Urban inland river turbid water body extraction method based on Landsat-8 image

Abstract

The invention discloses an urban inland river turbid water body extraction method based on a Landsat-8 image. The method comprises the following steps: acquiring an earth surface reflection characteristic value image of a ground feature after atmospheric influence is removed; respectively collecting samples of typical ground features and turbid water in the Landsat-8 image, and fitting a reflection spectrum curve of the ground features; a plurality of expressions are constructed to enhance signals of a turbid water body in an image and weaken signals of other ground objects in the image, and finally, the plurality of expressions are integrated to construct a new spectral index TWI; the method comprises the following steps: carrying out statistics on a turbid water body sample of a fitting spectrum curve in a TWI image to obtain an initial threshold value for extracting the turbid water body, combining visual observation and adjusting the threshold value to finally obtain an optimal threshold value capable of effectively detecting and extracting the turbid water body, extracting the turbid water body by utilizing the optimal threshold value, and detecting the turbid water body in the Landsat-8 image. The detection method provided by the invention is simple and convenient, can well detect the turbid water body information of the urban inland river in the image, and has a good application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of water body extraction, and particularly relates to a method for extracting turbid water bodies of urban inland rivers based on Landsat-8 images. Background Technique

[0002] China's climate belongs to a typical monsoon climate, with rain and heat in the same period. In summer, the precipitation in some areas is large and concentrated, which makes the sediment content in the rivers within the urban built-up areas increase rapidly during this period, resulting in the water bodies of some urban inland rivers being mostly turbid in summer. The turbid state of urban inland river water bodies reflects the risks of urban rainstorms and floods and environmental quality to a certain extent. Therefore, it is of great significance to identify and extract turbid water bodies of urban inland rivers based on remote sensing technology. At present, there is still a lack of specific and effective methods for detecting turbid water bodies of urban inland rivers. Through the image classification method, the turbid water bodies of urban inland rivers in the image can be detected and extracted to a certain extent, but the effect may be poor in terms of speed and accuracy, and it is not sufficient to meet the actual needs. To extract turbid water bodies by the classification method, it is necessary to determine the types of ground objects, select training samples of different types of ground objects, perform classification, and mask non-urban inland river water body land types, etc., before finally obtaining the information of turbid water bodies of urban inland rivers in the image. The process is complicated and time-consuming, and the extraction effect is not necessarily accurate.

[0003] In view of the current lack of methods for extracting turbid water bodies of urban inland rivers and the possible deficiencies in using traditional methods for extraction, there is an urgent need to propose a method for efficiently, practically, and accurately detecting and extracting turbid water bodies of urban inland rivers in order to effectively monitor urban rainstorms, floods, etc. Summary of the Invention

[0004] In order to overcome the problems such as the complicated process and time-consuming of the classification method for detecting turbid water bodies of urban inland rivers, the present invention uses Landsat-8 images as the data source, constructs a spectral index that can quickly and accurately extract turbid water bodies of urban inland rivers, and detects the turbid water bodies in the image by setting a threshold for the constructed spectral index, thereby improving the speed and accuracy of detecting turbid water bodies of urban inland rivers in the image.

[0005] The technical solution adopted by the present invention to solve the above technical problems is: a method for extracting turbid water bodies of urban inland rivers based on Landsat-8 images, including the following steps,

[0006] Step 1: Perform radiometric calibration, fusion, and atmospheric correction on the 8-band Landsat-8 image of the to-be-detected turbid water body to obtain the surface reflectance characteristic value image of the ground object after removing the atmospheric influence;

[0007] Step 2: Collect a certain number of pixels for typical land features and turbid water bodies in the Landsat-8 image, and fit the reflectance spectrum curve of the land features using the pixel mean of each land type sample.

[0008] Step 3: From the fitted spectral curves, identify band combinations that show significant differences in polygon area and straight-line slope between turbid water bodies and other image categories. Utilizing these differences, construct multiple expressions to enhance the signal of turbid water bodies in the images and weaken the signals of other land features. Finally, integrate these expressions to construct a new spectral index, TWI, capable of detecting turbid water bodies in urban rivers.

[0009] Step 4: Statistically analyze the turbid water samples with fitted spectral curves in TWI images to obtain the initial threshold for extracting turbid water. Combine visual observation with threshold adjustment to finally obtain the most suitable threshold for effectively detecting and extracting turbid water. Use the most suitable threshold to extract turbid water and detect turbid water in Landsat-8 images.

[0010] The above is the basic implementation of the present invention, and further improvements, refinements and limitations can be made on the basis of the above: the calculation method of the new spectral index TWI in step three is as follows,

[0011] S1: Since the spectral curves of turbid water bodies and other image categories show significant differences in the area of ​​the trapezoids formed between bands b6 and b7, we assume that the trapezoid formed by the spectral curve of turbid water bodies is ABCD with a smaller area; and that the trapezoid formed by the spectral curve of unused land is ABEF with a larger area. Using the trapezoid area calculation formula S=(upper base + lower base)×height / 2 (i.e.: (b6+b7)×(λ7-λ6) / 2), we can obtain the first type of feature layer that can distinguish turbid water bodies from other image categories.

[0012] S2: Since the area of ​​the trapezoid ABCD formed by the spectral curve of turbid water body between bands b6 and b7 is much smaller than the area of ​​the trapezoid formed by these two bands in other image categories, the brightness of turbid water body in the feature image calculated by the formula is lower, while the brightness of other categories in the image is higher; In order to increase the brightness of turbid water body and decrease the brightness of other categories, the feature layer obtained in step S1 can be multiplied by -1, which will result in a new feature layer (-1)×(b6+b7)×(λ7-λ6) / 2. In this layer, turbid water body appears as a bright tone, and other categories appear as a dark tone.

[0013] S3: Since the slope of the line segments formed by the turbid water body between the b4 and b7 bands and between the b4 and b6 bands is significantly different from the slopes formed by other image categories between these bands, the line segments GD and GC formed by the spectral curve of the turbid water body are significantly different from the line segments HF and HE formed by the spectral curve of the bare land. The former is negative and the latter is positive. Using the line segment slope calculation formula K = (X1 - X2) / (Y1 - Y2), the line segment slopes of different image categories between the bands b4 and b7 and between b4 and b6 can be calculated, and then the second type of feature layers (b4 - b6) / (λ4 - λ6) and (b4 - b7) / (λ4 - λ7) that can distinguish the turbid water body from other image categories can be obtained; <{

[0014] S4: Since the slope of the line segments formed by the spectral curve of the turbid water body between the b4 and b7 bands and between the b4 and b6 bands is negative and shows a dark tone in the new feature layer, in order to make the turbid water body show a bright tone, it can be achieved by multiplying the feature layer obtained in step S3 by -1. In this way, the second type of improved feature layers (-1)×(b4 - b6) / (λ4 - λ6) and (-1)×(b4 - b7) / (λ4 - λ7) can be obtained;

[0015] S5: Adding the two types of feature layers obtained in steps S2 and S4 can obtain a comprehensive layer that enhances the turbid water body and weakens the information of other ground objects; By adding the two types of layers, finally, the spectral index TWI that can enhance and is beneficial to the extraction of the turbid water body in the river is obtained, and its algorithm is as shown in the formula.

[0016]

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] First, the present invention uses Landsat-8 images as the data source, constructs a spectral index that can quickly and accurately extract the turbid water body in the urban inland river, and detects the turbid water body in the image by setting a threshold for the constructed spectral index, thereby improving the speed and accuracy of detecting the turbid water body in the urban inland river in the image.

[0019] Second, through trapezoidal area calculation, negative processing, and calculation and processing of the line segment slope between bands, the information of the turbid water body in the urban inland river is enhanced in the new feature image, the information of other categories is weakened, the interference of abnormal pixels is reduced, and a spectral index for detecting the turbid water body in the urban inland river is created, which is conducive to detecting and extracting the turbid water body in the image through the threshold method;

[0020] Third, the present invention can effectively detect the turbid water body in a complex urban environment. The detection process for this land type is simpler, more convenient, and faster than the classification method, and has significant advantages over the classification method. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1This is a Landsat-8 image of the test area for this invention.

[0022] Figure 2 The spectral curves of turbid water bodies and other land features are provided in this invention.

[0023] Figure 3 This is a schematic diagram illustrating the calculation of trapezoidal area and line segment slope in this invention.

[0024] Figure 4 This is the grayscale image of the TWI index extracted in this invention;

[0025] Figure 5 This invention relates to the detection results of turbid water bodies in urban rivers. Detailed Implementation

[0026] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and tables.

[0027] The embodiments of the present invention are as follows:

[0028] A method for extracting turbid water from urban rivers based on Landsat-8 imagery includes the following steps:

[0029] Step 1: Use ENVI 5.4 software to perform radiometric calibration, fusion, and rapid atmospheric correction on the 8-band Landsat-8 imagery (parameters shown in Table 1), and then crop out a test area. Figure 1 ).

[0030] Table 1: Landsat-8 Data Parameter Table

[0031]

[0032] Step 2: Collect samples of a certain number of pixels for each of the following image categories in the Landsat-8 imagery of the test area: turbid urban river water, vegetation, and shadows (72 samples for built-up land, 60 samples for unused land, 60 samples for vegetation, 80 samples for turbid urban river water, 50 samples for clear urban river water, and 40 samples for shadows). Fit the reflectance spectral curves of the six image categories using the pixel mean of the samples for each category. Figure 2 );

[0033] Step 3: From the fitted spectral curves, identify band combinations that show significant differences between turbid water bodies and other image categories in terms of polygon area and band difference. Utilizing these differences, construct multiple expressions to enhance the signal of turbid water bodies in the images and weaken the signals of other land features. Finally, integrate these expressions to construct a new spectral index, TWI (Turbidity Water Index), capable of detecting turbid water bodies in urban rivers.

[0034] As a further illustration, the calculation method of TWI described in Step 3 is as follows:

[0035] S1: Since the trapezoidal area difference formed between the spectral curves of turbid water bodies and other image categories in the b6 and b7 bands is relatively significant (e.g., Figure 3 , for the trapezoid ABCD formed by the spectral curve of the turbid water body, the area is smaller; for the trapezoid ABEF formed by the spectral curve of unused land, the area is larger), using the trapezoidal area calculation formula S=(upper base + lower base)×height / 2 (i.e., (b6 + b7)×(λ7 - λ6) / 2), the first type of feature layer that can distinguish turbid water bodies from other image categories can be obtained;

[0036] S2: Since the area of the trapezoid ABCD formed by the spectral curve of the turbid water body in the bands b6 and b7 is much smaller than the areas of the trapezoids formed by these two bands for other ground objects, the brightness of the turbid water body in the feature image obtained by formula calculation is lower, while the brightness of other categories in the image is higher. In order to make the brightness of the turbid water body higher and the brightness of other categories lower, it can be achieved by multiplying the feature layer obtained in Step S31 by -1. In this way, a new feature layer (-1)×(b6 + b7)×(λ7 - λ6) / 2 will be obtained. In this layer, the turbid water body shows a bright tone, and other categories show a dark tone;

[0037] S3: Since the slope differences of the line segments formed by the turbid water body between the b4 and b7 bands and between the b4 and b6 bands are relatively large compared with those of other image categories in these bands (e.g., Figure 3 , for the line segments GD and GC formed by the spectral curve of the turbid water body, which are significantly different from the line segments HF and HE formed by the spectral curve of bare land respectively, the former is negative and the latter is positive), using the line segment slope calculation formula K=(X1 - X2) / (Y1 - Y2), the line segment slopes of different image categories between the b4 and b7 bands and between the b4 and b6 bands can be calculated, and thus the second type of feature layer (b4 - b6) / (λ4 - λ6), (b4 - b7) / (λ4 - λ7) that can distinguish turbid water bodies from other image categories can be obtained;

[0038] S4: Since the slope of the line segments formed by the spectral curve of the turbid water body between the b4 and b7 bands and between the b4 and b6 bands is negative and shows a dark tone in the new feature layer, in order to make the turbid water body show a bright tone, it can be achieved by multiplying the feature layer obtained in Step S3 by -1. In this way, the second type of improved feature layer (-1)×(b4 - b6) / (λ4 - λ6), (-1)×(b4 - b7) / (λ4 - λ7) can be obtained;

[0039] S5: By adding the two feature layers obtained in steps S2 and S4, a comprehensive layer that enhances turbid water quality and weakens other land cover information can be obtained. Adding the two layers finally yields the spectral index TWI, which enhances and facilitates the extraction of turbid river water. The algorithm is shown in formula (1). The TWI grayscale image extracted using formula (1) is shown below. Figure 4 As shown.

[0040]

[0041] S4: Statistical analysis of turbid water samples with fitted spectral curves in TWI images was performed to obtain an initial threshold for extracting turbid water. This threshold was then adjusted based on visual observation to ultimately obtain the most suitable threshold for effectively detecting and extracting turbid water [1780, 2891]. This threshold was used to extract turbid water, detecting turbid water in Landsat-8 images. Figure 5 Furthermore, the overall accuracy of the method for detecting turbid water bodies in urban rivers was verified to be 98.72%, with a kappa coefficient of 0.97.

[0042] As can be seen from the above embodiments, the method for detecting turbid water bodies in urban rivers based on Landsat-8 imagery proposed in this invention is simple and convenient, and can effectively detect turbid water body information in urban rivers in the imagery, indicating that the method for detecting turbid water bodies in urban rivers based on Landsat-8 satellite imagery proposed in this invention has excellent performance.

[0043] The preferred embodiments and examples of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments and examples. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the concept of the present invention.

Claims

1. A method for extracting turbid water bodies in urban rivers based on Landsat-8 imagery, characterized in that: Includes the following steps, Step 1: Perform radiometric calibration, fusion, and atmospheric correction on the 8-band Landsat-8 image of the turbid water body to be detected to obtain the surface reflectance characteristic image of the ground features after removing atmospheric effects; Step 2: Collect a certain number of pixels for typical land features and turbid water bodies in the Landsat-8 image, and fit the reflectance spectrum curve of the land features using the pixel mean of each land type sample. Step 3: From the fitted spectral curves, identify band combinations that show significant differences in polygon area and straight-line slope between turbid water bodies and other image categories. Utilizing these differences, construct multiple expressions to enhance the signal of turbid water bodies in the images and weaken the signals of other land features. Finally, integrate these expressions to construct a new spectral index, TWI, capable of detecting turbid water bodies in urban rivers. Step 4: Statistically analyze the turbid water samples with fitted spectral curves in TWI images to obtain the initial threshold for extracting turbid water. Combine visual observation with threshold adjustment to finally obtain the most suitable threshold for effectively detecting and extracting turbid water. Use the most suitable threshold to extract turbid water and detect turbid water in Landsat-8 images.

2. The method for extracting turbid water bodies in urban rivers based on Landsat-8 imagery as described in claim 1, characterized in that: The calculation method for the new spectral index TWI in step three is as follows. S1: Since the spectral curves of turbid water bodies and other image categories show significant differences in the area of ​​the trapezoids formed between bands b6 and b7, we assume that the trapezoid formed by the spectral curve of turbid water bodies is ABCD with a smaller area; and that the trapezoid formed by the spectral curve of unused land is ABEF with a larger area. Using the trapezoid area calculation formula S=(upper base + lower base)×height / 2 (i.e.: (b6+b7)×(λ7-λ6) / 2), we can obtain the first type of feature layer that can distinguish turbid water bodies from other image categories. S2: Since the area of ​​the trapezoid ABCD formed by the spectral curve of turbid water body between bands b6 and b7 is much smaller than the area of ​​the trapezoid formed by these two bands in other image categories, the brightness of turbid water body in the feature image calculated by the formula is lower, while the brightness of other categories in the image is higher; In order to increase the brightness of turbid water body and decrease the brightness of other categories, the feature layer obtained in step S1 can be multiplied by -1, which will result in a new feature layer (-1)×(b6+b7)×(λ7-λ6) / 2. In this layer, turbid water body appears as a bright tone, and other categories appear as a dark tone. S3: Because the slope of the line segments formed by turbid water bodies between bands b4 and b7, and between bands b4 and b6, differs significantly from the slopes formed by other image categories between these bands, the line segments GD and GC formed by the spectral curves of turbid water bodies differ significantly from the line segments HF and HE formed by the spectral curves of bare land, respectively. The former is negative and the latter is positive. Using the line segment slope calculation formula K=(X1-X2) / (Y1-Y2), the line segment slopes of different image categories between bands b4 and b7, and between b4 and b6 can be calculated, thus distinguishing the second type of feature layers (b4-b6) / (λ4-λ6) and (b4-b7) / (λ4-λ7) between turbid water bodies and other image categories. S4: Since the slope of the line segment formed by the spectral curve of turbid water between bands b4 and b7, and b4 and b6 is negative, it appears dark in the new feature layer. In order to make the turbid water appear bright, the feature layer obtained in step S3 can be multiplied by -1. This will give us the second type of improved feature layer (-1)×(b4-b6) / (λ4-λ6) and (-1)×(b4-b7) / (λ4-λ7). S5: Adding the two types of feature layers obtained in steps S2 and S4 yields a comprehensive layer that enhances turbidity in the water and weakens information about other land features. Adding the two types of layers finally yields the spectral index TWI, which enhances and facilitates the extraction of turbid river water. The algorithm is shown in the formula.

Images