CN111339909A - Extraction algorithm of surface linear water body based on multi-scale region growth - Google Patents

Extraction algorithm of surface linear water body based on multi-scale region growth Download PDF

Info

Publication number
CN111339909A
CN111339909A CN202010111031.XA CN202010111031A CN111339909A CN 111339909 A CN111339909 A CN 111339909A CN 202010111031 A CN202010111031 A CN 202010111031A CN 111339909 A CN111339909 A CN 111339909A
Authority
CN
China
Prior art keywords
pixel
water body
scale
river
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010111031.XA
Other languages
Chinese (zh)
Other versions
CN111339909B (en
Inventor
金松
刘永学
孙超
刘永超
陆婉芸
李慧婷
许文轩
赵冰雪
吴伟
董雁伫
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.)
Nanjing University
Original Assignee
Nanjing University
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 Nanjing University filed Critical Nanjing University
Priority to CN202010111031.XA priority Critical patent/CN111339909B/en
Publication of CN111339909A publication Critical patent/CN111339909A/en
Application granted granted Critical
Publication of CN111339909B publication Critical patent/CN111339909B/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
    • G06V20/182—Network patterns, e.g. roads or rivers
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • 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
    • G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1793—Remote sensing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pathology (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Immunology (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Image Processing (AREA)

Abstract

The invention relates to a ground surface linear river extraction algorithm based on multi-scale region growth, and belongs to the technical field of remote sensing geography application. The algorithm is divided into four parts: (1) heterogeneous background equalization of complex rivers: river pixels in the image are enhanced through a normalized water body index (NDWI) and a spectral deviation correction fuzzy c-means clustering algorithm (BCFCM). (2) Water system enhancement of multi-scale local differential structure: and performing second-order Hessian matrix analysis on the image by using Frangi filtering to obtain a river pixel scale response diagram and a directional diagram. (3) And (3) growing and extracting river pixels in the multi-scale region: and establishing a region growing rule through the pixel scale characteristic and the direction characteristic to extract the water body pixels. (4) Removing pixel noise of the linear water body: and establishing a screening criterion, and deleting the water body pixels with excessive growth. The algorithm considers the characteristics of different river flow forms and complex background, and efficiently and accurately identifies and extracts the river.

Description

Extraction algorithm of surface linear water body based on multi-scale region growth
Technical Field
The invention relates to an extraction algorithm of a surface linear water body based on multi-scale region growth, and belongs to the technical field of remote sensing geography application.
Background
Surface rivers are an important part of the life of the earth and are the basis for human survival and development. The occurrence and development of human beings and social ecosystems thereof are interdependent and inseparable with rivers. As population grows and climate changes, surface rivers change from surface morphological features to internal, inherent ecological particles. Accurate river extraction is a necessary premise for knowing and effectively managing rivers, and a river distribution map can provide important information for monitoring and evaluating flood disasters, flood forecasting, water pollution monitoring and geographic information database updating.
The main feature of satellite remote sensing is its high detection capability, which makes it relatively easy and cost effective to capture large areas of surface information. The Sentinel-2MSI is used as a Landsat series follow-up satellite, and the spatial resolution, the spectral resolution and the signal-to-noise ratio of the Sentinel-2MSI are greatly improved. The Sentinel-2MSI has a spatial resolution of 10-30m and a short return visit period, and provides an optimal data set for rapidly monitoring surface rivers. The river system is usually composed of rivers with different sizes, the width of the rivers is from 10m to thousands of meters, and the river systems are communicated and staggered with each other. A linear river extraction algorithm based on multi-scale region growth is developed, and a river system distribution map covering the whole river can be obtained efficiently and accurately through pixel seed point growth.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method overcomes the defect of monitoring tiny rivers in the existing river extraction algorithm, and provides an extraction algorithm of the surface linear water body based on multi-scale region growth. The method combines background equalization and multi-scale water body enhancement to establish a region growth criterion for identifying different water body scales, and accurately and efficiently extracts linear water bodies with different scales.
In order to solve the technical problems, the extraction algorithm of the surface linear water body based on the multi-scale region growth provided by the invention comprises the following steps:
firstly, preparing a Sentinel-2MSI image, performing atmospheric correction to eliminate the influence of atmospheric scattering and absorption, performing mask operation on the image by using a target region vector file, and cutting an interested region;
secondly, performing normalized water body index operation by using a green light wave band and a near infrared wave band of the image to obtain an NDWI image and highlight linear water body information of a target area;
thirdly, performing clustering operation on the NDWI image according to a spectral deviation correction fuzzy C-means clustering algorithm, and further enhancing a linear river gray value to obtain a BCFCM image;
setting N scales sigma related to the river width as convolution scales, and performing filtering processing on the BCFCM image by using a Frangi filtering algorithm of a Hessian matrix to obtain a river response graph L and a directional diagram D, wherein the river response graph L is formed by overlapping N layers, each layer corresponds to one scale sigma, and the scales sigma are sequentially arranged from large to small;
fifthly, selecting a point 1% before the brightness value in the river response image L as an initial seed point with the maximum scale, and visually selecting a plurality of seed points in the corresponding image layer;
and sixthly, establishing a region growth principle, and growing the seed points, wherein the convergence conditions are as follows: for the seed point q, if the eight neighborhood pixels p of the same scale simultaneously satisfy the following two formulas, the pixel p is determined as a water body pixel:
Figure BDA0002389998800000021
Figure BDA0002389998800000022
in the formula, D (p) and D (q) are direction information of the pixel elements p and q respectively and are directly obtained from a directional diagram D, L (p) is river response of the pixel element p and is directly obtained from a river response diagram L, and η is a key threshold value for water body region growth, and the value of the threshold value is 0.8;
step seven, taking the grown water body pixel as a seed point of the next scale, continuing to grow by using the growth principle of the step six, and continuously repeating the step until the growth of the water body pixel of the last scale is finished;
eighthly, creating a 10 x 10 sliding window, wherein a pixel q with the largest brightness value in the window is a determined water body pixel, the brightness value of the pixel q is I (q), the pixel q is directly obtained from the BCFCM image, if I (q) is not less than M, the pixel q is a large-scale bright linear water body, otherwise, the pixel q is a small-scale weak water body, and M is the first 20% pixel value of the BCFCM image gray level histogram;
for the pixel p in the window, if the following formula is met, the pixel p is determined as a water body pixel to be determined, and the pixel which does not meet the following formula is removed
Figure BDA0002389998800000031
In the formula, D (p) and D (q) are direction information of the pixel p and q respectively and are directly obtained from a directional diagram D;
if the pixel q with the maximum brightness value in the window is a large-scale bright linear water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T1If the water body pixel p is the determined water body pixel, the water body pixel p is reserved, otherwise, the water body pixel p is noiseCarry out deletion, T1For setting the large-scale gray level threshold, the value range is as follows: pixel values of 50% -60% of a BCFCM image gray value histogram;
if the pixel q with the maximum brightness value in the window is a small-scale weak water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T2If the noise is not the noise, the water body pixel p is reserved for the determined water body pixel, otherwise, the noise is deleted, T2In order to set a small-scale gray threshold value, the value range is 80% of the pixel value of the BCFCM image gray value histogram;
and ninthly, performing binarization processing on the image formed by the determined water body pixels to obtain a multi-scale river distribution map.
The multi-scale linear water body algorithm extraction adopted by the invention belongs to the innovation of algorithm application, and the core idea of the algorithm is derived from the extraction of linear blood vessels and retinas in the medical field; meanwhile, the problems of linear characteristic blurring, high background heterogeneity and the like of rivers in images are considered, and the difference between the linear water body characteristics and the background is increased by introducing a spectral deviation correction fuzzy C mean value clustering algorithm. Background noise in the image is eliminated through background equalization to enhance linear water body characteristics, a multi-scale region growth criterion is established, and river pixels are extracted from large scale to small scale. Compared with the traditional river extraction algorithm, the algorithm improves the linear water body identification precision, robustness and universality, and provides an idea for fine linear river extraction. All steps are realized by Matlab programming, and manual participation is reduced.
The main innovation of the algorithm is the following three aspects:
1. in the third step, an improved spectrum deviation correction fuzzy C mean value clustering algorithm (BCFCM) is used for processing the NDWI image, the overlapped area of the target water and the heterogeneous background reflectivity in the image is separated, and reflectivity confusion among pixels is reduced.
2. And in the fourth step, a linear filtering enhancer is established by utilizing the characteristic value of Hessian, linear target enhancement is carried out by traversing each scale, and the maximum response of the detection scale space is recorded into the linear characteristic index of the optimal scale.
3. And in the sixth step, a region growing rule is constructed according to river pixel multiscale response characteristics and direction information obtained by a Hessian matrix, 8 neighborhood pixels around the seed point pixel are used for seed point growing, and when all pixels to be determined in the neighborhood of the seed point meet set conditions under corresponding scales, the pixels to be determined are classified into a water body and are included in a seed point set of the next scale.
The method has the advantages of strong robustness, high precision and high sensitivity to a linear river system in river extraction of different scales, and simultaneously has a good river extraction effect in Landsat and Modis images.
Drawings
The invention will be further described with reference to the accompanying drawings.
FIG. 1 is a flow chart of a surface linear water body extraction algorithm based on multi-scale region growing.
FIG. 2 is a normalized water body index effect graph, a fuzzy c-clustering algorithm graph and a corresponding gray level histogram.
Fig. 3 is a target enhancement of river differential structure at different scales.
Fig. 4 is a river direction characteristic diagram.
Fig. 5 is a flow chart of a linear water body multi-scale region growing process.
Fig. 6 is a multi-scale linear river noise removal flowchart.
FIG. 7 is a diagram of a multi-scale region growing process.
Detailed Description
The technical route and the operation steps of the present invention will be more clearly understood from the following detailed description of the present invention with reference to the accompanying drawings. An example study area of the invention is the Welland river in UK, 5 months and 7 days in 2018.
The extraction algorithm (fig. 1) of the surface linear water body based on the multi-scale region growth in the embodiment comprises the following steps:
the first step is as follows: and (3) a Sentinel-2A image (the resolution is 10m) of the British region of 5,7 and 2018 is downloaded, and the high resolution and the wide range of the data can identify rivers with different scales in the research area. According to the image date and the monitoring position, Flash atmospheric correction parameters including sensor types, atmospheric models, vapor inversion, aerosol models and the like are input, and atmospheric scattering and absorption influences are eliminated. And performing mask operation on the image by using the target area vector file in the ENVI, and cutting to obtain the region of interest.
The second step is that: reading the cut image by using Matlab, carrying out normalization processing on 2 nd (green light) and 8 th (near infrared) wave bands, and calculating a normalized water body index by using the following formula to obtain an NDWI image so as to highlight linear water body information of a target area:
Figure BDA0002389998800000051
in the formula, ρ2Intensity value in the green band, p8The intensity value of the near red wave band.
Fig. 2a and 2b are NDWI images of a research area and corresponding NDWI histogram distribution diagrams, respectively, and the histograms are in a multi-peak state, and the water characteristics and the background characteristics are difficult to distinguish.
And thirdly, performing a spectrum deviation corrected fuzzy C-means clustering algorithm (BCFCM) written by Matlab on the NDWI image to obtain a BCFCM image, wherein system parameters vi are set to be 0.6 to 0.6, the maximum iteration times are 1000 and α are equal to 50 in the experiment, the difference between the water body characteristic and the background information is increased, and the image gray frequency has bimodal characteristics (figure 2C and figure 2d), the spectrum deviation corrected fuzzy C-means clustering algorithm is the prior art, and can be used for processing a river by a Landsat 8 image based fuzzy water flow and a river characteristic, and the background characteristic is further enhanced by the BCFCM (a bimodal fuzzy C-means clustering algorithm) and a river characteristic fuzzy C-means histogram.
The fourth step: writing a Frangi filtering program based on a Hessian matrix, and performing filtering processing on a BCFCM image by using a Frangi filtering algorithm of the Hessian matrix to obtain a river response diagram L and a directional diagram D, wherein the river response diagram L is formed by superposing N layers, each layer corresponds to a scale sigma, and the scales sigma are sequentially arranged from large to small. The pixel scale σ of the Frangi filter is 1 to 10. In the step, a high-brightness linear characteristic indicator is established by using the characteristic value of the Hessian matrix.
The Frangi filtering algorithm of the Hessian matrix is an existing method, see the paper (Multiscale Vessel enhancement filtering, Medical Image Computing and Computer-assisted reconstruction-MICCAI' 98(pp.130-137): Springer; Vessel enhancement differentiation: area space representation of Vessel structures, Medical Image Analysis,10, 815-825; partial differential equation-based Medical Image enhancement and segmentation method research, Hunan university, 2012, Liangfei). The filtering process is roughly as follows:
for each pixel scale (i.e. convolution scale), the Hessian matrix of each pixel is calculated, the eigenvalue (lambda) of the Hessian matrix1、λ2) And a feature vector (e)1、e2) Can identify the bright background and the dark background of the dark target, and the pixel value of a certain pixel in the river meets the lambda 10 and lambda2<And 0, the water body where the target pixel is located is regarded as a bright feature, and the surrounding heterogeneous objects are regarded as dark features. And calculating to obtain a river response value of the pixel according to the Hessian matrix, and taking the maximum river response value as the final water body scale after calculating a series of scales so as to obtain a river response graph L. Assuming that the river response value is the largest when the scales of the pixels P1 and P3 are 2, and the river response value is the largest when the scales of the pixels P2 and P4 are 4, the pixels P1 and P3 are in the same layer (which can be set as the 2 nd layer), the pixels P2 and P4 are in the same layer (which can be set as the 4 th layer), and each scale corresponds to one layer.
Hessian matrix eigenvector e1And e2Can reflect the local shape of a linear river, e1The direction of the potentially linear structure is represented by the low second derivative, e2The normal direction of a potentially linear structure is represented by a high second derivative. When the river response value is maximum, the river directions at the pixel positions are approximately parallel, otherwise, the pixel directions are disordered, and a directional diagram D is obtained. When the algorithm selection scale is the optimal scale of the river, the directions of the water body pixels and the neighborhood water body pixels are consistent, and the continuity and consistency of each pixel and the adjacent pixels are ensured.
Fig. 3 is a river response graph, where the Frangi filtering used for small scale σ ═ 1 only enhances small rivers, and larger width rivers are identified as the scale σ grows. Line1, Line2 and Line3 represent rivers of three scales, small, medium and large, respectively. Line1 σ ═ 1 is the best metric, Line2 σ ═ 3, and Line3 σ ═ 5.
Fig. 4 is a river pixel directional diagram. The pixel directions of the partial enlarged view in fig. 4a are consistent; FIG. 4b corresponds to Line1 in FIG. 4a calculated from eigenvectors e1 and e2 of the Hessian matrix; fig. 4c corresponds to Line2 in fig. 4a and shows the magnitude of the component of the river pixel response L in the direction D.
The fifth step: and selecting a point 1% before the brightness value in the river response graph L as an initial seed point, locating the initial seed point in a large-scale river, and manually adding the seed point to a linear water body with discrete local regions of the tributaries in the same graph layer. This step corresponds to fig. 7 a.
And a sixth step: establishing a region growth principle, and growing the seed points, wherein the convergence conditions are as follows: for the seed point q, if the eight neighborhood pixels p of the same scale simultaneously satisfy the following two formulas, the pixel p is determined as a water body pixel:
Figure BDA0002389998800000071
Figure BDA0002389998800000081
in the formula, D (p) and D (q) are direction information of the pixel elements p and q respectively and are directly obtained from the directional diagram D, L (p) is river response of the pixel element p and is directly obtained from a river response diagram L, and η is a key threshold value for water body region growth, and the value of the threshold value is 0.8.
And step seven, taking the grown water body pixel as a seed point of the next scale, continuing to grow by using the growth principle of the step six, and continuously repeating the step until the growth of the water body pixel of the last scale is finished.
In the growth process of the two steps of water body seed points, the water body seed points grow from the maximum dimension sigma of 10 to the maximum dimension sigma of 1, and when the undetermined pixels in the neighborhood of the seed points meet the dimension growth condition, the undetermined pixels are summarized into water body pixels until the minimum dimension growth is finished. The algorithm calculation process is given by fig. 5, and the picture element growth process diagram corresponds to fig. 7 b-7 e.
Eighth step: and (3) creating a 10 x 10 sliding window, wherein a pixel q with the maximum brightness value in the window is a determined water body pixel, the brightness value of the pixel q is I (q), and the pixel q is directly obtained from the BCFCM image, if I (q) is more than or equal to M, the pixel q is a large-scale bright linear water body, otherwise, the pixel q is a small-scale weak water body, and M is the first 20% pixel value of the BCFCM image gray level histogram. In this example, the strategy for noise elimination at different scales is: and setting a gray threshold value of 0.6, and when the target pixel meets the condition that I (q) is more than or equal to 0.6, positioning the target pixel around the large-scale bright linear water body, otherwise, setting the target pixel as a small-scale weak water body.
For the pixel p in the window, if the following formula is met, the pixel p is determined as a water body pixel to be determined, and the pixel which does not meet the following formula is removed
Figure BDA0002389998800000082
In the formula, D (p) and D (q) are direction information of the pixel p and q respectively and are directly obtained from a directional diagram D;
if the pixel q with the maximum brightness value in the window is a large-scale bright linear water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T1If the noise is not the noise, the water body pixel p is reserved for the determined water body pixel, otherwise, the noise is deleted, T1For setting the large-scale gray level threshold, the value range is as follows: and the 50% -60% pixel value of the BCFCM image gray value histogram. In this example, T1The value is 0.3.
If the pixel q with the maximum brightness value in the window is a small-scale weak water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T2If the noise is not the noise, the water body pixel p is reserved for the determined water body pixel, otherwise, the noise is deleted, T2In order to set a small-scale gray threshold, the value range is 80% of the pixel value of the BCFCM image gray value histogram. In this example, T2The value is-0.1.
The large scale gray scale threshold value T1And small scale gray scale threshold T2The parameters are required to be tested and confirmed for multiple times according to actual image data, a threshold with good river extraction effect is selected, good effect and poor effect are visually identified, the effect is good if the river is clear and has few noise points, and the effect is poor otherwise. The specific flow of the algorithm is given by fig. 6, which corresponds to fig. 7 f.
The ninth step: and carrying out binarization operation on the river result to obtain a multi-scale river distribution map.
In addition to the above embodiments, the present invention may have other embodiments. All technical solutions formed by adopting equivalent substitutions or equivalent transformations fall within the protection scope of the claims of the present invention.

Claims (6)

1. An extraction algorithm of a surface linear water body based on multi-scale region growth comprises the following steps:
preparing a multispectral satellite image, performing atmospheric correction, eliminating the influence of atmospheric scattering and absorption, performing mask operation on the image by using a target area vector file, and cutting an interested area;
secondly, performing normalized water body index operation by using a green light wave band and a near infrared wave band of the image to obtain an NDWI image and highlight linear water body information of a target area;
thirdly, performing clustering operation on the NDWI image according to a spectral deviation correction fuzzy C-means clustering algorithm, and further enhancing a linear river gray value to obtain a BCFCM image;
fourthly, setting N pixel scales sigma related to the river width as convolution scales, and performing filtering processing on the BCFCM image by using a Frangi filtering algorithm of a Hessian matrix to obtain a river response graph L and a directional diagram D, wherein the river response graph L is formed by overlapping N layers, each layer corresponds to one scale sigma, and the scales sigma are sequentially arranged from large to small;
fifthly, selecting a point 1% before the brightness value in the river response image L as an initial seed point with the maximum scale, and visually selecting a plurality of seed points in the corresponding image layer;
and sixthly, establishing a region growth principle, and growing the seed points, wherein the convergence conditions are as follows: for the seed point q, if the eight neighborhood pixels p of the same scale simultaneously satisfy the following two formulas, the pixel p is determined as a water body pixel:
Figure FDA0002389998790000011
Figure FDA0002389998790000012
in the formula, D (p) and D (q) are direction information of the pixel elements p and q respectively and are directly obtained from a directional diagram D, L (p) is river response of the pixel element p and is directly obtained from a river response diagram L, and η is a key threshold value for water body region growth, and the value of the threshold value is 0.8;
step seven, taking the grown water body pixel as a seed point of the next scale, continuing to grow by using the growth principle of the step six, and continuously repeating the step until the growth of the water body pixel of the last scale is finished;
eighthly, creating a 10 x 10 sliding window, wherein a pixel q with the largest brightness value in the window is a determined water body pixel, the brightness value of the pixel q is I (q), the pixel q is directly obtained from the BCFCM image, if I (q) is not less than M, the pixel q is a large-scale bright linear water body, otherwise, the pixel q is a small-scale weak water body, and M is the first 20% pixel value of the BCFCM image gray level histogram;
for the pixel p in the window, if the following formula is met, the pixel p is determined as a water body pixel to be determined, and the pixel which does not meet the following formula is removed
Figure FDA0002389998790000021
In the formula, D (p) and D (q) are direction information of the pixel p and q respectively and are directly obtained from a directional diagram D;
if the pixel q with the maximum brightness value in the window is a large-scale bright linear water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T1Then the water body pixel p is the determined water bodyThe pixels are retained, otherwise, the noise is deleted, T1For setting the large-scale gray level threshold, the value range is as follows: pixel values of 50% -60% of a BCFCM image gray value histogram;
if the pixel q with the maximum brightness value in the window is a small-scale weak water body, judging the brightness value I (p) of the pixel p of the water body to be determined, and if I (p) is more than or equal to T2If the noise is not the noise, the water body pixel p is reserved for the determined water body pixel, otherwise, the noise is deleted, T2In order to set a small-scale gray threshold value, the value range is 80% of the pixel value of the BCFCM image gray value histogram;
and ninthly, performing binarization processing on the image formed by the determined water body pixels to obtain a multi-scale river distribution map.
2. The multi-scale region growth-based surface linear water body extraction algorithm of claim 1, characterized in that: the calculation formula of the normalized water body index is as follows:
Figure FDA0002389998790000022
in the formula, ρ2Intensity value in the green band, p8The intensity value of the near red wave band.
3. The multi-scale region growth-based surface linear water body extraction algorithm of claim 1, characterized in that: the multispectral satellite image is a Sentinel-2A image with the respective rate of 10 m.
4. The multi-scale region growth-based surface linear water body extraction algorithm of claim 1, characterized in that: the system parameter v of the spectral deviation correction fuzzy C-means clustering algorithmiIs set to be [0.6, -0.6]The maximum number of iterations is set to 1000 and the neighborhood pel influence parameter α is set to 50.
5. The multi-scale region growth-based surface linear water body extraction algorithm of claim 1, characterized in that: the N scales sigma are natural numbers of 1-10.
6. The multi-scale region growth-based surface linear water body extraction algorithm of claim 1, characterized in that: the scale σ takes the form 1,3,5,7,9 or 2,4,6,8, 10.
CN202010111031.XA 2020-02-24 2020-02-24 Surface linear water body extraction algorithm based on multi-scale region growth Active CN111339909B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010111031.XA CN111339909B (en) 2020-02-24 2020-02-24 Surface linear water body extraction algorithm based on multi-scale region growth

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010111031.XA CN111339909B (en) 2020-02-24 2020-02-24 Surface linear water body extraction algorithm based on multi-scale region growth

Publications (2)

Publication Number Publication Date
CN111339909A true CN111339909A (en) 2020-06-26
CN111339909B CN111339909B (en) 2023-05-05

Family

ID=71183672

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010111031.XA Active CN111339909B (en) 2020-02-24 2020-02-24 Surface linear water body extraction algorithm based on multi-scale region growth

Country Status (1)

Country Link
CN (1) CN111339909B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112766075A (en) * 2020-12-31 2021-05-07 中国冶金地质总局矿产资源研究院 Hyperspectral remote sensing black and odorous water body grading method based on semi-supervised learning strategy

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101546431A (en) * 2009-05-07 2009-09-30 同济大学 Extraction method of water body thematic information of remote sensing image based on sequential nonlinear filtering
US20160307073A1 (en) * 2015-04-20 2016-10-20 Los Alamos National Security, Llc Change detection and change monitoring of natural and man-made features in multispectral and hyperspectral satellite imagery
CN108537795A (en) * 2018-04-23 2018-09-14 中国科学院地球化学研究所 A kind of mountain stream information extracting method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101546431A (en) * 2009-05-07 2009-09-30 同济大学 Extraction method of water body thematic information of remote sensing image based on sequential nonlinear filtering
US20160307073A1 (en) * 2015-04-20 2016-10-20 Los Alamos National Security, Llc Change detection and change monitoring of natural and man-made features in multispectral and hyperspectral satellite imagery
CN108537795A (en) * 2018-04-23 2018-09-14 中国科学院地球化学研究所 A kind of mountain stream information extracting method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YONGXING WANG, YONGXUE LIU, SONG JIN, CHAO SUN, XIANGLIN WEI: "Evolution of the topography of tidal flats and sandbanks along the Jiangsu coast from 1973 to 2016 observed from satellites" *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112766075A (en) * 2020-12-31 2021-05-07 中国冶金地质总局矿产资源研究院 Hyperspectral remote sensing black and odorous water body grading method based on semi-supervised learning strategy

Also Published As

Publication number Publication date
CN111339909B (en) 2023-05-05

Similar Documents

Publication Publication Date Title
CN109446992B (en) Remote sensing image building extraction method and system based on deep learning, storage medium and electronic equipment
CN109800736B (en) Road extraction method based on remote sensing image and deep learning
Sahebjalal et al. Analysis of land use-land covers changes using normalized difference vegetation index (NDVI) differencing and classification methods
CN112016436A (en) Remote sensing image change detection method based on deep learning
Cablk et al. Detecting and discriminating impervious cover with high-resolution IKONOS data using principal component analysis and morphological operators
CN109978032B (en) Bridge crack detection method based on space pyramid cavity convolution network
CN108564588B (en) Built-up area automatic extraction method based on depth features and graph segmentation method
Lu et al. Detection of urban expansion in an urban-rural landscape with multitemporal QuickBird images
CN111881816B (en) Long-time-sequence river and lake ridge culture area monitoring method
CN113033670A (en) Method for extracting rice planting area based on Sentinel-2A/B data
Im et al. An automated binary change detection model using a calibration approach
CN112488046A (en) Lane line extraction method based on high-resolution images of unmanned aerial vehicle
CN112418506B (en) Coastal zone wetland ecological safety pattern optimization method and device based on machine learning
CN114926797A (en) Transformer double-branch road extraction method and device based on edge constraint and feature adaptation
CN112037244B (en) Landsat-8 image culture pond extraction method combining index and contour indicator SLIC
CN113780307A (en) Method for extracting blue-green space information with maximum regional year
Chouari Wetland land cover change detection using multitemporal Landsat data: A case study of the Al-Asfar wetland, Kingdom of Saudi Arabia
CN111310771A (en) Road image extraction method, device and equipment of remote sensing image and storage medium
CN111339909B (en) Surface linear water body extraction algorithm based on multi-scale region growth
CN111882573B (en) Cultivated land block extraction method and system based on high-resolution image data
CN113705441A (en) High-spatial-temporal-resolution surface water body extraction method cooperating with multispectral and SAR images
Liu A SAM-based method for large-scale crop field boundary delineation
Kulinan et al. Rapid wildfire damage estimation using integrated object-based classification with auto-generated training samples from Sentinel-2 imagery on Google Earth Engine
Magee Segmentation, object-oriented applications for remote sensing land cover and land use classification
CN115661675A (en) Multi-cloud area earthquake landslide remote sensing identification method, system, equipment and storage medium

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