CN112464746A - Water quality monitoring method and system based on satellite images and machine learning - Google Patents

Water quality monitoring method and system based on satellite images and machine learning Download PDF

Info

Publication number
CN112464746A
CN112464746A CN202011243483.XA CN202011243483A CN112464746A CN 112464746 A CN112464746 A CN 112464746A CN 202011243483 A CN202011243483 A CN 202011243483A CN 112464746 A CN112464746 A CN 112464746A
Authority
CN
China
Prior art keywords
water quality
satellite image
data
image data
reflectivity
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202011243483.XA
Other languages
Chinese (zh)
Other versions
CN112464746B (en
Inventor
何炜琪
李继影
吴志杰
董世元
张仁泉
黄佳慧
薛媛媛
郭超硕
陈蓉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Suzhou Environmental Monitoring Center
Research Institute For Environmental Innovation (suzhou) Tsinghua
Original Assignee
Jiangsu Suzhou Environmental Monitoring Center
Research Institute For Environmental Innovation (suzhou) Tsinghua
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu Suzhou Environmental Monitoring Center, Research Institute For Environmental Innovation (suzhou) Tsinghua filed Critical Jiangsu Suzhou Environmental Monitoring Center
Priority to CN202011243483.XA priority Critical patent/CN112464746B/en
Publication of CN112464746A publication Critical patent/CN112464746A/en
Application granted granted Critical
Publication of CN112464746B publication Critical patent/CN112464746B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/18—Water
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G06V10/247—Aligning, centring, orientation detection or correction of the image by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Multimedia (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Databases & Information Systems (AREA)
  • Remote Sensing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Chemical & Material Sciences (AREA)
  • Image Processing (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Food Science & Technology (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Medicinal Chemistry (AREA)
  • Analytical Chemistry (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Astronomy & Astrophysics (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)

Abstract

The invention discloses a water quality monitoring method based on satellite images and machine learning, which comprises the following steps: respectively preprocessing the acquired image data of the first satellite and the acquired image data of the second satellite to generate reflectivity data; establishing radiation normalization models of different sensors according to different seasons and different ground objects; establishing a space-time fusion model based on weight filtering; establishing a water quality parameter inversion database based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms; and outputting a water quality inversion result according to the generated reflectivity image data set and the established water quality parameter inversion model. A water quality parameter inversion model based on machine learning is established, the model precision is high, and the obtained water quality parameter inversion result can reflect the spatial distribution condition of the water quality parameters.

Description

Water quality monitoring method and system based on satellite images and machine learning
Technical Field
The invention belongs to the technical field of urban surface water environment water quality monitoring, and particularly relates to a long-time continuous surface water quality monitoring method and system based on satellite images and machine learning.
Background
Surface water environment water quality monitoring is one of main contents of environment monitoring, accurately, timely and comprehensively reflects the current water quality situation and development trend, provides scientific basis for water environment management, pollution source control, environment planning and the like, and plays a vital role in the aspects of water environment protection, water pollution control and water environment health maintenance. At present, the surface water environment water quality monitoring mainly comprises the following methods:
the traditional urban surface water environment water quality monitoring method mainly adopts manual and automatic site monitoring, the time and labor cost is high, and the point data is difficult to be well applied in the aspects of analysis of the overall water quality distribution condition of a drainage basin and a lake body, environment quality supervision and control.
Satellite remote sensing image data is used for monitoring water environment quality, for example, Chinese patent with publication number CN 108507949A discloses a river water quality monitoring method based on a high-resolution remote sensing satellite, which is characterized in that a high-resolution remote sensing image a of a river to be monitored is obtained from a ground station based on the high-resolution remote sensing satellite, and a digital topographic map of the area of the river to be monitored is obtained; the preprocessing module is used for preprocessing the high-resolution remote sensing image a to obtain n high-resolution remote sensing images c which are positioned in different wave bands and can display the complete river channel image to be monitored; the controller conducts regression analysis on the actual mass concentration of each reference position point and the reflectivity of each reference position point on each high-resolution remote sensing image c to obtain a corresponding regression equation, and then establishes a relation function between the actual mass concentration of the reference position point and the reflectivity of each reference position on the corresponding position of each high-resolution remote sensing image c based on the regression equation with the minimum discrete degree; the controller calculates the reflectivity of the monitoring point on the n high-resolution remote sensing images c, and calculates the mass concentration of the water quality parameter at the monitoring point according to the relation function. The water environment quality monitoring is carried out by satellite remote sensing image data, and a water quality parameter inversion model is generally established by utilizing water body reflectivity data and ground station water quality monitoring data. Usually, a characteristic wave band or a characteristic parameter is selected based on water color characteristics, a regression relation between the water quality parameter and an image is established through an experience or semi-experience method, a regression model mainly comprises a multiple linear model, and the internal association between the water quality parameter and water body reflectivity data is difficult to dig deeply by artificially participating in the selection of the characteristic wave band or the characteristic parameter.
The satellite remote sensing image is used for monitoring the ground environment, and the influence of the time resolution and the space resolution of the remote sensing load causes the following two problems:
a, the satellite load transit period is short, the satellite load spatial resolution is low, and effective monitoring on a tiny target ground object is difficult;
b, the image has high spatial resolution, but the satellite-borne load usually has a long visiting period, and the target ground object cannot be continuously monitored in a short time. Meanwhile, the satellite remote sensing image is also influenced by cloud cover, so that less data can be effectively utilized, and long-time sequence continuous monitoring on fine ground objects is difficult to realize. The invention is achieved accordingly.
Disclosure of Invention
Aiming at the technical problems, the invention aims to provide a water quality monitoring method and a water quality monitoring system based on satellite images and machine learning, wherein the monitoring data of a monitoring station and the spectrum data of remote sensing images are subjected to multi-source effective fusion, a water quality parameter inversion model based on machine learning is established, the model precision is high, and the obtained water quality parameter inversion result can reflect the spatial distribution condition of water quality parameters. And spatial data support is provided in the aspects of pollution source tracing, environmental quality control and the like.
The technical scheme of the invention is as follows:
a water quality monitoring method based on satellite images and machine learning comprises the following steps:
s01: acquiring first satellite image data and second satellite image data;
s02: respectively preprocessing the acquired first satellite image data and the acquired second satellite image data, and generating first satellite image reflectivity data and second satellite image reflectivity data;
s03: establishing a radiation normalization model among different sensors according to different seasons and different ground objects, assimilating the first satellite image reflectivity data and the second satellite image reflectivity data, and generating first satellite image reflectivity data and second satellite image reflectivity data which are small in radiation difference and same in wave band number;
s04: based on the assimilation results of different source image data, a space-time fusion model based on weight filtering is established by using the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data, and a long-time sequence continuous high-spatial-resolution remote-sensing reflectivity image data set is generated;
s05: based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station, screening reflectivity image data and water quality monitoring data, and establishing a water quality parameter inversion database; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
s06: and outputting a water quality parameter inversion result according to the reflectivity image data set generated in the step S04 and the established water quality parameter inversion model.
In a preferred technical solution, the establishing of the radiation normalization models of the different sensors in step S03 includes:
s31: classifying the ground objects based on the first satellite image data with high spatial resolution, and classifying the images according to the types of the ground objects;
s32: generating a first satellite image data set and a second satellite image data set according to the seasons and the ground feature classification;
s33: according to the generated image data set, a radiation normalization model of the reflectivity data of the first satellite image is established based on the correlation between adjacent wave bands, the second reflectivity data is assimilated into the first reflectivity data, and a pair of the first satellite image data and the assimilated second satellite image data is established.
In a preferred technical solution, the establishing of the weight filtering-based spatio-temporal fusion model in step S04 includes:
s41: setting a search frame in a neighborhood range by taking a target pixel as a central pixel, and selecting pixels which have spectrum and time consistency with the central pixel as similar pixels one by one in the search frame according to a set search rule;
s42: calculating the weight ratio of each similar pixel according to the difference of the similar pixel and the central pixel in the spectral dimension and the spatial dimension;
s43: and calculating the target pixel value at the prediction moment according to a set fusion rule, and calculating all pixels in the image one by one to generate the high-spatial-resolution image at the prediction moment.
In a preferred technical scheme, the step S41 of selecting, pixel by pixel in the search box according to a set search rule, a pixel having spectral and temporal consistency with the center pixel as a similar pixel includes:
calculating the difference value of each pixel and the central pixel value in each reference moment search frame, and comparing the difference value with a set threshold value;
if the difference value between each pixel and the central pixel value in the search frame is less than or equal to the threshold value, the pixel is judged to be a similar pixel, namely | F (x)i,yj,tk,B)-F(xw/2,yw/2,tk,B)|≤2ω(B)/m；
In the formula, F (x)i,yj,tkB) and F (x)w/2,yw/2,tkB) respectively at a reference time tkSimilar pixels (x) within the first satellite search boxi,yj) And a center pixel (x)w/2,yw/2) The pixel value of the B-th wave band, w represents the size of a search window in a neighborhood, omega (B) represents the standard deviation of the whole first satellite image in the B-th wave band, and m represents the classification number;
selecting similar pixels at each reference moment, and taking an intersection as a final similar pixel selection result;
in a preferred embodiment, the step S42 of calculating the weight includes:
s421: calculating the spatial distance D of the first satellite image based on the reference timeijk：
Figure BDA0002769137170000041
Wherein D isijkRepresenting the space distance between the similar pixel and the central pixel, wherein A is a constant;
s422: calculating (x) on different source imagesi,yj) Sum wave of similar pixels at all reference timeCorrelation coefficient R between segmentsij：
Figure BDA0002769137170000042
Wherein E represents a mean value, G represents a variance,
Figure BDA0002769137170000043
and
Figure BDA0002769137170000044
respectively representing similar picture elements (x)i,yj) A set of pixel values of each band and all reference times on the first satellite image and the second satellite image;
s423: considering the two factors and carrying out normalization processing to form the weight W of the similar pixelijk：
Figure BDA0002769137170000045
Wherein, Pijk＝(1-Rij)×Dijk。
In a preferred technical solution, the step S43 of calculating the target pixel value at the predicted time includes:
s431: for a certain reference time tkCalculating the predicted time t of the image datapThe formula of the central pixel value is as follows:
Figure BDA0002769137170000046
in the formula, C (x)i,yj,tkB) and C (x)i,yj,tpB) respectively denote the reference time tkAnd predicting time tpSimilar picture elements (x) within a low spatial resolution image search boxi,yj) The pixel value in the B-th band, V represents the conversion coefficient;
s432: and synthesizing a plurality of reference time prediction results to obtain a final target pixel value:
Figure BDA0002769137170000047
wherein, TkComprises the following steps:
Figure BDA0002769137170000051
in a preferred technical solution, the step S05 of establishing an inverse model of the water quality parameter includes:
based on a space-time fusion result, constructing pairwise ratios and normalized value variables between wave bands as independent variables, monitoring data as dependent variables, constructing a water quality inversion model by adopting various machine learning algorithms, and selecting an optimal precision inversion model; the model precision calculation formula is as follows:
Figure BDA0002769137170000052
where MAPE is the mean relative error, yiIn order to actually monitor the value of the current,
Figure BDA0002769137170000053
and n is the number of dependent variables of the test set.
The invention also discloses a water quality monitoring system for satellite image and machine learning, which comprises:
the remote sensing image acquisition module is used for acquiring first satellite image data and second satellite image data;
the remote sensing image preprocessing module is used for respectively preprocessing the acquired first satellite image data and the acquired second satellite image data and generating first satellite image reflectivity data and second satellite image reflectivity data;
the different-source image radiation normalization module is used for establishing radiation normalization models among different sensors according to different seasons and different ground objects, assimilating the first satellite image reflectivity data and the second satellite image reflectivity data and generating the first satellite image reflectivity data and the second satellite image reflectivity data which are small in radiation difference and same in wave band quantity;
the remote sensing image space-time fusion module is used for establishing a space-time fusion model based on weight filtering by utilizing the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data based on the assimilation result of different source image data, and generating a long-time sequence continuous high-spatial-resolution remote sensing reflectivity image data set;
the water quality parameter inversion model module is used for screening reflectivity image data and water quality monitoring data and establishing a water quality parameter inversion database based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
and the water quality parameter inversion result output module outputs a water quality parameter inversion result according to the generated reflectivity image data set and the established water quality parameter inversion model.
In a preferred technical solution, the establishing of the radiation normalization models of different sensors in the different source image radiation normalization modules includes:
s31: classifying the ground objects based on the first satellite image data with high spatial resolution, and classifying the images according to the types of the ground objects;
s32: generating a first satellite image data set and a second satellite image data set according to the seasons and the ground feature classification;
s33: according to the generated image data set, a radiation normalization model of the reflectivity data of the first satellite image is established based on the correlation between adjacent wave bands, the second reflectivity data is assimilated into the first reflectivity data, and a pair of the first satellite image data and the assimilated second satellite image data is established.
In an optimized technical scheme, the building of a weight filtering-based space-time fusion model in the remote sensing image space-time fusion module comprises the following steps:
s41: setting a search frame in a neighborhood range by taking a target pixel as a central pixel, and selecting pixels which have spectrum and time consistency with the central pixel as similar pixels one by one in the search frame according to a set search rule;
s42: calculating the weight ratio of each similar pixel according to the difference of the similar pixel and the central pixel in the spectral dimension and the spatial dimension;
s43: and calculating the target pixel value at the prediction moment according to a set fusion rule, and calculating all pixels in the image one by one to generate the high-spatial-resolution image at the prediction moment.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention constructs a water quality parameter monitoring method based on remote sensing images, effectively fuses monitoring data of monitoring sites and spectrum data of the remote sensing images, generates water quality spatial data through a water quality parameter inversion algorithm, and provides spatial data support in the aspects of pollution source tracing, environmental quality control and the like.
2. In order to relieve the problem of the time-space contradiction of the satellite remote sensing load, a different-source image reflectivity data assimilation method system is firstly established, the high-time attribute characteristics of the sentinel No. 2 satellite image and the high-time resolution attribute characteristics of the sentinel No. 3 satellite image are effectively fused by using a space-time fusion method based on spatial filtering, remote sensing image data with the high-time resolution attribute characteristics are generated, and a long-time continuous fusion image data set is provided for urban water environment research.
3. According to the invention, a water quality parameter inversion model is established by using a fusion image data set with attribute characteristics of high time and high spatial resolution and long-time sequence water quality monitoring data and using machine learning algorithms such as random forests, neural networks and the like, compared with the traditional algorithm, the water quality parameter inversion precision is improved, and a water quality spatial distribution data set is provided for urban water environment monitoring and management.
Drawings
The invention is further described with reference to the following figures and examples:
FIG. 1 is a flow chart of a water quality monitoring method of satellite images and machine learning according to the present invention;
FIG. 2 is a schematic view of a fusion process of remote sensing images from different sources according to the present invention;
FIG. 3 is a schematic diagram of a water quality parameter inversion model construction process according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings in conjunction with the following detailed description. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.
Example (b):
a water quality monitoring system for satellite images and machine learning mainly comprises a remote sensing image acquisition module, a remote sensing image preprocessing module, a different-source remote sensing image radiation normalization module, a remote sensing image space-time fusion module, a water quality parameter inversion model module and a water quality parameter inversion result output module. The method aims to obtain a water quality parameter inversion result capable of representing the quality spatial distribution characteristics of the surface water environment by using remote sensing satellite image reflectivity data and through a water quality parameter inversion model. In order to solve the problem of contradiction between time resolution and space resolution caused by the limitation of remote sensing images by load hardware, an assimilation system of reflectivity among different sensors is constructed, a space-time fusion algorithm model based on spatial filtering is established, fusion of high-space low-time resolution images and low-space high-time resolution images is achieved, and a high-space high-time resolution image data set is generated. According to the water quality parameter inversion method, the water quality parameter inversion model based on machine learning is established, the model precision is high, and the obtained water quality parameter inversion result can reflect the spatial distribution condition of the water quality parameters.
The functions of the modules are as follows:
the remote sensing image acquisition module is used for acquiring first satellite image data and second satellite image data;
the remote sensing image preprocessing module is used for respectively preprocessing the acquired first satellite image data and the acquired second satellite image data and generating first satellite image reflectivity data and second satellite image reflectivity data;
the different-source image radiation normalization module is used for establishing radiation normalization models of different sensors according to different seasons and different ground objects and generating first satellite image reflectivity data and second satellite image reflectivity data which are small in radiation difference and same in wave band quantity;
the remote sensing image space-time fusion module is used for establishing a space-time fusion model based on weight filtering by utilizing the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data based on the assimilation result of different source image data, and generating a long-time sequence continuous high-spatial-resolution remote sensing reflectivity image data set;
the water quality parameter inversion model module is used for screening reflectivity image data and water quality monitoring data and establishing a water quality parameter inversion database based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
and the water quality parameter inversion result output module outputs a water quality parameter inversion result according to the generated reflectivity image data set and the established water quality parameter inversion model.
As shown in fig. 1, the water quality monitoring method based on satellite images and machine learning of the present invention comprises the following steps:
s01: acquiring first satellite image data and second satellite image data;
s02: respectively preprocessing the acquired first satellite image data and the acquired second satellite image data, and generating first satellite image reflectivity data and second satellite image reflectivity data;
s03: establishing a radiation normalization model among different sensors according to different seasons and different ground objects, and generating first satellite image reflectivity data and second satellite image reflectivity data which are small in radiation difference and same in wave band quantity;
s04: based on the assimilation results of different source image data, a space-time fusion model based on weight filtering is established by using the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data, and a long-time sequence continuous high-spatial-resolution remote-sensing reflectivity image data set is generated;
s05: based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station, screening reflectivity image data and water quality monitoring data, and establishing a water quality parameter inversion database; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
s06: and outputting a water quality parameter inversion result according to the reflectivity image data set generated in the step S04 and the established water quality parameter inversion model.
The specific satellite image and machine learning water quality monitoring system and method are as follows:
the method comprises the following steps: automatically crawling remote sensing load sentinel No. 2 and sentinel No. 3 image data, and acquiring a water quality monitoring result of a ground monitoring station to form an image database and a water quality monitoring database;
the first satellite (sentinel number 2) and the second satellite (sentinel number 3) image acquisition may include acquiring raw image data of the sensors for sentinel number 2 and sentinel number 3 from the european space agency official (https:// sciihub.
The load running height of the sentinel No. 2 is 786 kilometers, the time resolution is 5 days, the spatial resolution is 10, 20 and 60 meters, the spectral range covers visible light, near infrared and short wave infrared, and the load is a multispectral satellite imaging load;
the load operation height of the sentinel No. 3 is 815 kilometers, the time resolution is 1 day, the spatial resolution is 300 meters, the spectral range covers visible light and near infrared, and the sentinel is a multispectral satellite imaging load.
The original image data is obtained by a sensor load, and converted into original data of DN value by an electric signal.
Step two: the image preprocessing comprises radiation correction, geometric correction and cloud coverage removal, and reflectivity data are generated;
remote sensing image preprocessing, namely preprocessing downloaded original image data through radiation correction and geometric correction to obtain image data capable of representing surface feature reflectivity information;
the radiation correction comprises sensor correction and atmospheric correction, and the reflection characteristic and the radiation characteristic of the ground object are correctly evaluated;
sensor correction is to eliminate the influence caused by the photoelectric system of the sensor;
atmospheric correction, which is to solve the radiation transmission equation and solve approximate real reflectivity data in order to eliminate the influence caused by atmospheric water vapor concentration and aerosol density in the process of the surface feature reflectivity to the sensor;
geometric correction means that the original remote sensing image usually has severe geometric deformation, the geometric information of the original image is corrected by using an image correction method, and the image correction means that the geographic coordinate of the original image is corrected by means of a group of ground control points.
Step three: establishing radiation normalization models of different sensors, and generating sentinel No. 3 reflectivity simulation data based on the sentinel No. 2;
and establishing a radiation normalization model of different sensors, wherein the radiation normalization model is used for different sensors to cause that the reflectivity data of the sensors in the same central wavelength are different due to different spectral response functions. According to the method, a ground feature classification result is utilized to establish radiation normalization models of different source remote sensing images in different seasons, and first satellite image reflectivity data and second satellite image reflectivity data which are small in radiation difference and identical in wave band number are generated.
The spectral response function reflects the response degree of the ground object spectrum in a certain waveband range, and the spectral response functions of different sensors are inconsistent, so that the reflectivity values acquired in the same waveband range are different greatly.
The ground object classification result is that the ground object is divided into three types of vegetation, water and impervious surface based on the high-spatial-resolution sentinel No. 2 image, and the ground object classification is carried out on the sentinel No. 2 and sentinel No. 3 image data according to the classification result;
building a radiation normalization model of remote sensing images of different sources in different seasons, wherein the radiation normalization model of the reflectivity data of the sentinel No. 2 image is built based on the correlation between wave bands according to the generated image data set, namely, the sentinel No. 3 reflectivity data is assimilated into the sentinel No. 2 reflectivity data, and a sentinel No. 2 and assimilated sentinel No. 3 image data pair is built. The problem of reflectivity difference caused by sensor hardware among data is solved, the accuracy evaluation is carried out by utilizing average relative errors, the accuracy can be kept within 20%, and the following table shows radiation normalization models and the accuracy of different ground objects in different seasons, wherein Sn represents the nth wave band of the sentinel No. 2 image, and Bm represents the mth wave band of the sentinel No. 3 image.
Figure BDA0002769137170000101
Figure BDA0002769137170000111
Figure BDA0002769137170000121
Step four: establishing a space-time fusion model of sentinel No. 2 and sentinel No. 3
As shown in fig. 2, the spatio-temporal fusion model of the sentinel No. 2 and the sentinel No. 3 is to establish a spatio-temporal fusion model based on weight filtering, fuse the high spatial attribute of the sentinel No. 2 and the high temporal attribute of the sentinel No. 3, generate image data having both high spatial and high temporal attributes, and achieve the purpose of image supplementation. Comprises the following steps: selecting similar pixels, calculating the weight of the similar pixels, calculating a space-time conversion coefficient and outputting a fusion result.
The weight filtering-based spatio-temporal fusion model can generate the sentinel No. 2 image at the predicted time tp through a weight filtering-based spatio-temporal fusion algorithm according to a reference time t1 and t2 image pair which is input and simultaneously has the sentinel No. 2 and the sentinel No. 3 and a spatio-temporal fusion model which is input and predicted at the predicted time tp and only has the sentinel No. 3 image.
The method for establishing the space-time fusion model based on the weight filtering comprises the following steps: the method comprises the steps of setting a search frame in a neighborhood range by taking a target pixel as a center pixel, selecting pixels which have spectrum and time phase change consistency with the center pixel from pixel to pixel in the search frame according to a certain search rule as similar pixels, calculating the weight ratio of auxiliary information provided by each similar pixel according to the difference of the similar pixels and the center pixel in spectrum and space dimensions, finally calculating a target pixel value at a prediction time according to a set fusion rule, and calculating all pixels in an image one by one to generate a high-spatial resolution image at the prediction time.
The pixel which has spectrum and time phase change consistency with the central pixel is selected as a similar pixel from pixel to pixel in the search frame according to a certain search rule, and the method comprises the following steps: firstly, similar pixel selection is carried out aiming at each reference moment, and then intersection is taken as a final similar pixel selection result. The similar pixel selection calculation mode at the single reference moment is as follows:
|F(xi,yj,tk,B)-F(xw/2,yw/2,tk,B)|≤2ω(B)/m
in the formula, F (x)i,yj,tkB) and F (x)w/2,yw/2,tkB) respectively at a reference time tkSimilar picture element (x) in sentinel No. 2 search boxi,yj) And a center pixel (x)w/2,yw/2) And in the pixel value of the B-th wave band, w represents that the size of a search window in the neighborhood is set to be 51, omega (B) represents the standard deviation of the whole image of the sentinel No. 2 in the B-th wave band, and m represents that the classification number is set to be 3. It can be seen that, when the similar pixel is selected, based on the sentinel No. 2 image in each reference moment, a search box with a certain size is set in the neighborhood range by taking the target pixel as the center, the difference value between the pixel value of each pixel in the search box and the pixel value of the center is judged, and if the difference value is smaller than the threshold value, the pixel is the similar pixel.
The weight ratio of the auxiliary information provided by each similar pixel element is calculated according to the difference between the similar pixel element and the central pixel element in the spectral dimension and the spatial dimension, and the weight ratio comprises the following steps: and for the selected similar image elements, carrying out weight distribution on the similar image elements so as to be used for calculating the target image element value at the prediction moment. The weight calculation steps are as follows:
1) calculating space distance D based on number 2 image of sentinel at reference momentijk：
Figure BDA0002769137170000131
In the above formula DijkAnd the space distance between the similar image element and the central image element is represented, A is a constant for controlling the size of the space distance, and the width of the search box is usually set.
2) Calculating (x) on different source imagesi,yj) The correlation coefficient R of similar image elements between all reference time instants and wave bandsij：
Figure BDA0002769137170000132
Figure BDA0002769137170000133
Figure BDA0002769137170000134
Wherein E represents a mean, G represents a variance, and FijCijRespectively representing similar picture elements (x)i,yj) And the pixel values of each wave band and all reference moments on the sentinel No. 2 image and the sentinel No. 3 image are collected.
Considering the two factors and carrying out normalization processing to form the weight W of the similar pixelijk：
Figure BDA0002769137170000135
Wherein, Pijk＝(1-Rij)×Dijk. It can be seen that the similar pixel weight calculation considers the spatial difference between the similar pixel and the central pixel, and the correlation of similar pixel values between different source images in the reference phase.
The method comprises the following steps of calculating a target pixel value at the prediction time according to a set fusion rule, calculating all pixels in an image one by one to generate a sentinel No. 2 image at the prediction time, and comprises the following steps: for a certain reference instant tkCalculating the predicted time t of the image datapThe formula of the central pixel value is as follows:
Figure BDA0002769137170000141
in the formula, C (x)i,yj,tkB) and C (x)i,yj,tpB) represents similar picture elements (x) in the low spatial resolution image search box at the reference time tk and the prediction time tp respectivelyi,yj) And V represents a conversion coefficient at the pixel value of the B-th wave band, linear regression analysis is carried out by using similar pixel values of two groups of different source images before and after the prediction time, and the calculated slope is the conversion coefficient, thereby representing the conversion relation between the mixed pixel of the sentinel No. 3 image and the pure pixel of the sentinel No. 3 image.
And then, synthesizing a plurality of reference time prediction results to obtain a final target pixel value:
Figure BDA0002769137170000142
wherein, TkComprises the following steps:
Figure BDA0002769137170000143
step five: screening and extracting water quality monitoring data
Screening and extracting water quality monitoring data, inverting a database based on established water quality parameters, extracting longitude and latitude information of monitoring points, and extracting the reflectivity data of preprocessed images (including sentinel No. 2 images and fusion images) according to a spatial position relationship.
Step six: method for establishing water quality parameter inversion model by using machine learning method
As shown in fig. 3, the establishment of the water quality parameter inversion model by using the machine learning method means that based on the sentinel No. 2 data and the spatio-temporal fusion result of the sentinel No. 3 data, pairwise ratios and normalized value variables between wave bands are established as independent variables, and monitoring data are used as dependent variables. And (3) adopting machine learning methods such as a neural network, a random forest, a support vector machine, a decision tree, a gradient lifting tree and the like to construct a water quality inversion model, and selecting an optimal precision inversion model. Wherein 70% of data is used for model training, 30% of data is used for model precision verification, and the model precision calculation steps are as follows:
Figure BDA0002769137170000151
in the formula, MAPE represents the average relative error (%), yiIn order to actually monitor the value of the current,
Figure BDA0002769137170000152
and n is the number of dependent variables of the test set.
Step seven: water quality parameter space image output
The water quality parameter spatial image output means that the remote sensing inversion image is subjected to graded color rendering according to the water quality classification regulation in the ground surface water environment quality standard GB3838-2002, and a water quality parameter spatial distribution diagram is output.
It is to be understood that the above-described embodiments of the present invention are merely illustrative of or explaining the principles of the invention and are not to be construed as limiting the invention. Therefore, any modification, equivalent replacement, improvement and the like made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. Further, it is intended that the appended claims cover all such variations and modifications as fall within the scope and boundaries of the appended claims or the equivalents of such scope and boundaries.

Claims (10)

1. A water quality monitoring method based on satellite images and machine learning is characterized by comprising the following steps:
s01: acquiring first satellite image data and second satellite image data;
s02: respectively preprocessing the acquired first satellite image data and the acquired second satellite image data, and generating first satellite image reflectivity data and second satellite image reflectivity data;
s03: establishing a radiation normalization model among different sensors according to different seasons and different ground objects, assimilating the first satellite image reflectivity data and the second satellite image reflectivity data, and generating first satellite image reflectivity data and second satellite image reflectivity data which are small in radiation difference and same in wave band number;
s04: based on the assimilation results of different source image data, a space-time fusion model based on weight filtering is established by using the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data, and a long-time sequence continuous high-spatial-resolution remote-sensing reflectivity image data set is generated;
s05: based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station, screening reflectivity image data and water quality monitoring data, and establishing a water quality parameter inversion database; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
s06: and outputting a water quality parameter inversion result according to the reflectivity image data set generated in the step S04 and the established water quality parameter inversion model.
2. The satellite imagery and machine learned water quality monitoring method of claim 1, wherein said building radiation normalization models for different sensors in step S03 comprises:
s31: classifying the ground objects based on the first satellite image data with high spatial resolution, and classifying the images according to the types of the ground objects;
s32: generating a first satellite image data set and a second satellite image data set according to the seasons and the ground feature classification;
s33: according to the generated image data set, a radiation normalization model of the reflectivity data of the first satellite image is established based on the correlation between adjacent wave bands, the second reflectivity data is assimilated into the first reflectivity data, and a pair of the first satellite image data and the assimilated second satellite image data is established.
3. The satellite image and machine learning water quality monitoring method according to claim 2, wherein the step S04 of establishing a weight filtering-based space-time fusion model comprises:
s41: setting a search frame in a neighborhood range by taking a target pixel as a central pixel, and selecting pixels which have spectrum and time consistency with the central pixel as similar pixels one by one in the search frame according to a set search rule;
s42: calculating the weight ratio of each similar pixel according to the difference of the similar pixel and the central pixel in the spectral dimension and the spatial dimension;
s43: and calculating the target pixel value at the prediction moment according to a set fusion rule, and calculating all pixels in the image one by one to generate the high-spatial-resolution image at the prediction moment.
4. The satellite image and machine learning water quality monitoring method according to claim 3, wherein the step S41 of selecting pixels having spectral and temporal consistency with the center pixel as similar pixels pixel by pixel in the search box according to the set search rules comprises:
calculating the difference value of each pixel and the central pixel value in each reference moment search frame, and comparing the difference value with a set threshold value;
if the difference value of each pixel and the central pixel value in the search frame is less than or equal to the threshold value, the pixel is judged to be a similar pixelI.e. | F (x)i,yj,tk,B)-F(xw/2,yw/2,tk,B)|≤2ω(B)/m；
In the formula, F (x)i,yj,tkB) and F (x)w/2,yw/2,tkB) respectively at a reference time tkSimilar pixels (x) within the first satellite search boxi,yj) And a center pixel (x)w/2,yw/2) The pixel value of the B-th wave band, w represents the size of a search window in a neighborhood, omega (B) represents the standard deviation of the whole first satellite image in the B-th wave band, and m represents the classification number;
selecting similar pixels at each reference moment, and taking an intersection as a final similar pixel selection result;
5. the satellite image and machine learning water quality monitoring method according to claim 3, wherein the step S42 of calculating the weight comprises:
s421: calculating the spatial distance D of the first satellite image based on the reference timeijk：
Figure FDA0002769137160000021
Wherein D isijkRepresenting the space distance between the similar pixel and the central pixel, wherein A is a constant;
s422: calculating (x) on different source imagesi,yj) The correlation coefficient R of similar image elements between all reference time instants and wave bandsij：
Figure FDA0002769137160000022
Wherein E represents a mean value, G represents a variance,
Figure FDA0002769137160000031
and
Figure FDA0002769137160000032
respectively representing similar picture elements (x)i,yj) A set of pixel values of each band and all reference times on the first satellite image and the second satellite image;
s423: considering the two factors and carrying out normalization processing to form the weight W of the similar pixelijk：
Figure FDA0002769137160000033
Wherein, Pijk＝(1-Rij)×Dijk。
6. The satellite image and machine learning water quality monitoring method according to claim 5, wherein the step S43 of calculating the target pixel value at the predicted time comprises:
s431: for a certain reference time tkCalculating the predicted time t of the image datapThe formula of the central pixel value is as follows:
Figure FDA0002769137160000034
in the formula, C (x)i,yj,tkB) and C (x)i,yj,tpB) respectively denote the reference time tkAnd predicting time tpSimilar picture elements (x) within a low spatial resolution image search boxi,yj) The pixel value in the B-th band, V represents the conversion coefficient;
s432: and synthesizing a plurality of reference time prediction results to obtain a final target pixel value:
Figure FDA0002769137160000035
wherein, TkComprises the following steps:
Figure FDA0002769137160000036
7. the satellite image and machine learning water quality monitoring method according to claim 1, wherein the step S05 of establishing an inverse model of the water quality parameters comprises:
based on a space-time fusion result, constructing pairwise ratios and normalized value variables between wave bands as independent variables, monitoring data as dependent variables, constructing a water quality inversion model by adopting various machine learning algorithms, and selecting an optimal precision inversion model; the model precision calculation formula is as follows:
Figure FDA0002769137160000041
where MAPE is the mean relative error, yiIn order to actually monitor the value of the current,
Figure FDA0002769137160000042
and n is the number of dependent variables of the test set.
8. The utility model provides a water quality monitoring system of satellite image and machine learning which characterized in that includes:
the remote sensing image acquisition module is used for acquiring first satellite image data and second satellite image data;
the remote sensing image preprocessing module is used for respectively preprocessing the acquired first satellite image data and the acquired second satellite image data and generating first satellite image reflectivity data and second satellite image reflectivity data;
the different-source image radiation normalization module is used for establishing radiation normalization models among different sensors according to different seasons and different ground objects, assimilating the first satellite image reflectivity data and the second satellite image reflectivity data and generating the first satellite image reflectivity data and the second satellite image reflectivity data which are small in radiation difference and same in wave band quantity;
the remote sensing image space-time fusion module is used for establishing a space-time fusion model based on weight filtering by utilizing the time information of the reflectivity of the ground object of the second satellite image and the space information of the first satellite image data based on the assimilation result of different source image data, and generating a long-time sequence continuous high-spatial-resolution remote sensing reflectivity image data set;
the water quality parameter inversion model module is used for screening reflectivity image data and water quality monitoring data and establishing a water quality parameter inversion database based on time information of a time-space fusion result and spatial information of water quality monitoring data of a ground monitoring station; screening a water quality parameter inversion database, extracting training set and test set data, and establishing an inversion model of the water quality parameters by utilizing various machine learning algorithms;
and the water quality parameter inversion result output module outputs a water quality parameter inversion result according to the generated reflectivity image data set and the established water quality parameter inversion model.
9. The satellite image and machine learning water quality monitoring system according to claim 8, wherein the building of the radiation normalization models of different sensors in the different source image radiation normalization module comprises:
s31: classifying the ground objects based on the first satellite image data with high spatial resolution, and classifying the images according to the types of the ground objects;
s32: generating a first satellite image data set and a second satellite image data set according to the seasons and the ground feature classification;
s33: according to the generated image data set, a radiation normalization model of the reflectivity data of the first satellite image is established based on the correlation between adjacent wave bands, the second reflectivity data is assimilated into the first reflectivity data, and a pair of the first satellite image data and the assimilated second satellite image data is established.
10. The satellite image and machine learning water quality monitoring system according to claim 8, wherein the building of the weight filtering-based space-time fusion model in the remote sensing image space-time fusion module comprises:
s41: setting a search frame in a neighborhood range by taking a target pixel as a central pixel, and selecting pixels which have spectrum and time consistency with the central pixel as similar pixels one by one in the search frame according to a set search rule;
s42: calculating the weight ratio of each similar pixel according to the difference of the similar pixel and the central pixel in the spectral dimension and the spatial dimension;
s43: and calculating the target pixel value at the prediction moment according to a set fusion rule, and calculating all pixels in the image one by one to generate the high-spatial-resolution image at the prediction moment.
CN202011243483.XA 2020-11-10 2020-11-10 Water quality monitoring method and system for satellite image and machine learning Active CN112464746B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011243483.XA CN112464746B (en) 2020-11-10 2020-11-10 Water quality monitoring method and system for satellite image and machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011243483.XA CN112464746B (en) 2020-11-10 2020-11-10 Water quality monitoring method and system for satellite image and machine learning

Publications (2)

Publication Number Publication Date
CN112464746A true CN112464746A (en) 2021-03-09
CN112464746B CN112464746B (en) 2023-09-12

Family

ID=74826376

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011243483.XA Active CN112464746B (en) 2020-11-10 2020-11-10 Water quality monitoring method and system for satellite image and machine learning

Country Status (1)

Country Link
CN (1) CN112464746B (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113255961A (en) * 2021-04-16 2021-08-13 长江水利委员会长江科学院 Lake water environment monitoring site optimized layout method based on time sequence multi-source spectrum remote sensing data
CN113324923A (en) * 2021-06-07 2021-08-31 郑州大学 Remote sensing water quality inversion method combining time-space fusion and deep learning
CN113534158A (en) * 2021-06-10 2021-10-22 国家卫星气象中心(国家空间天气监测预警中心) Static orbit meteorological satellite radar reflectivity factor inversion method based on deep learning
CN113536606A (en) * 2021-09-15 2021-10-22 成都同飞科技有限责任公司 Water quality change detection method and system, storage medium and device
CN113624778A (en) * 2021-09-18 2021-11-09 重庆星视空间科技有限公司 Water pollution tracing system and method based on remote sensing image inversion
CN114965918A (en) * 2022-04-20 2022-08-30 重庆两江生态渔业发展有限公司 Water quality analysis method based on satellite remote sensing image
CN115439757A (en) * 2022-08-26 2022-12-06 南方海洋科学与工程广东省实验室(广州) Water quality pollution source tracing method, device and equipment based on remote sensing image
CN115830473A (en) * 2023-02-20 2023-03-21 江苏省生态环境监测监控有限公司 Water quality inversion method and system based on satellite remote sensing and automatic monitoring
CN117351373A (en) * 2023-12-04 2024-01-05 厦门大学 Intelligent fusion processing method for estuary and bay water quality remote sensing products
CN117422987A (en) * 2023-09-06 2024-01-19 中国热带农业科学院环境与植物保护研究所 River water quality monitoring method based on satellite remote sensing image and machine learning
CN117470789A (en) * 2023-11-06 2024-01-30 中国自然资源航空物探遥感中心 Method, device, equipment and readable storage medium for identifying dawsonite
WO2024041560A1 (en) * 2022-08-24 2024-02-29 武汉大学 Surface water quality monitoring method based on high-spatial-resolution satellite

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103927455A (en) * 2014-04-24 2014-07-16 中国科学院遥感与数字地球研究所 Land aerosol optical property retrieval method based on Gaofen-1 satellite
CN105184076A (en) * 2015-09-02 2015-12-23 安徽大学 Space-time integrated fusion method for remote sensing earth surface temperature data
CN107316017A (en) * 2017-06-22 2017-11-03 河海大学 Non-point pollution source strength evaluation method based on Crop Information remote sensing exact inversion
CN107610050A (en) * 2017-08-31 2018-01-19 北京航天绘景科技有限公司 A kind of high-spatial and temporal resolution land table Temperature fusion method based on multi-source satellite data
CN108613933A (en) * 2018-06-13 2018-10-02 中南林业科技大学 Forest land arid space-time dynamic monitoring method based on multi-sources RS data fusion
CN108985959A (en) * 2018-08-09 2018-12-11 安徽大学 Wheat powdery mildew remote sensing monitoring method based on surface temperature inversion technology
CN109933873A (en) * 2018-11-02 2019-06-25 中国林业科学研究院资源信息研究所 A kind of method of desertificated area Soil Organic Carbon Density high spatial remote sensing estimation
CN110865040A (en) * 2019-11-29 2020-03-06 深圳航天智慧城市系统技术研究院有限公司 Sky-ground integrated hyperspectral water quality monitoring and analyzing method
CN111444351A (en) * 2020-03-24 2020-07-24 清华苏州环境创新研究院 Method and device for constructing knowledge graph in industrial process field
CN111639719A (en) * 2020-06-08 2020-09-08 安徽大学 Footprint image retrieval method based on space-time motion and feature fusion
CN111666896A (en) * 2020-06-09 2020-09-15 中国科学院地理科学与资源研究所 Remote sensing image space-time fusion method based on linear fusion model
CN111797186A (en) * 2020-06-17 2020-10-20 南开大学 Method for inverting COD (chemical oxygen demand) of water body by remote sensing

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103927455A (en) * 2014-04-24 2014-07-16 中国科学院遥感与数字地球研究所 Land aerosol optical property retrieval method based on Gaofen-1 satellite
CN105184076A (en) * 2015-09-02 2015-12-23 安徽大学 Space-time integrated fusion method for remote sensing earth surface temperature data
CN107316017A (en) * 2017-06-22 2017-11-03 河海大学 Non-point pollution source strength evaluation method based on Crop Information remote sensing exact inversion
CN107610050A (en) * 2017-08-31 2018-01-19 北京航天绘景科技有限公司 A kind of high-spatial and temporal resolution land table Temperature fusion method based on multi-source satellite data
CN108613933A (en) * 2018-06-13 2018-10-02 中南林业科技大学 Forest land arid space-time dynamic monitoring method based on multi-sources RS data fusion
CN108985959A (en) * 2018-08-09 2018-12-11 安徽大学 Wheat powdery mildew remote sensing monitoring method based on surface temperature inversion technology
CN109933873A (en) * 2018-11-02 2019-06-25 中国林业科学研究院资源信息研究所 A kind of method of desertificated area Soil Organic Carbon Density high spatial remote sensing estimation
CN110865040A (en) * 2019-11-29 2020-03-06 深圳航天智慧城市系统技术研究院有限公司 Sky-ground integrated hyperspectral water quality monitoring and analyzing method
CN111444351A (en) * 2020-03-24 2020-07-24 清华苏州环境创新研究院 Method and device for constructing knowledge graph in industrial process field
CN111639719A (en) * 2020-06-08 2020-09-08 安徽大学 Footprint image retrieval method based on space-time motion and feature fusion
CN111666896A (en) * 2020-06-09 2020-09-15 中国科学院地理科学与资源研究所 Remote sensing image space-time fusion method based on linear fusion model
CN111797186A (en) * 2020-06-17 2020-10-20 南开大学 Method for inverting COD (chemical oxygen demand) of water body by remote sensing

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
JUNYU HE 等: "Space-time chlorophyll-a retrieval in optically complex waters that accounts for remote sensing and modeling uncertainties and improves remote estimation accuracy", 《WATER RESEARCH》, vol. 171, pages 1 - 17 *
WEIQI HE 等: "Water quality monitoring in a slightly-polluted inland water body through remote sensing — Case study of the Guanting Reservoir in Beijing, China", 《FRONTIERS OF ENVIRONMENTAL SCIENCE & ENGINEERING IN CHINA》, vol. 2, pages 163 *
段广拓: "基于多源遥感数据的水环境参量反演算法研究与应用", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, no. 6, pages 028 - 51 *
董世元 等: "ESTARFM 相似像元选取方法的改进研究", 《遥感技术与应用》, vol. 35, no. 1, pages 185 - 193 *

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113255961B (en) * 2021-04-16 2022-05-10 长江水利委员会长江科学院 Lake water environment monitoring site optimized layout method based on time sequence multi-source spectrum remote sensing data
CN113255961A (en) * 2021-04-16 2021-08-13 长江水利委员会长江科学院 Lake water environment monitoring site optimized layout method based on time sequence multi-source spectrum remote sensing data
CN113324923A (en) * 2021-06-07 2021-08-31 郑州大学 Remote sensing water quality inversion method combining time-space fusion and deep learning
CN113324923B (en) * 2021-06-07 2023-07-07 郑州大学 Remote sensing water quality inversion method combining space-time fusion and deep learning
CN113534158A (en) * 2021-06-10 2021-10-22 国家卫星气象中心(国家空间天气监测预警中心) Static orbit meteorological satellite radar reflectivity factor inversion method based on deep learning
CN113534158B (en) * 2021-06-10 2024-06-11 国家卫星气象中心(国家空间天气监测预警中心) Static orbit meteorological satellite radar reflectivity factor inversion method based on deep learning
CN113536606A (en) * 2021-09-15 2021-10-22 成都同飞科技有限责任公司 Water quality change detection method and system, storage medium and device
CN113624778A (en) * 2021-09-18 2021-11-09 重庆星视空间科技有限公司 Water pollution tracing system and method based on remote sensing image inversion
CN114965918A (en) * 2022-04-20 2022-08-30 重庆两江生态渔业发展有限公司 Water quality analysis method based on satellite remote sensing image
WO2024041560A1 (en) * 2022-08-24 2024-02-29 武汉大学 Surface water quality monitoring method based on high-spatial-resolution satellite
CN115439757A (en) * 2022-08-26 2022-12-06 南方海洋科学与工程广东省实验室(广州) Water quality pollution source tracing method, device and equipment based on remote sensing image
CN115830473A (en) * 2023-02-20 2023-03-21 江苏省生态环境监测监控有限公司 Water quality inversion method and system based on satellite remote sensing and automatic monitoring
CN117422987A (en) * 2023-09-06 2024-01-19 中国热带农业科学院环境与植物保护研究所 River water quality monitoring method based on satellite remote sensing image and machine learning
CN117470789A (en) * 2023-11-06 2024-01-30 中国自然资源航空物探遥感中心 Method, device, equipment and readable storage medium for identifying dawsonite
CN117351373B (en) * 2023-12-04 2024-03-12 厦门大学 Intelligent fusion processing method for estuary and bay water quality remote sensing products
CN117351373A (en) * 2023-12-04 2024-01-05 厦门大学 Intelligent fusion processing method for estuary and bay water quality remote sensing products

Also Published As

Publication number Publication date
CN112464746B (en) 2023-09-12

Similar Documents

Publication Publication Date Title
CN112464746B (en) Water quality monitoring method and system for satellite image and machine learning
Wang et al. Spatio-temporal fusion for daily Sentinel-2 images
CN113128134B (en) Mining area ecological environment evolution driving factor weight quantitative analysis method
CN110222903B (en) Rice yield prediction method and system based on unmanned aerial vehicle remote sensing
Carmel et al. Combining Location and Classification Error Sources for Estimating Multi-Temporal
CN111767865A (en) Method for inverting mangrove forest biomass by using aerial image and laser data
KR102193108B1 (en) Observation method for two-dimensional river mixing using RGB image acquired by the unmanned aerial vehicle
CN112699959B (en) Multi-source multi-scale precipitation data fusion method and device based on energy functional model
CN110927120B (en) Early warning method for vegetation coverage
Kaivosoja et al. A case study of a precision fertilizer application task generation for wheat based on classified hyperspectral data from UAV combined with farm history data
US20230186503A1 (en) Method for troubleshooting hidden dangers of trees near power transmission lines by combining icesat-2 with jl-1 images
CN111937016B (en) City internal poverty-poor space measuring method and system based on street view picture and machine learning
CN110363246A (en) A kind of fusion method of high-spatial and temporal resolution vegetation index NDVI
CN106779067A (en) Soil moisture method for reconstructing and system based on multi- source Remote Sensing Data data
CN112380781B (en) Satellite observation completion method based on reanalysis data and unbalanced learning
Engerer et al. Himawari-8 enabled real-time distributed PV simulations for distribution networks
CN111723525B (en) PM2.5 inversion method based on multi-source data and neural network model
KR102476565B1 (en) A floating waste detection system for lake and swamp using statistical characteristics of SAR radar images and methods using the same
CN114120101A (en) Soil moisture multi-scale comprehensive sensing method
CN117687416A (en) Path planning method and system for river network water safety detection device
RU2716477C1 (en) System for monitoring forest use and forest pathological changes
CN113779863B (en) Ground surface temperature downscaling method based on data mining
CN115730718A (en) Atmospheric NO combining hyper-spectral satellite and artificial intelligence 2 Space-time prediction algorithm
CN112733746B (en) Collaborative classification method for fusing InSAR coherence and multispectral remote sensing
CN117826112B (en) Soil water content inversion method based on sar

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