KR102501581B1 - Method for predicting concentration of pathogenic microorganism using remote sensing hiperspectral images - Google Patents

Abstract

In the present invention, a first water quality index is calculated using a remote sensing hyperspectral image, a water quality database is constructed using the first water quality index and the actually measured concentration data of pathogenic microorganisms, and then the water quality database is used for actual measurement. The main purpose is to provide a method that can predict the concentration of pathogenic microorganisms in untreated water systems.
To achieve the above object, the method for predicting the concentration of pathogenic microorganisms using a hyperspectral image for remote sensing of the present invention includes the steps of acquiring a hyperspectral image of a target water system through remote imaging; Calculating a first water quality index composed of one or more water quality items that can be calculated through a regression analysis model using the obtained hyperspectral image; constructing a water quality database by combining the calculated primary water quality index and actually measured concentration data of pathogenic microorganisms in the target water system as one set; and predicting the concentration of pathogenic microorganisms that have not been actually measured using a first water quality indicator calculated from a new hyperspectral image taken remotely from the target water system and a machine learning model trained through the water quality database.

Description

A method for predicting the concentration of pathogenic microorganisms using remote sensing hyperspectral imagery

The present invention relates to a method for predicting the concentration of pathogenic microorganisms using remote sensing hyperspectral imagery, and more particularly, to a water quality database using a primary water quality index that can be calculated using remotely monitored hyperspectral imagery and the actually measured concentration of pathogenic microorganisms. and how to accurately predict the concentration of pathogenic microorganisms in the future using this water quality database.

Recently, interest in water sports such as yachting, water skiing, canoeing, and rowing has increased, and as waterside spaces in rivers have been improved, the number of families playing water has increased. However, since water activities in rivers may infect users with water-borne diseases through contact with water bodies, safety for water quality must be ensured. To this end, water quality information on rivers, etc. system is being developed.

Conventional water quality information has mainly been collected by actually measuring water quality information data targeting several survey points due to the characteristics of topography or water system. This water quality inspection method has a problem in that it is difficult to grasp the comprehensive water quality of the entire water system. Moreover, the concentration of pathogenic microorganisms, which are the main cause of water-borne diseases, is collected through sampling in the field, and then cultured for 18 to 24 hours after administering the water quality sample to a medium (plate, Colilert, Enterolert, etc.) It has been measured by reading the concentration of microorganisms, and this analysis process not only consumes considerable time and manpower, but also has limitations in providing water quality information to the general public in real time due to the time delay according to the analysis.

In order to solve this problem, a method of quickly monitoring water quality information using a video image remotely photographed from an aircraft has been developed. An example thereof is disclosed in Korean Patent Registration No. 10-1116462 (title of invention: method for monitoring water quality using remote sensing data and apparatus for monitoring water quality using the same) (Patent Document 1). According to this, a regression analysis model (correlation function) such as transparency, chlorophyll a concentration, and total phosphorus concentration was created using remote sensing image data and actual measurement data, and this regression analysis model was used to determine the transparency and chlorophyll a A method of obtaining data of at least one of concentration and total phosphorus concentration, determining a eutrophication index using this data, and evaluating the water quality of a water system using the determined eutrophication index.

However, since the above-mentioned conventional water quality monitoring method uses a video image taken of a water system at a very high altitude, such as from an aircraft, in order to remove noise from the video image, it is determined according to the sun altitude, atmospheric condition, ground surface cover state, and geometric topography. It had to go through a complex correction process. Moreover, only water quality indicators correlated with remote sensing images, such as transparency, chlorophyll a concentration, and total phosphorus concentration, can be predicted, and the concentration of pathogenic microorganisms, which are the main cause of waterborne infectious diseases, cannot be predicted.

KR 10-1116462 B

The present invention was developed to solve these conventional problems, and a first water quality index is calculated using a hyperspectral image of remote sensing, and a water quality database is obtained using the first water quality index and the actually measured concentration data of pathogenic microorganisms. After constructing the water quality database, the main purpose is to provide a method for predicting the concentration of pathogenic microorganisms in unmeasured water systems using the water quality database.

To achieve the above object, the method for predicting the concentration of pathogenic microorganisms using a hyperspectral image for remote sensing of the present invention includes the steps of acquiring a hyperspectral image of a target water system through remote imaging; Calculating a first water quality index composed of one or more water quality items that can be calculated through a regression analysis model using the obtained hyperspectral image; constructing a water quality database by combining the calculated primary water quality index and actually measured concentration data of pathogenic microorganisms in the target water system as one set; and predicting the concentration of pathogenic microorganisms that have not been actually measured using a first water quality indicator calculated from a new hyperspectral image taken remotely from the target water system and a machine learning model trained through the water quality database.

In addition, in the hyperspectral image acquisition step, the target water system may be closely photographed using an unmanned aerial vehicle or a drone.

In addition, the water quality items include water depth, turbidity, total suspended solids (TSS), volatile suspended solids (VSS), phytoplankton, colored dissolved organic matter (CDOM), non-algal particles (NAP), total phosphorus (TP), total nitrogen (TN) ), one or more of Cyanobacteria, Microcystin, Phycocyanin, and Chlorophyll-a.

In addition, the pathogenic microorganism may be one or two or more of fecal coliform, enterococci, Staphylococcus aureus, Vibrio enteritis, Salmonella, Bacillus cereus, Clostridium perfringens, enterovirus, diphtheria, and Shigella.

In addition, the machine learning model may be characterized in that it is any one of an artificial neural network model and a support vector machine model.

In addition, the step of constructing the water quality database may further include automatically correcting the machine learning model using an additional water quality database to increase prediction accuracy.

According to the method for predicting the concentration of pathogenic microorganisms using remote sensing hyperspectral imagery of the present invention configured as described above, the concentration of pathogenic microorganisms in a target water system can be predicted in real time using hyperspectral imagery taken by an unmanned aerial vehicle or a drone. This enables rapid forecasting of water quality information.

In addition, since the concentration of pathogenic microorganisms can be accurately predicted without actually measuring the concentration of pathogenic microorganisms, it is possible to reduce time, manpower, and cost previously required to actually measure the concentration of pathogenic microorganisms.

In addition, as the data of primary water quality indicators that can be built in real time using unmanned aerial vehicles or drones is vast, it is possible to continuously calibrate and improve the accuracy of the model. As a result, the prediction accuracy for the concentration of pathogenic microorganisms is further improved over time.

1 is a flow chart showing a method for predicting the concentration of pathogenic microorganisms according to the present invention.
2 is a diagram illustrating a first water quality index calculation process according to the present invention.
3 is a diagram illustrating the results of calculating the first water quality index according to the present invention.
4 is a diagram illustrating a machine learning process according to the present invention.

Hereinafter, a method for predicting the concentration of pathogenic microorganisms using a remote sensing hyperspectral image according to the present invention will be described in more detail with reference to the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms, only the present embodiments make the disclosure of the present invention complete, and common knowledge in the art to which the present invention pertains. It is provided to completely inform the person who has the scope of the invention, and the present invention is only defined by the scope of the claims. For reference, in describing the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

The main steps of the present invention will be described in detail with reference to FIG. 1, which is a flow chart showing a method for predicting the concentration of pathogenic microorganisms according to the present invention.

First, a hyperspectral video image of a target water system is acquired through remote shooting (S10). The remote photographing is generally possible through a remote photographing means such as an aircraft or a satellite. The video image obtained by this long-distance shooting has the advantage that it can be used to measure the water quality of large-area water systems such as lakes and coasts. It has a disadvantage that it needs to go through a complicated correction process according to the method, and the prediction accuracy of the water quality information is low.

Considering this point, it is preferable to use a video image taken from a short distance of the target water system using a small unmanned aerial vehicle or a drone. The biggest reason is that there is no need to go through the complicated correction procedure that was performed on video images taken from a distance.

On the other hand, the hyperspectral image is an image captured by a hyperspectral sensor, and allows the spectral characteristics of an object to be captured to be expressed through hundreds of spectral bands that are continuous and narrow in wavelength. says Since each spectral band represents the spectral characteristics of one atom or a specific substance, the hyperspectral sensor has the advantage of being able to simultaneously detect several substances. In addition, conventionally, water quality information was acquired through one-point body water, but there is also an advantage that surface-level monitoring is possible by using a hyperspectral sensor. As a result, the spatial distribution and movement of contaminants on the surface of the water system can be more clearly identified.

Next, a first water quality index composed of one or more water quality items that can be calculated through a regression analysis model is calculated using the obtained hyperspectral image (S20). As described above, Patent Document 1 discloses a method of acquiring water quality items such as transparency, concentration of chlorophyll-a, and total phosphorus through a regression analysis model using a remote sensing hyperspectral image.

In the present invention, more diverse water quality items are calculated as primary water quality indicators in order to predict the final target, the concentration of pathogenic microorganisms that cause water-borne infectivity. For example, water depth, turbidity, total suspended solids (TSS), volatile suspended solids (VSS), phytoplankton, colored dissolved organic matter (CDOM), non-algal particulates (NAP), total phosphorus (TP), total nitrogen (TN), One or more water quality items among Cyanobacteria, Microcystin, Phycocyanin, and Chlorophyll-a are used as primary water quality indicators.

If the number of water quality items used as the primary water quality index increases, it is possible to build a more extensive water quality database, thereby increasing the accuracy of prediction. Furthermore, it is possible to further improve the prediction accuracy of a specific pathogenic microorganism by selectively using a water quality item that has been proven to be associated with a specific pathogenic microorganism.

2 and 3 briefly show the first water quality index calculation process and results according to the present invention. Change characteristics over time of a specific spectral band obtained using a hyperspectral sensor as shown in FIG. 2 (a) and change characteristics over time of the concentration data of Chlorophyll-a measured as shown in FIG. shows the same change trend when comparing . Considering the correlation between the specific spectral band and Chlorophyll-a, as shown in (c) of FIG. 2, it can be seen that the specific spectral band and Chlorophyll-a have a first-order direct proportional relationship.

In this way, the distribution of Chlorophyll-a concentration in the water system can be predicted by analyzing the hyperspectral image obtained by taking the entire target water system area by surface using the regression analysis model using the high correlation between the specific spectral band and Chlorophyll-a. . Figure 3 compares the predicted data distribution and the actual data distribution for the concentration of chlorophyll-a with the same contour line. According to this, it can be seen that relatively accurate concentration prediction is possible through the regression analysis model of the hyperspectral image and water quality items.

Next, a water quality database is constructed by combining the calculated primary water quality index and the actually measured concentration data of pathogenic microorganisms in the target water system as one set (S30).

Pathogenic microorganisms refer to microorganisms such as bacteria and viruses that can transmit disease to humans. In particular, pathogenic microorganisms that cause water-borne infectious diseases include fecal coliform, enterococci, Staphylococcus aureus, Vibrio enteritis, Salmonella, Bacillus cereus, Clostridium perfringens, enterovirus, diphtheria, and dysentery. The present invention provides a method for predicting in real time the concentration of one or two or more of the pathogenic microorganisms at risk of developing water-borne infectious diseases in a target water system.

To this end, the concentrations of pathogenic microorganisms actually measured from the target water system at the same time as the plurality of primary water quality indicators calculated from the hyperspectral image taken by remote sensing are grouped into a database as one set. For example, Bacillus cereus, which is present in a high proportion in crops and vegetables including grains and causes water-borne diarrhea, dizziness and abdominal pain, has a clear relationship with the content of colored dissolved organic matter (CDOM) and total phosphorus (TP) among water quality items. If there is a correlation, data values of a plurality of water quality items, including colored dissolved organic matter (CDOM) and total phosphorus (TP), and the actually measured concentration of Bacillus cereus are set as a water quality database. When a vast amount of water quality database is established through this iterative process, it is possible to accurately predict the concentration of pathogenic microorganisms without actually measuring it.

According to the present invention, a step of automatically correcting the machine learning model using an additional water quality database may be further included in order to increase the accuracy of prediction (S40). As described above, the water quality database is a collection and storage of primary water quality indicators obtained from hyperspectral imagery and concentrations of pathogenic microorganisms actually measured. By calibrating, higher accuracy can be secured. The machine learning model refers to a method in which a computer derives an accurate result value by deriving an optimal function through repetitive training by imitating a human learning process. Representatively, an artificial neural network and a support vector machine ( Support Vector Machine) and the like, and details thereof will be described later with reference to FIG. 4 .

Finally, the concentration of pathogenic microorganisms that have not been actually measured is predicted using the first water quality index calculated from the new hyperspectral image taken remotely in the target water system and the machine learning model repeatedly trained from the water quality database (S50).

That is, a plurality of predetermined primary water quality indicators are calculated from the hyperspectral video image remotely photographed using an unmanned aerial vehicle or drone through the step S20. The water quality database is searched using the plurality of primary water quality indicators, the primary water quality indicators having the closest data values are extracted, and the concentration values of pathogenic microorganisms grouped together with the primary water quality indicators are output. As a result, it is possible to calculate and forecast the concentration of pathogenic microorganisms in the target water system in real time without actually measuring the concentration.

In the step of predicting the concentration of pathogenic microorganisms, the concentration of the pathogenic microorganisms may be predicted using the machine learning model. In other words, by applying a machine learning model such as artificial intelligence to the process of extracting the primary water quality index with the closest data value, more accurate prediction is possible. In particular, it is generally known that the accuracy of prediction through machine learning is higher than that of the regression model.

4 schematically shows two representative methods of the machine learning model. First, the artificial neural network model partially mimics the human neural network by simplifying it. Among them, links with high correlation are connected with weights, and each neuron receives information from several neurons, but as a result, only one value is output.

The support vector machine model is a method mainly used for classification and regression analysis, and is composed of a hyperplane or a set of hyperplanes. Intuitively, since the hyperplane has a large difference from the nearest training data point and the classification error is small, for a good classification, find the hyperplane that has the furthest distance from the closest training data point for any classified point. That is, learning is performed while maximizing the distance between the two classes on the N-dimensional space and searching for the N-1-dimensional hyperplane that positions the data belonging to the same class on the same side.

Since the above two machine learning models exemplify the most commonly used methods, it is possible to increase the accuracy of predicting the concentration of pathogenic microorganisms according to the present invention, and it is natural that other machine learning models are also applicable.

Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art to which the present invention pertains can be implemented in other specific forms without changing the technical spirit or essential features of the present invention. will be able to understand Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. The scope of the present invention is indicated by the claims to be described later rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts thereof should be construed as being included in the scope of the present invention. do.

Claims (6)

Obtaining a hyperspectral image of a target water system through remote imaging;
Calculating a first water quality index composed of one or more water quality items that can be calculated through a regression analysis model using the obtained hyperspectral image;
determining a correlation between the one or more water quality items constituting the calculated primary water quality index and actually measured concentration data of pathogenic microorganisms in the target water system;
constructing a water quality database by combining the one or more water quality items and the concentration data of the pathogenic microorganisms into one set according to the correlation; and
Predicting the concentration of pathogenic microorganisms that have not been actually measured using a primary water quality index calculated from a new hyperspectral image image taken remotely in the target water system and a machine learning model trained through the water quality database;
The pathogenic microorganisms are remote sensing seconds, characterized in that one or two or more of fecal coliform, enterococci, Staphylococcus aureus, Vibrio enteritis, Salmonella, Bacillus cereus, Clostridium perfringens, enterovirus, diphtheria, and Shigella. A method for predicting the concentration of pathogenic microorganisms using spectroscopic images.
The method of claim 1,
The hyperspectral image acquisition step is a method for predicting the concentration of pathogenic microorganisms using a remote sensing hyperspectral image image, characterized in that for taking a close-up of the target water system using an unmanned aerial vehicle or a drone. The method of claim 1,
The water quality items include water depth, turbidity, total suspended solids (TSS), volatile suspended solids (VSS), phytoplankton, colored dissolved organic matter (CDOM), non-algal particles (NAP), total phosphorus (TP), total nitrogen (TN), A method for predicting the concentration of pathogenic microorganisms using remote sensing hyperspectral image images, characterized in that one or more of Cyanobacteria, Microcystin, Phycocyanin, and Chlorophyll-a. delete The method of claim 1,
The machine learning model is an artificial neural network model or a support vector machine model. The method of claim 1,
The water quality database construction step further comprises automatically correcting the machine learning model using an additional water quality database to increase the accuracy of the prediction.

Images