimal ELM feature classification model set in the remote sensing image data as target image data, wherein the region corresponding to the target image data is a coal mine region.
And S51, in the process of classifying the remote sensing image data by using the optimal ELM feature classification model set, classifying the remote sensing image data by the N optimal classification models respectively.
In this step, the remote sensing image data is classified by using the optimal ELM classification model feature set containing 11 optimal classification models obtained in step S43. When classification is carried out, each single optimal classification model carries out classification and identification on the remote sensing image data, and due to the fact that parameters of each optimal classification model are different, classification and identification results of the same remote sensing image data are possibly different. The whole classification process can be understood in the classification process, the first optimal classification model carries out primary classification and identification on the remote sensing image data, and the remote sensing image data is judged to contain coal characteristic data, so that the remote sensing image data is determined to be of type I; and carrying out primary classification recognition on the remote sensing image data by the second optimal classification model, judging that the remote sensing image data does not contain coal characteristic data, and determining that the remote sensing image data is II. After the remote sensing image data are classified by the 11 optimal classification models respectively, the remote sensing image data are distinguished one by one each time and classified.
Step S52, when the remote sensing image data is greater than or equal to
And classifying and identifying the optimal classification model as target image data, wherein the region corresponding to the target image data is a coal mine region.
After the 11 classification identifications in step S51 are completed, when the number of times of determination of having the coal characteristic data is equal to or greater than 6 times, the remote sensing image data is the target image data, that is, the location of the remote sensing image data is the coal mine area. For example, if there are 6 times of classification and 5 times of classification and identification as class i and ii, the remote sensing image data is the target image data. For another example, if 2 of 11 classification identifications are classified as class i and 9 are classified as class ii, the remote sensing image data does not include coal feature data and is not target image data.
The computer program instructions corresponding to the coal mine area identification method in this embodiment may be stored in a storage medium such as an optical disc, a hard disk, or a usb disk.
Example four
As shown in fig. 4, the coal mine area identification system disclosed in this embodiment is used in cooperation with the coal mine area identification method in the third embodiment.
The difference between the present embodiment and the second embodiment is that the present embodiment further includes a deviation rectifying unit 5, and the deviation rectifying unit 5 is used for performing correction processing on the remote sensing image data, and includes a radiation deviation rectifying subunit 51 and an atmospheric deviation rectifying subunit 52. The radiation correction subunit 51 is used to obtain more real surface feature reflectivity and radiance of the remote sensing image data. The atmospheric deviation rectifying subunit 52 is used for eliminating the influence of factors such as atmosphere and illumination on the reflection of the ground objects.
The correction unit 5 may be disposed at any position between the acquisition unit 1 and the classification unit 4, as long as the correction and correction processing of the remote sensing image data is completed before the remote sensing image data is classified and identified. In this embodiment, the deviation rectifying unit 5 is specifically disposed between the obtaining unit 1 and the training unit 3, and the deviation rectifying unit 5 and the screening unit 2 respectively process the remote sensing image data and the measured spectral data of the coal in the target area.
The coal mine area identification method and the system thereof disclosed by the invention are used for acquiring remote sensing image data and actual measurement spectrum data of a target area, training by utilizing the actual measurement spectrum data to acquire an optimal ELM characteristic classification model set and classifying the remote sensing image data by using the set so as to identify the coal mine area in the target area. By using the method, operators do not need to go to the field for surveying, so that the limitation of manual surveying is avoided, and the manpower, financial resources and material resources are saved. Meanwhile, the method can effectively improve the coal mine area identification accuracy and efficiently and quickly identify whether coal mine distribution exists in the target area.
It will be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the specific working processes of the system, the apparatus and the unit described above, reference may be made to the corresponding processes in the foregoing method embodiments, and details are not described here again.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a Universal Serial Bus flash drive (USB flash drive), a portable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A coal mine area identification method is characterized by comprising the following steps:
acquiring remote sensing image data of a target area and actually measured spectrum data of coal in the target area, wherein the actually measured spectrum data is acquired by a spectrometer through collecting the coal in the target area;
by screening the measured spectral data, acquiring spectral data which is consistent with the band of the remote sensing image data in the measured spectral data as sample spectral data according to the corresponding relation of the band of the remote sensing image data and the measured spectral data, wherein the sample spectral data comprises training data and test data;
training the training data by using a network training set containing coal and non-coal spectral data and adopting a preset extreme learning machine to obtain an optimal ELM feature classification model set with the optimal classification recognition rate aiming at the training data;
and classifying the remote sensing image data by using the optimal ELM feature classification model set, and acquiring the remote sensing image data which is classified and identified as having coal feature data by the optimal ELM feature classification model set in the remote sensing image data as target image data, wherein the region corresponding to the target image data is a coal mine region.
2. The coal mine area identification method of claim 1,
performing N rounds of training on the training data by adopting a preset extreme learning machine, performing T times of training on the training data in each round of training, and obtaining a classification recognition rate corresponding to each training after each training is completed;
after each round of training is finished, selecting the highest classification recognition rate from the T classification recognition rates as the optimal classification recognition rate, and taking the classification model corresponding to the optimal classification recognition rate as the optimal classification model;
and after N rounds of training are finished, forming a set by the N optimal classification models to serve as the optimal ELM characteristic classification model set.
3. The coal mine area identification method as claimed in claim 1, wherein the excitation function between the input layer and the hidden layer in the preset extreme learning machine is an Ln function;
the formula of the Ln function is as follows:
Wherein, x is input data;
ai-the ith neuron inputs a weight;
bi-the ith neuron deviation value.
4. The coal mine area identification method of claim 2, wherein the training data is trained with a number of rounds, N, ranging from 5 to 101, N being an odd number.
5. The coal mine area identification method according to claim 4, wherein in the process of classifying the remote sensing image data by using the optimal ELM feature classification model set, N optimal classification models are used for classifying the remote sensing image data respectively;
when the remote sensing image data is more than or equal to
And if the optimal classification model is classified and identified as the target image data, the region corresponding to the target image data is a coal mine region.
6. The coal mine area identification method as claimed in any one of claims 1 to 5,
after obtaining the remote sensing image data, the remote sensing image data is corrected before being classified.
7. The coal mine area identification method of claim 2, wherein the training data is trained T times in each round of training, T having a value in the range of 100 to 1000.
8. A coal mine area identification system for performing coal mine area identification in combination with the coal mine area identification method according to any one of claims 1 to 7, comprising:
the device comprises an acquisition unit, a data acquisition unit and a data processing unit, wherein the acquisition unit is used for acquiring remote sensing image data of a target area and actually measured spectrum data of coal in the target area, and the actually measured spectrum data is acquired by a spectrometer through collecting the coal in the target area;
the screening unit is used for screening the actually measured spectrum data to obtain sample spectrum data;
the training unit is used for training data in the sample spectrum data to obtain an optimal ELM characteristic classification model set;
the classification unit is used for classifying the remote sensing image data to obtain target image data;
the acquisition unit, the screening unit, the training unit and the classification unit are connected in sequence.
9. The coal mine area identification system of claim 8, further comprising a rectification unit disposed between the acquisition unit and the classification unit for rectifying the remote sensing image data;
the deviation rectifying unit comprises a radiation deviation rectifying subunit and an atmosphere deviation rectifying subunit.
10. The coal mine area identification system of claim 8 wherein the acquisition unit comprises a remote sensing image acquisition subunit and a measured spectrum acquisition subunit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710698775.4A CN107895136B (en) | 2017-08-15 | 2017-08-15 | Coal mine area identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710698775.4A CN107895136B (en) | 2017-08-15 | 2017-08-15 | Coal mine area identification method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107895136A CN107895136A (en) | 2018-04-10 |
| CN107895136B true CN107895136B (en) | 2020-09-15 |
Family
ID=61803193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710698775.4A Active CN107895136B (en) | 2017-08-15 | 2017-08-15 | Coal mine area identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107895136B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108489912B (en) * | 2018-05-11 | 2019-08-27 | 东北大学 | A kind of coal constituent analysis method based on coal spectroscopic data |
| CN109034239B (en) * | 2018-07-23 | 2021-01-26 | 北京金风科创风电设备有限公司 | Remote sensing image classification method, and site selection method and device for distributed wind power plant |
| CN109522788B (en) * | 2018-09-30 | 2020-11-06 | 广州地理研究所 | City range extraction method and device based on random forest classification algorithm and electronic equipment |
| CN111742329B (en) * | 2020-05-15 | 2023-09-12 | 安徽中科智能感知科技股份有限公司 | Mining typical feature dynamic monitoring method and platform based on multi-source remote sensing data fusion and deep neural network |
| CN111595792B (en) * | 2020-06-12 | 2023-07-25 | 鞍钢集团矿业有限公司 | Mine TFe content detection method based on hyperspectrum and TELM |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103345707A (en) * | 2013-06-04 | 2013-10-09 | 中国科学院遥感与数字地球研究所 | Crop maturation stage remote sensing prediction method based on multi-source remote sensing data |
| CN103970996A (en) * | 2014-05-04 | 2014-08-06 | 江苏优图空间信息科技有限公司 | Method for building general type vegetation index model |
| CN106600589A (en) * | 2016-12-09 | 2017-04-26 | 中国科学院合肥物质科学研究院 | Coastline area remote sensing map-based satellite borne spectrograph pointing registration method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103942749B (en) * | 2014-02-24 | 2017-01-04 | 西安电子科技大学 | A kind of based on revising cluster hypothesis and the EO-1 hyperion terrain classification method of semi-supervised very fast learning machine |
| CN105279556B (en) * | 2015-11-05 | 2017-11-07 | 国家卫星海洋应用中心 | A kind of Enteromorpha detection method and device |
| CN106228197A (en) * | 2016-08-15 | 2016-12-14 | 南京信息工程大学 | A kind of satellite image cloud amount recognition methods based on self adaptation extreme learning machine |
| CN106560695A (en) * | 2016-10-20 | 2017-04-12 | 中国计量大学 | Wuyi rock tea production place identification method through combination of three detection methods |
-
2017
- 2017-08-15 CN CN201710698775.4A patent/CN107895136B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103345707A (en) * | 2013-06-04 | 2013-10-09 | 中国科学院遥感与数字地球研究所 | Crop maturation stage remote sensing prediction method based on multi-source remote sensing data |
| CN103970996A (en) * | 2014-05-04 | 2014-08-06 | 江苏优图空间信息科技有限公司 | Method for building general type vegetation index model |
| CN106600589A (en) * | 2016-12-09 | 2017-04-26 | 中国科学院合肥物质科学研究院 | Coastline area remote sensing map-based satellite borne spectrograph pointing registration method |
Non-Patent Citations (1)
| Title |
|---|
| Research on Coal Exploration Technology Based on Satellite Remote Sensing;Dong Xiao 等;《Journal of Sensors》;20161231;第1-9页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107895136A (en) | 2018-04-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107895136B (en) | Coal mine area identification method and system | |
| US11521380B2 (en) | Shadow and cloud masking for remote sensing images in agriculture applications using a multilayer perceptron | |
| Halme et al. | Utility of hyperspectral compared to multispectral remote sensing data in estimating forest biomass and structure variables in Finnish boreal forest | |
| AU2021323934A1 (en) | System and method for image-based remote sensing of crop plants | |
| CN114219847B (en) | Method and system for determining crop planting area based on phenological characteristics and storage medium | |
| Krishna et al. | Hyperspectral satellite data analysis for pure pixels extraction and evaluation of advanced classifier algorithms for LULC classification | |
| Zhang et al. | Predicting soil moisture content over partially vegetation covered surfaces from hyperspectral data with deep learning | |
| Norman et al. | Spatio-statistical optimization of image segmentation process for building footprint extraction using very high-resolution WorldView 3 satellite data | |
| CN111582387A (en) | Rock spectral feature fusion classification method and system | |
| CN112668613A (en) | Satellite infrared imaging effect prediction method based on weather forecast and machine learning | |
| CN103278467A (en) | Rapid nondestructive high-accuracy method with for identifying abundance degree of nitrogen element in plant leaf | |
| CN111222545B (en) | Image classification method based on linear programming incremental learning | |
| Aymerich et al. | Detection of tephra layers in Antarctic sediment cores with hyperspectral imaging | |
| CN116863341B (en) | Crop classification and identification method and system based on time sequence satellite remote sensing image | |
| Jadhav et al. | Segmentation analysis using particle swarm optimization-self organizing map algorithm and classification of remote sensing data for agriculture | |
| CN109726679B (en) | Remote sensing classification error spatial distribution mapping method | |
| CN115015258B (en) | Crop growth vigor and soil moisture association determination method and related device | |
| CN110807387A (en) | Object classification method and system based on hyperspectral image characteristics | |
| Nidamanuri et al. | Existence of characteristic spectral signatures for agricultural crops–potential for automated crop mapping by hyperspectral imaging | |
| Vikström | A comparison of different machine learning algorithms applied to hyperspectral data analysis | |
| Rajeswari et al. | Classification based land use/land cover change detection through Landsat images | |
| Devidhanshrii et al. | Multiobjective Convolution Neural Network Towards Soil Nutrients Classification For Crop Recommendation On Based On Spectral And Spatial Properties Using Landsat Hyperspectral Images | |
| Schneider et al. | Classification of hyperspectral imagery using GPs and the OAD covariance function with automated endmember extraction | |
| CA3221595A1 (en) | A method and system for logging data for a mineral sample | |
| Idris | Predicting Grain Yield by Utilizing Multispectral Images and Convolutional Neural Network |
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 |