KR102175195B1 - System and Method for characterizing spectral library built for hazardous chemicals mixed in river flow using hyperspectral image - Google Patents

Abstract

The present invention relates to a device for characterizing a spectral library to improve an identification capability for harmful chemical substances mixed in river water using a hyperspectral image and a method therefor, wherein the device is configured to construct a standard feature spectral library of harmful chemical substances mixed in river water so as to increase accuracy in identifying the harmful chemical substances. The device comprises: a group-specific correlation coefficient comparison unit which compares group-specific correlation coefficients of a spectral library of harmful chemical substances mixed in river water, constructed by a construction means for a spectral library of harmful chemical substances mixed in river water, and the newly photographed spectral spectrum of harmful chemical substances to each other; a group selection unit which selects a group belonging to an upper value set as a result of the comparison of the group-specific correlation coefficient comparison unit; a recognition rate evaluation unit which evaluates a recognition rate between samples using a reflectance of the group selected by the group selection unit, and re-performs the steps from the group selection step when the recognition rate fails to reach a predetermined recognition rate; and a standard feature spectral library determination unit which determines a standard feature spectral library of harmful chemical substances mixed in river water when the finally-set recognition rate is satisfied.

Description

System and Method for characterizing spectral library built for hazardous chemicals mixed in river flow using hyperspectral image}

The present invention relates to the monitoring of hazardous chemicals, and specifically improves the ability to identify hazardous chemicals mixed in river water using hyperspectral images that improve the accuracy of identifying hazardous chemicals by establishing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water. It relates to an apparatus and method for characterizing a spectral library for.

Recently, the social interest related to water pollution such as green algae phenomena due to climate change and high temperatures in summer, and chemical substances and oil spills due to various accidents is increasing.

Among the water pollution cases, the leakage of hazardous chemicals due to a chemical accident adversely affects the human body when contacted, contaminates the air, water and soil, and causes discoloration or necrosis of surrounding crops.

In particular, chemicals that may spill into rivers are often colorless and water-soluble, making it difficult to confirm the spill with the naked eye. In the event of a chemical accident, chemical substances are directly detected by using simple detection equipment such as pH paper, simple identification detection kit, detection tube type detection equipment, electronic detection equipment portable ion spectrometer, portable FT-IR, etc.

Such a contact sensor is dependent on on-site personnel, and the installed detection sensor is also limitedly installed in a place where there is a concern about the leakage of chemical substances, so it is difficult to actively detect an uninstalled area.

The introduction of a remote sensing technique capable of overcoming the limitations of such a high-risk contact detection technology and the use of the latest sensor technology are being introduced.

The prior art remote sensing technology uses a line scan hyperspectral sensor mounted on a mobile platform such as a satellite, an aircraft, and a drone, and a hyperspectral image captured through a camera as a distribution of spatial spectral characteristics at the same time.

Spectroscopic libraries of the prior art have been built mostly through indoor spectrometers, outdoor spot spectroradiometers, and aerial hyperspectral images.

There is a problem in that different post-processing methods have to be applied depending on the acquisition method of the spectral spectrum because there is a separate post-processing technique for correcting the hyperspectral image or normalizing the spectral spectrum according to each device.

In addition, with the recent development of lightweight hyperspectral sensors that are mounted on drones rather than satellites or aerial vehicles, cases of acquiring and utilizing hyperspectral images using drones are increasing, but it is difficult to build a spectral library of acquired images. to be.

In addition, it is relatively easy to construct a spectral library for land coverings that do not have permeability such as hardwoods, conifers, paddy fields, fields, asphalt, concrete, rocks, etc., and although the process is common, there are no cases of spectral libraries for liquid substances such as hazardous chemicals. In addition, the process of construction is also difficult to normalize.

In particular, when using hyperspectral images for material identification, it is essential to construct a unique spectral library according to the object to be detected and classified.

However, the hyperspectral library performed so far was generally used after constructing a spectral library for single objects such as minerals, soils, vegetation, artifacts, and liquids (Yang et al ., 2019), but in a solvent such as chemicals mixed with river water. There have been few examples that can be referenced in the construction of libraries for underlying materials.

Examples of building similar spectroscopic libraries that distinguish specific substances mixed with water, although not chemical substances, include USGS (Kokaly et al. , 2017), Johns Hopkins University (Baldride et al ., 2009), and NASA (Baldridge et al ., 2009). And The ECOSTRESS spectral library version 1.0 (Meerdink et al ., 2019).

Specifically, for example, Montmor (4 types), H ₂ O Ice (1 type), Melting snow (16 types), Red Coated Algea Water mixed with a water solvent constructed by USGS (Kokaly et al ., 2017) (1 type), Seawater Open Ocean (1 type), Seawater Coast Chl (1 type) and JHU (9 types) Frost, Ice, Distilled Water, Sea Foam, Sea water, Coarse Granular Snow, Fine Snow, Medium Granular Snow , A unique spectral library was built based on the spectral information collected with an indoor spectrometer by collecting samples for tap water.

However, the problem is that in other cases, including the USGS case, the case of a spectral library based on an indoor spectrometer is predominantly, so it is not exactly a spectral library containing a water solvent that can be applied in a natural river under sunlight.

As described above, spectral libraries of the prior art have been used as data for grasping the types or characteristics of indicators, but there is no case of spectral libraries for liquid substances such as hazardous chemicals, and there is a problem that it is difficult to normalize the construction process.

In particular, the detection of hazardous chemical substances mixed in river water using the standard spectroscopic library of hazardous chemical substances mixed in river water using hyperspectral images uses the entire wavelength of 400 to 1000 nm, which is the wavelength band of the equipment according to the technical standard.

At this time, in the case of some hazardous chemicals, there may be a problem that it is difficult to identify them as corresponding substances.

Therefore, there is a need to develop a technology for detecting hazardous chemical substances mixed in river water and a technology for characterizing a spectroscopic library that can increase the accuracy of identification of hazardous chemical substances.

Korean Patent Registration No. 10-2073394 Korean Patent Registration No. 10-1700144 Korean Patent Registration No. 10-1414045

The present invention is to solve the problems of the conventional technology for monitoring hazardous chemicals, and by constructing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water, the harmful chemicals are mixed with river water using hyperspectral images that increase the accuracy of identification of hazardous chemicals. An object of the present invention is to provide an apparatus and method for characterizing a spectroscopic library to improve chemical identification ability.

The present invention uses a pre-built spectral library for hazardous chemicals mixed with river water and optimized according to a series of processes to establish a spectral library for standard characteristics of hazardous chemicals. An object of the present invention is to provide an apparatus and method for characterizing a spectral library for improving capability.

The present invention is a river water mixing using a hyperspectral image to standardize the process for constructing a standard characteristic spectroscopic library that is easy to identify hazardous chemicals by using a spectral library of hazardous chemicals. It is an object of the present invention to provide an apparatus and method for characterizing spectral libraries for improving the ability to identify hazardous chemicals.

The present invention enables high-accuracy identification by establishing a standard characteristic spectroscopic library of hazardous chemical substances mixed in river water, so that when a chemical accident occurs, the accident occurrence is immediately confirmed and the accidental substance is quickly identified. The purpose of this is to provide an apparatus and method for characterizing a spectroscopic library to improve the ability to identify hazardous chemical substances mixed with river water used.

The present invention builds a standard characteristic spectroscopic library of hazardous chemicals mixed in river water and applies it to real-time monitoring of hazardous chemicals, thereby effectively detecting and responding to chemical accidents. An object of the present invention is to provide an apparatus and method for characterizing a spectroscopic library to improve chemical identification ability.

Other objects of the present invention are not limited to the objects mentioned above, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

The apparatus for characterizing a spectral library for improving the ability to identify hazardous chemical substances in river water mixture using the hyperspectral image according to the present invention to achieve the above object is a river water mixing constructed by means of building a spectral library for hazardous chemical substances in river water mixing. A group-specific correlation coefficient comparison unit that compares the group-specific correlation coefficient between the hazardous chemical spectral library and the newly photographed spectral spectrum of the toxic chemical; ; A recognition rate evaluation unit that evaluates the recognition rate between samples using the reflectance of the group selected by the group selection unit, and re-performs it from the group selection step when the set recognition rate is not met; It characterized in that it comprises a; standard characteristic spectral library determination unit for determining the standard characteristic spectral library of the substance.

The method for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed with river water using a hyperspectral image according to the present invention for achieving another object is to use a super spectral image in the means of building a spectral library for hazardous chemical substances mixed with river water. Constructing a mixed hazardous chemical spectral library; Comparing the correlation coefficient for each group between the river water mixed toxic chemical spectral library and newly photographed spectral spectra of the toxic chemical substance constructed by the river water mixed toxic chemical spectral library construction means; Selecting a group belonging to an upper value set as a result of comparison in the correlation coefficient comparison step for each individual; evaluating the recognition rate between samples using the reflectance of the selected group, and re-performing from the group selection step if the recognition rate is not reached; And determining a standard characteristic spectral library of hazardous chemical substances mixed with river water when the set recognition rate is satisfied.

The apparatus and method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed with river water using the hyperspectral image according to the present invention as described above have the following effects.

First, it is possible to increase the accuracy of identification of hazardous chemicals by establishing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water.

Second, it is possible to build a standard characteristic spectral library of hazardous chemicals by optimizing it according to a series of processes using the pre-built spectral library for hazardous chemicals mixed with river water.

Third, it is possible to standardize the process for constructing a standard characteristic spectral library that is easy to identify hazardous chemicals by using the spectral library of hazardous chemicals mixed in river water constructed with hyperspectral images.

Fourth, by establishing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water, high-accuracy identification is possible, so that when a chemical accident occurs, the occurrence of an accident can be immediately identified and the accident substance can be quickly identified.

Fifth, by constructing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water and applying them to real-time monitoring of hazardous chemicals, the detection and response of chemical accidents can be made efficiently.

1A is a block diagram of an apparatus for characterizing a spectroscopic library for improving identification of hazardous chemical substances mixed with river water using hyperspectral images according to the present invention
1b is a configuration diagram of a spectral library construction unit according to the present invention
2A is a flow chart showing a method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed with river water using hyperspectral images according to the present invention
Figure 2b is a flow chart showing a spectral library construction process according to the present invention
3 is a spectral spectrum comparison graph
Figure 4 is a library of primary characteristics of sodium fluoride
5 is a spectral library of the fourth characteristic of sodium fluoride

Hereinafter, a preferred embodiment of an apparatus and method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed with river water using hyperspectral images according to the present invention will be described in detail as follows.

Features and advantages of the apparatus and method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed with river water using hyperspectral images according to the present invention will be apparent through detailed descriptions of each embodiment below.

1A is a configuration diagram of an apparatus for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed with river water using a hyperspectral image according to the present invention, and FIG. 1B is a configuration diagram of a spectral library construction unit according to the present invention.

The apparatus and method for characterizing a spectroscopic library for improving the ability to identify hazardous chemicals mixed in river water using hyperspectral images according to the present invention is to improve the accuracy of identification of hazardous chemicals by constructing a standard characteristic spectroscopic library of hazardous chemicals mixed in river water. It was made to increase.

To this end, the present invention may include a configuration of constructing a standard characteristic spectral library of hazardous chemicals by optimizing through a series of processes using a pre-built spectral library for hazardous chemicals mixed with river water.

The present invention may include a configuration for standardizing a process for constructing a standard characteristic spectral library for easy identification of hazardous chemicals by using a spectroscopic library of hazardous chemicals mixed with river water constructed from hyperspectral images.

The apparatus for characterizing a spectral library for improving the ability to identify hazardous chemicals mixed with river water using the hyperspectral image according to the present invention is as shown in FIG. 1A. Group-specific correlation coefficient comparison unit 100 that compares the group-specific correlation coefficient between the substance spectral library and the newly photographed spectral spectrum of hazardous chemicals, and the group-specific correlation coefficient comparison unit 100 selects the group that is the set higher value as a result of comparison A recognition rate evaluation unit 300 that evaluates the recognition rate between samples using the reflectivity of the group selected by the group selection unit 200 and the group selection unit 200 to perform the recognition rate again from the group selection step when the recognition rate is not reached. ), and a standard characteristic spectral library determination unit 400 for determining a standard characteristic spectral library of hazardous chemical substances mixed with river water when the finally set recognition rate is satisfied.

Here, the means for constructing a spectral library of hazardous chemical substances mixed with river water may include the following configuration.

As shown in Fig. 1b, a hyperspectral image capture unit 10 for taking a hyperspectral image of the prepared sample, and a hyperspectral image that is stored by converting the intensity of light into a digital number (a number obtained by normalizing the relative size) in the image storage step A hyperspectral image radiation correction unit (20) that converts an image into a reflectance value through radiometric calibration, and a river water mixed hazardous chemical sample taken from a hyperspectral image that is not affected by light scattering as much as possible. A spectral spectrum extraction unit 30 that extracts a spectral spectrum of a region, and a hazardous chemical characteristic extraction unit that removes spectral information of the river water in order to extract the characteristics of only the hazardous chemicals from a sample in which river water and hazardous chemicals are mixed (40) And, the outlier removal unit 50 that examines the effectiveness of the obtained spectral spectrum and removes the outlier, and the spectral spectrum remaining after the outlier removal in the outlier removal unit 50 is super smoothered to determine one spectral spectrum. And a vector normalization unit 70 for constructing the determined spectral spectrum as a final library and performing vector normalization in order to compare each hazardous chemical substance with each other.

2A is a flow chart showing a method for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed with river water using a hyperspectral image according to the present invention, and FIG. 2B is a flow chart showing a spectral library construction process according to the present invention. to be.

The method for characterizing a spectral library for improving the ability to identify hazardous chemicals in mixed river water using hyperspectral images according to the present invention is a spectral library for hazardous chemicals mixed in river water using hyperspectral images in the means for establishing a spectral library for hazardous chemicals mixed in river water. It is to build a standard characteristic spectral library using the result of constructing.

As shown in Fig. 2a, a river water mixed hazardous chemicals spectral library is constructed using a hyperspectral image in the river water mixed hazardous chemicals spectral library construction means (S201).

Then, the correlation coefficient for each group is compared between the river water mixed hazardous chemicals spectral library established by the river water mixed hazardous chemicals spectral library construction means and the newly photographed hazardous chemicals spectral spectrum (S202).

Then, a group belonging to the upper value set as a result of the comparison in the correlation coefficient comparison step for each group is selected (S203)

Subsequently, the recognition rate between the samples is evaluated using the reflectance of the selected group (S204), and if the recognition rate is not reached, the re-performation is performed from the group selection step (S205).

When the finally set recognition rate is satisfied, the standard characteristic spectroscopic library of hazardous chemical substances mixed with river water is determined (S206).

In addition, the process of constructing the river water mixed hazardous chemicals spectral library using the hyperspectral image in the means for constructing the river water mixed hazardous chemicals spectral library is as shown in FIG. 2B.

First, a sample is prepared and the hyperspectral image is captured by the hyperspectral image capturing unit 10 (S211).

Subsequently, in the image storage step in the hyperspectral image capture unit 10, the stored hyperspectral image by converting the light intensity into a digital number (a number obtained by normalizing the relative size) is radiated by the hyperspectral image radiation correction unit 20 ( Radiometric Calibration) is converted into a reflectance value (S212).

In addition, the spectral spectrum extraction unit 30 extracts a spectral spectrum of a region that is not affected by light scattering as much as possible from the river water mixed hazardous chemical substance sample photographed in the hyperspectral image (S213)

Subsequently, the spectral information of the river water is removed in order to extract the characteristics of only the hazardous chemical from the sample in which the river water and the hazardous chemical are mixed in the hazardous chemical property extracting unit 40 (S214).

Then, outliers are removed by examining the validity of the spectral spectrum obtained by the outlier removal unit 50 (S215).

Then, the spectral spectrum remaining after the outlier is removed in the spectral spectrum determination unit 60 is processed as a super smoother to determine one spectral spectrum (S216).

Then, the spectral spectra determined by the vector normalization unit 70 are constructed as a final library, and a spectral library is constructed by performing vector normalization in order to compare each of the hazardous chemicals with each other (S217).

In the process of constructing a spectral library for hazardous chemical substances mixed with river water using such hyperspectral images, a step of correcting the hyperspectral images will be described as follows.

The hyperspectral image captured with the hyperspectral sensor is converted to a reflectance value through radiometric calibration because the intensity of light is converted into a digital number (a number obtained by normalizing the relative size) in the image storage step. .

In order to convert the spectral information stored as a DN value into a reflectance value, it is corrected by applying a linear regression equation using a specially manufactured reflective cloth to show a constant reflectance.

: 파장,

: 파장

에 대한 반사율(%),

: 초분광 영상을 반사율로 보정하기 위한 기울기,

: 저장된 이미지 값(Digital Number),

: 절편값이다.here, (

: wavelength,

: wavelength

Reflectance (%) for

: Tilt to correct hyperspectral image with reflectance,

: Saved image value (Digital Number),

: It is the intercept value.

In order to effectively use hyperspectral images, radiation correction of the data must be performed to remove atmospheric absorption and scattering effects.

In addition to eliminating atmospheric effects, this process involves converting the hyperspectral data from the sensor's radiance to reflectance.

It is desirable to collect the spectral radiation measurements using an in situ spectroradiometer on the ground similar to the time when the spectral data of the object is collected with the hyperspectral sensor.

However, when performing remote sensing, access to the site may not be possible, and simultaneous operation of both hyperspectral sensors and spectroradiometers may be difficult. Therefore, use a standard reference panel (Spectralon) or a calibration panel to convert spectral data into reflectivity. Convert.

In addition, since different results may be derived for the same chemical substance depending on the optical and equipment conditions at the time of hyperspectral imaging, a reference value as a reference value when converting the measured value (DN, Radiance) into reflectance is required.

Therefore, in the present invention, when taking a hyperspectral image of a hazardous chemical, a reflective cloth made of a special material having a standard reflectivity of 55%, 44%, 22%, and 5% is photographed together with a correction panel. Radiation correction is performed to convert the recorded luminance to reflectance.

Since the used reflective cloth already knows the exact reflectivity, it can be set as a 0-100% reflectance relationship by taking pictures under the same conditions as the hyperspectral imaging of chemicals and establishing a relational expression.

For example, the relational expression applied modeling according to the normal distribution so as to rise according to the measured value of the spectroradiometer in the wavelength range of 400 to 420 nm is as follows.

는 평균,

는 분산이다.here,

Is the average,

Is the variance.

에서

값을 가져야 하므로

을 만족해야 한다. only,

in

It must have a value

Should be satisfied.

를 산정하여 방사보정에 적용한다.In the present invention, using the spectral information of the actual spectrometer

Is calculated and applied to radiation correction.

In addition, spectral information of river water is removed in order to extract the characteristics of only the hazardous chemical from the sample in which river water and hazardous chemicals are mixed.

The luminance of the liquid collected by the hyperspectral image includes bottom reflection, absorption by water columns, water surface reflection, and absorption by the atmosphere. In order to exclude the effect of luminance due to other conditions other than the characteristics of the sample, the present invention The characteristics of hazardous chemicals are separated by removing the reflectivity value of river water (solvent) taken under the same conditions as the sample, which can be called the base, from the reflectivity of the mixture.

는 파장,

는 총 휘도,

는 바닥반사,

는 물기둥에 의한 흡수,

는 수표면 반사,

는 대기에 의한 흡수이다.here,

Is the wavelength,

Is the total luminance,

Is the floor reflection,

Is absorption by the water column,

Is the water surface reflection,

Is absorption by the atmosphere.

는 하천수와 유해화학물질 혼합물의 반사도,

는 하천수의 반사도이다.here,

Is the reflectivity of the mixture of river water and hazardous chemicals,

Is the reflectivity of river water.

자료만을 남기고, 아웃라이어에 해당하는 반사도를 포함한 스펙트럼을 제거한다.And the hyperspectral data with each reflectivity value at n wavelengths are used within the reliability interval of 99.73% by using the standard deviation at each wavelength.

Only the data is left, and the spectrum including the reflectivity corresponding to the outlier is removed.

It is natural that the reliability interval can be set differently.

The effective reflectivity range is as in Equation 6.

는 파장,

는 Super Smoother 처리된 화학물질의 반사도를 나타낸다.here,

Is the wavelength,

Represents the reflectivity of super smoother-treated chemicals.

Hereinafter, the characterization of the spectral library for improving the ability to identify chemical substances using the spectral library will be described in detail.

3 is a spectral spectrum comparison graph.

The method of classifying hyperspectral images using a spectral library is being used in various studies to identify and classify indicator materials such as land cover classification.

The spectral library of hazardous chemicals constructed in the present invention is also a comparison of the spectral library and 18 kinds of hazardous chemicals with hyperspectral images taken with a time difference of 2 hours in order to identify and classify chemicals in hyperspectral images. .

3 is a spectral library (Lib) of (1) hydrofluoric acid, (2) bromine, (3) phosphorus oxychloride, and (4) ammonium difluoride, respectively, and a spectral spectrum (New HSI) extracted from a new hyperspectral image.

In addition, in order to compare the spectral library constructed for the same material and the spectral spectrum extracted from the new hyperspectral image, an example of calculating the correlation coefficient of the two spectral spectra is shown in Table 1.

The chromatic bromine showed a high correlation of 0.979, but the remaining 17 kinds of hazardous chemicals had a correlation coefficient of 0.9 or less, and in some cases, a negative correlation.

The present invention is for identifying hazardous chemical substances using a spectral library, and includes a configuration for upgrading a spectral library to enable identification of chemical substances and increase the recognition rate.

That is, for each hazardous chemical substance, the wavelength band showing the characteristic reflectivity of each substance is selected, and the correlation coefficient between the characteristic wavelength band of the spectral library and the new spectral spectrum is calculated to determine whether to identify it.

In addition, the recognition rate is calculated to evaluate the accuracy of the characteristic spectral library.

Recognition rate is, for example, when the new spectral spectrum is compared with all the characteristic spectral libraries of 18 kinds of hazardous chemicals, the correlation coefficient is calculated, and the correlation coefficient is ranked, and recognition is possible if the target substance is ranked within 3 rankings. I judge that it is.

In the actual identification, the possibility of the substance being ranked 1st, 2nd, and 3rd is suggested, and if chemical substances handled by nearby factories or facilities near rivers are identified, it will be possible to identify spilled chemicals.

In an embodiment of the present invention, a characteristic spectral library is produced with a target of 80% recognition rate.

As a result of evaluating the recognition rate using the first constructed characteristic spectral library, if the recognition rate of the library does not reach 80%, a trial and error method is performed until the recognition rate reaches 80% to satisfy the recognition rate of 80%.

When evaluating the recognition rate, it is evaluated that it is recognizable if the correlation coefficient rank is within the 3rd rank, but when constructing a characteristic spectral library, it is desirable to adjust the correlation coefficient of the hazardous chemical as much as possible so that it becomes the 1st rank.

Substances that do not fall within the 3rd rank in the 1st characteristic spectral library are improved to be within the 3rd rank by modifying the set wavelength range, and the harmful chemicals with the 1st rank correlation coefficient have the wavelength range widened to widen the difference in correlation coefficient with the 2nd rank substance. Modify it.

The 2nd, 3rd, 4th, etc. process is continuously performed, and if the recognition rate does not change as a result of the recognition rate evaluation, the standard feature library is determined.

In addition, in order to build a characteristic spectral library, for example, a total wavelength of 400 to 1,000 nm is set as a group of five adjacent wavelengths, and 146 wavelength groups are set.

As shown in Equation 8, the correlation coefficient between the spectral library of the set group and the new spectral spectrum is calculated, and all 146 wavelength groups are performed in the same manner.

와

는 각각 분광라이브러리의 반사도와 신규 초분광영상에서 추출한 분광 스펙트럼의 반사도의 평균값이며,

와

는 분광라이브러리와 신규 초분광영상에서 추출한 분광 스펙트럼의 반사도의 표준편차이다.here,

Wow

Is the average value of the reflectivity of each spectral library and the reflectance of the spectral spectrum extracted from the new hyperspectral image,

Wow

Is the standard deviation of the reflectivity of the spectral spectrum extracted from the spectral library and the new hyperspectral image.

Table 2 shows the reflectivity correlation coefficient for each group of sodium fluoride.

In addition, a first-order characteristic spectral library is constructed by designating a group having a higher correlation coefficient among the set wavelength groups as a characteristic section.

Different wavelength groups are set for each hazardous chemical, and the number of wavelength groups is also set differently.

As an example, the primary characteristic spectroscopic library of sodium fluoride was set to 5 characteristic groups, as shown in FIG. 4, and the correlation coefficient for each section is shown in Table 3.

Table 4 shows the results of calculating the correlation coefficient of the new spectrum of sodium fluoride and the characteristic spectroscopic library of 18 types of hazardous chemicals.

In the same way, the new spectrum of 18 kinds of hazardous chemicals is calculated as shown in Table 5, and the correlation coefficient with 18 kinds of characteristic spectral libraries, respectively, and the recognition rate is determined as shown in Table 6.

Except for toluene, all 17 hazardous chemicals showed correlation coefficient values of 0.8 or higher. In addition, as a result of calculating the recognition rate, 10 of the 18 hazardous chemicals were recognized, showing a recognition rate of 55.6%, and the recognition rate was evaluated using the total wavelength of 6 types and 33.3% when recognized as the first target material. The recognition rate was greatly improved than when it was done.

Sodium fluoride was judged to have the 7th correlation coefficient ranking and needed to be improved, and the recognition rate set by the present invention should be satisfied by raising the recognition rate by modifying all the primary characteristic spectral libraries of 18 hazardous chemicals. do.

In addition, in the case of hazardous chemicals that are determined to require improvement of the characteristic spectral library as a result of the correlation calculation and recognition rate evaluation of the primary characteristic spectral library, the characteristic spectral library is modified by adding or excluding the wavelength group having the next-order correlation coefficient.

As a result of calculating the recognition rate using the modified secondary characteristic spectral library, 14 of the 18 hazardous chemicals were recognized, resulting in a 77.8% recognition rate, and 11 types were recognized as the first target material. The recognition rate was improved at 66.1%, compared to when the recognition rate was evaluated using the primary characteristic spectral library.

In order to satisfy the recognition rate value set as the standard in the present invention and avoid confusion with the material recognized as the second priority, the improvement of the characteristic spectral library is performed again.

In the same way as when constructing the secondary characteristic spectral library, modify the characteristic spectral library by adding or excluding the wavelength group having the next-order correlation coefficient of the hazardous chemical that is determined to need improvement.

In the case of sodium fluoride as an example, as shown in Table 7, the primary characteristic spectral library used 5 wavelength bands, but the secondary characteristic spectral library used 4 wavelength bands, and the third characteristic spectral library used 3 wavelength bands. do.

As a result of calculating the recognition rate using the modified tertiary characteristic spectral library, all 18 substances out of 18 hazardous chemicals were recognized, showing a recognition rate of 100.0%, and 16 cases were recognized as the first target substance. As 88.9%, it can be seen that the recognition rate is improved compared to when the recognition rate is evaluated using the secondary characteristic spectral library.

In addition, although the recognition rate value set as a reference in the present invention is satisfied, in order to avoid confusion with the material recognized as the second priority, the improvement of the characteristic spectral library is performed again to verify the set section.

The finally constructed sodium fluoride characteristic spectroscopic library is shown in FIG. 5.

The result of calculating the recognition rate of all hazardous chemicals using the final characteristic library is shown in Table 8. All 17 substances except ammonium difluoride were recognized as the first priority, showing a recognition rate of 94.4%.

If viewed as the third priority, which is the recognition rate standard set in the embodiment of the present invention, all substances were recognized and showed a recognition rate of 100.0%.

Table 9 shows the result of calculating the recognition rate for each order of the characteristic spectral library.

The apparatus and method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed in river water using the hyperspectral image according to the present invention described above are accurate by constructing a spectroscopic library for standard characteristics of hazardous chemical substances mixed in river water. The high level of identification is possible, so that when a chemical accident occurs, the occurrence of the accident can be immediately identified and the accident substance can be quickly identified.

The present invention constructs a standard characteristic spectral library of hazardous chemicals mixed in river water and applies it to real-time monitoring of hazardous chemicals, thereby enabling efficient detection and response of chemical accidents.

As described above, it will be understood that the present invention is implemented in a modified form without departing from the essential characteristics of the present invention.

Therefore, the specified embodiments should be considered from a descriptive point of view rather than a limiting point of view, and the scope of the present invention is indicated in the claims rather than the above description, and all differences within the scope equivalent thereto are included in the present invention. It will have to be interpreted.

100. Group correlation coefficient comparison unit
200. Group Selection Department
300. Recognition Rate Evaluation Department
400. Standard characteristic spectroscopic library confirmation section

Claims (13)

A correlation coefficient comparison unit for each group comparing the correlation coefficient for each group between the river water mixed hazardous chemical spectral library constructed by the river water mixed hazardous chemical spectral library construction means and the newly photographed hazardous chemical spectral spectrum;
A group selection unit for selecting a group belonging to an upper value set as a result of the comparison of the correlation coefficient comparison unit for each group;
A recognition rate evaluating unit for evaluating a recognition rate between samples using the reflectance of the group selected by the group selecting unit, and performing re-performing from the group selecting step when the recognition rate is not reached;
To improve the ability to identify hazardous chemicals mixed in river water using hyperspectral images, comprising: a standard characteristic spectral library determination unit that determines the standard characteristic spectral library of hazardous chemicals mixed with river water when the finally set recognition rate is satisfied. A device for characterizing spectral libraries. The method of claim 1, wherein the river water mixed hazardous chemical substance spectroscopic library construction means,
A hyperspectral image capture unit for taking a hyperspectral image of the prepared sample,
In the image storage step, the hyperspectral image radiation correction unit converts the stored hyperspectral image into a reflectance value through radiometric calibration by converting the intensity of light into a normalized digital number (DN). ,
A spectral spectrum extraction unit that extracts a spectral spectrum of a region that is not affected by light scattering as much as possible from a sample of hazardous chemical substances mixed with river water taken from a hyperspectral image;
A hazardous chemical property extracting unit that removes spectral information of river water in order to extract the characteristics of only the hazardous chemical from a sample in which river water and hazardous chemicals are mixed;
An outlier removal unit for removing outliers by reviewing the effectiveness of the acquired spectral spectrum,
A spectral spectrum determination unit that determines one spectral spectrum by super smoother processing the spectral spectra remaining after the outlier removal in the outlier removal unit, and
Identification of hazardous chemicals mixed in river water using hyperspectral images, characterized in that it includes a vector normalization unit that performs vector normalization in order to construct a final library of confirmed spectral spectra and compare each hazardous chemicals with each other. A device for characterizing spectral libraries to improve capabilities. Constructing a river water mixed hazardous chemicals spectral library using hyperspectral images in a means for building a river water mixed hazardous chemicals spectral library;
Comparing the correlation coefficient for each group between the river water mixed hazardous chemicals spectral library constructed by the river water mixed hazardous chemicals spectral library construction means and the newly photographed hazardous chemicals spectral spectrum;
Selecting a group belonging to an upper value set as a result of the comparison in the group correlation coefficient comparison step;
Evaluating the recognition rate between samples by using the reflectance of the selected group, and re-performing from the group selection step if the recognition rate is not reached;
When the finally set recognition rate is satisfied, the step of determining a standard characteristic spectral library of hazardous chemical substances mixed with river water; for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed with river water using hyperspectral images, characterized in that it includes Way. The method of claim 3, wherein in the step of re-performing from the group selection step when the set recognition rate is not reached,
Substances that do not fall within the 3rd rank in the 1st characteristic spectral library are improved to be within the 3rd rank by modifying the set wavelength range, and the harmful chemicals with the 1st rank correlation coefficient have the wavelength range widened to widen the difference in correlation coefficient with the 2nd rank substance. Modified,
Characterization of a spectral library to improve the ability to identify hazardous chemicals mixed with river water using hyperspectral images, characterized by continuously performing the process by increasing the order and determining a standard characteristic library when the recognition rate does not change as a result of the recognition rate evaluation Way for. The method of claim 3, wherein the evaluation of the recognition rate between samples using the reflectance of the selected group,

A method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed in river water using hyperspectral images, characterized in that performing as a method. The method of claim 3, wherein in the step of comparing the correlation coefficient for each group between the constructed spectral library of mixed hazardous chemicals of river water and the spectral spectrum of newly photographed hazardous chemicals,
The correlation coefficient between the spectral library of the set group and the new spectral spectrum,

Is calculated as,
here,

Wow

Is the average value of the reflectivity of each spectral library and the reflectance of the spectral spectrum extracted from the new hyperspectral image,

Wow

Is a method for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed in river water using a hyperspectral image, characterized in that it is the standard deviation of the reflectivity of the spectroscopic library and the spectroscopic spectrum extracted from the new hyperspectral image. The method of claim 6, wherein a group whose correlation coefficient is ranked higher among the set wavelength groups is designated as a characteristic section to construct a first-order characteristic spectral library,
A method for characterizing a spectral library for improving the ability to identify mixed hazardous chemicals in river water using hyperspectral images, characterized in that a wavelength group in a different section is set for each hazardous chemical, and the number of wavelength groups is set differently. The characteristic spectral library according to claim 7, wherein in the case of a hazardous chemical that is determined to require improvement of the characteristic spectral library as a result of the correlation calculation and recognition rate evaluation of the primary characteristic spectral library, a wavelength group having a correlation coefficient of the next order is added or excluded. And modify
Characterization of spectral libraries for improving the ability to identify hazardous chemicals mixed with river water using hyperspectral images, characterized in that this configuration is applied in the same way to the correlation calculation and recognition rate evaluation of the characteristic spectral library of increased order later. Way for you. The method of claim 3, wherein the step of constructing a spectral library for hazardous chemical substances mixed with river water by using a hyperspectral image comprises:
Preparing a sample of hazardous chemical substances and photographing a hyperspectral image in a hyperspectral imaging unit,
Converting the spectral information stored as a DN (Digital Number) value in the hyperspectral image radiation correction unit into a reflectance value through radiometric calibration, and
Extracting a spectral spectrum of a region not affected by light scattering from a sample of hazardous chemical substances mixed with river water taken from the hyperspectral image by the spectral spectrum extraction unit,
Removing the spectral information of the river water in order to extract the characteristics of the hazardous chemicals only from the samples in which the river water and the hazardous chemicals are mixed in the hazardous chemical property extraction unit; and
Determining the validity of the spectral spectrum obtained by the outlier removal unit and removing outliers; and
Comprising the step of constructing a spectral library by processing the spectral spectra remaining after removing the outlier in the spectral spectrum determination unit as a single spectral spectrum by super smoother processing, and performing vector normalization of the spectral spectra determined in the vector normalization unit. A method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed in river water using hyperspectral images, characterized in that. The method of claim 9, wherein in the step of converting to a reflectance value,
In order to convert the spectral information stored as a DN value into a reflectance value, a linear regression equation is applied and corrected using a reflective cloth representing a constant reflectance.

here,

: wavelength,

: wavelength

Reflectance (%) for

: Tilt to correct hyperspectral image with reflectance,

: Saved image value (Digital Number),

: A method for characterizing a spectral library for improving the ability to identify hazardous chemical substances mixed with river water using hyperspectral images characterized by being an intercept value. The method of claim 9, wherein when photographing a hyperspectral image of a hazardous chemical substance, radiation correction is performed to convert a brightness recorded by a hyperspectral sensor into a reflectance by photographing a standard reflectivity together with a reflective cloth,
The relational expression that applied modeling according to the normal distribution so as to rise according to the measured value of the spectroradiometer in the wavelength range of 400 to 420 nm is:

ego,
here,

Is the average,

Is the variance, provided that

in

It must have a value

A method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed in river water using hyperspectral images, characterized in that satisfying. The method of claim 9, wherein in the step of removing spectral information of river water in order to extract characteristics of only the corresponding hazardous chemical,
Full luminance

Defined as,
here,

Is the wavelength,

Is the total luminance,

Is the floor reflection,

Is absorption by the water column,

Is the water surface reflection,

Is absorption by the atmosphere,

To extract the characteristics of hazardous chemicals,
here,

Is the reflectivity of the mixture of river water and hazardous chemicals,

A method for characterizing a spectroscopic library for improving the ability to identify hazardous chemical substances mixed in river water using hyperspectral images, characterized in that is the reflectivity of river water. The method of claim 9, wherein in the step of removing outliers by determining the validity of the spectral spectrum, the effective reflectivity range is,

Is defined as,
here,

Is the wavelength,

Is a method for characterizing a spectroscopic library to improve the ability to identify hazardous chemical substances mixed in river water using hyperspectral images, characterized in that the reflectivity of chemical substances treated with Super Smoother.

Images