A water extraction method integrating spectral and spatial information

Through the method of integrating spectral and spatial information, combined with water body index and roughness, the problems of leakage and misrelease in complex environments of traditional water body extraction methods are solved, achieving higher accuracy and completeness of water body extraction.

CN114170495BActive Publication Date: 2024-12-06CHINA UNIV OF GEOSCIENCES (WUHAN)
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111507682.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-10
Publication Date
2024-12-06
Estimated Expiration
2041-12-10

AI Technical Summary

Technical Problem

Traditional water body extraction methods based on spectral information are prone to leakage and misrelease in complex geographical environments, and cannot effectively ensure the accuracy of water bodies, especially the extraction accuracy at different scales and times.

Method used

Using the method of integrating spectral and spatial information, the water body research area is divided into initially determined water bodies, initial potential water bodies and initial non-water bodies. Through reclassification and spatial areas, the water body index and roughness are connected, and the water body is further screened and judged to finally obtain the water body. accurate classification results.

Benefits of technology

It improves the accuracy and completeness of water body extraction, reduces leakage and errors in lifting, and is suitable for water body extraction of different scales and times.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114170495B_ABST
    Figure CN114170495B_ABST
Patent Text Reader

Abstract

The present invention provides a water body extraction method integrating spectral and spatial information. First, the water body research area is divided into three categories according to spectral information: initial determined water body, initial potential water body and initial non-water body; then the initial potential water body is reclassified to obtain potential water body and non-water body, and the final determined water body and the final potential water body are further screened out according to the roughness of the potential water body; finally, according to the final determined water body and the final potential water body, the area to be determined is obtained, the target pixels in the area to be determined are connected to the spatial region, and then each connected region is judged to obtain the final water body classification result of the water body research area. The beneficial effects of the present invention are: the water body extraction method is simple and easy to operate, and the accuracy and completeness of water body extraction are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of water body extraction, and in particular to a water body extraction method integrating spectral and spatial information. Background Art

[0002] Water bodies are the basic guarantee for economic development and ecological environmental protection. In recent years, due to global warming, extreme weather has occurred frequently, and natural disasters caused by abnormal rainfall (such as floods) have brought huge economic losses and threats to society. Therefore, it is of great significance to quickly and accurately obtain surface water information before and after disasters for rapid disaster assessment and prevention. In addition, monitoring the time series changes of surface water and understanding the spatial distribution characteristics and temporal change laws of water bodies are conducive to the management and planning of water resources in the country. Remote sensing technology provides an effective and feasible technical means for surface water detection. Extracting water bodies using remote sensing images has become one of the important topics in remote sensing research. Among them, optical remote sensing has the characteristics of "what you see is what you get". It provides rich spectral and geometric information of ground objects, which makes it possible to quickly and accurately draw surface water. Generally speaking, the methods of extracting water bodies using optical images can be divided into two categories: classification method and threshold method.

[0003] The water extraction method based on classification technology identifies water bodies based on the different reflectance spectral characteristics of water bodies and other land objects in different bands. With the development of artificial intelligence technology, machine learning algorithms have played an important role in surface water mapping. At present, the classification methods based on machine learning that are widely used mainly include support vector machine (SVM), decision tree method and object-oriented method. However, the accuracy of water mapping using supervised classification depends on the quality of training samples. It costs a lot to obtain comprehensive and extensive training samples, which to some extent limits the widespread application of classification methods in water extraction.

[0004] Unlike the classification method, the threshold method only uses spectral information to distinguish target objects from background objects. It has a simple principle and high efficiency, and can be effectively used for water monitoring and time series analysis. In optical images, the absorption rate of water bodies in the blue-green light band is relatively low, and it has a strong absorption effect in other bands, especially the infrared band, which makes water bodies obviously different from other objects. Therefore, water body extraction can be simply considered as a binary classification problem. Based on the spectral characteristics of water bodies, the threshold method obtains surface water information by selecting appropriate thresholds in one or more bands, and has been widely used. However, different background information has different effects on the accuracy of water body extraction. It is difficult to maintain high accuracy by setting thresholds in only one or more bands. In order to ensure the effective use of the threshold method, many researchers have constructed a variety of water body indexes through mathematical operations based on different background objects and combined with the spectral characteristics of water bodies, further increasing the spectral difference between water bodies and other objects. In practical applications, the water body index can be selected according to the characteristics of the objects in the study area, and the appropriate threshold can be selected on this basis to finally achieve high-precision water body extraction.

[0005] The accuracy of water body extraction depends not only on the selected water body index, but also on the threshold. Among them, visually selecting the threshold can obtain more accurate results, but this method is time-consuming and labor-intensive and is not suitable for global use. Automatic threshold selection has stronger universality. Among them, the Otsu threshold method is considered to be one of the best methods for threshold selection. This method obtains the threshold by maximizing the difference between the target and the background. It does not require parameters, is simple to calculate, and is not affected by image brightness and contrast. Therefore, it is often used for binary classification. However, in some complex geographical environments, relying on only one threshold often leads to the phenomenon of wrong extraction and omission of water bodies. The threshold method cannot guarantee the accuracy of extracting water bodies at different scales and times. On the one hand, some low-reflectivity objects such as dark roads and building shadows have spectral characteristics similar to those of water bodies, and it is difficult for the water body index to distinguish this part of the noise. Although high-reflectivity objects such as metal buildings have different spectral characteristics from water bodies, they may also have water body index values ​​similar to those of water bodies. These errors are often introduced when drawing water bodies using a single threshold. On the other hand, due to the influence of climatic conditions, surrounding objects, and water quality, the spectral characteristics of water bodies in different regions are different. Even in the same region, there are slight differences in the spectral characteristics of boundary water bodies and central water bodies. The variability of water body spectral information often causes the threshold method to miss some information.

[0006] Therefore, the traditional water body extraction method based on spectral information has certain limitations, and it cannot effectively guarantee the accuracy of the water body. In order to extract complete and accurate water body information, the present invention proposes a water body extraction method integrating spatial information, which further reduces the omission and wrong extraction errors caused by the threshold method by utilizing the spatial distribution information of the water body. Summary of the invention

[0007] In order to solve the above problems, the present invention provides a water body extraction method integrating spectral and spatial information, which mainly includes the following steps:

[0008] S1: The water body study area is divided into three categories according to the spectral information: initial determined water body, initial potential water body and initial non-water body;

[0009] S2: reclassifying the initial potential water body to obtain potential water bodies and non-water bodies, and further screening out final determined water bodies and final potential water bodies according to the roughness of the potential water body;

[0010] S3: According to the final determined water body and the final potential water body, the area to be determined is obtained, the target pixels in the area to be determined are connected in spatial regions, and then each connected region is judged to obtain the final water body classification result of the water body study area.

[0011] Furthermore, the classification criteria in step S1 are:

[0012] C1={i∈V|WI i >T pure} 1-1

[0013] P1={i∈V|T mixed <WI i ≤T pure} 1-2

[0014] N1={i∈V|WI i ≤T mixed} 1-3

[0015] Where V is the pixel set of the water body research area, T pure is a larger threshold, T mixed For a smaller threshold, WI i is the water index value of the ith pixel, C1 is the initial determined water body, P1 is the initial potential water body, and N1 is the initial non-water body.

[0016] Furthermore, the threshold T pure , T mixed The calculation method is shown in formula 1-4 and 1-5:

[0017] T pure =max{T s ,0.5×(M1+std1)} 1-4

[0018] T mixed = min{T s ,0.5×(M2-std2)} 1-5

[0019] Where, T S is a given global threshold, using the global threshold T S The water index WI is divided into two parts. The water index is greater than or equal to T S The part and water index are less than T S The part, M1, std1 represent the water index greater than or equal to T S The median and standard deviation of the part, M2, std2 represent the water index less than T S The median and standard deviation of the parts.

[0020] Furthermore, the specific implementation process of step S2 is:

[0021] S2.1: Use the short-wave infrared band to re-judge the initial potential water body. The re-judgement criteria are:

[0022] P2={i∈P1|ρ SWIR,i <T SWIR} 2-1

[0023] In the formula, ρ SWIR,i is the reflectivity of the i-th pixel in the SWIR band, T SWIR For a given empirical threshold, the part that meets the formula is classified as potential water body P2, and the part that does not meet the formula is classified as non-water body N2;

[0024] S2.2: Calculate the roughness of the potential water body P2. The calculation formula is shown in Formula 2-2:

[0025]

[0026] Where WIR i is the roughness of the i-th pixel, N i is the neighborhood of pixel i, j is the neighborhood N i The pixel j in WI j is the water index value of pixel j, For the neighborhood N i The maximum water index value within the

[0027] S2.3: Further screen out and determine water body C2 and potential water body P according to the roughness of potential water body P2. The specific discrimination formulas are shown in formulas 2-3 and 2-4:

[0028] C2={i∈P2|WIR i >T WIR} 2-3

[0029] P={i∈P2|WIR i ≤T WIR} 2-4

[0030] Where, T WIR is the roughness threshold;

[0031] Step 2.4: Calculate the union of the initially determined water bodies C1 and C2 to obtain the final determined water body C.

[0032] Furthermore, the specific implementation process in step S3 is:

[0033] S3.1: Calculate the union of the final potential water body and the final determined water body to obtain the area to be judged, and the pixels contained in the area are called target pixels;

[0034] S3.2: Connecting the target pixel to a spatial region, that is, traversing all target pixels by a neighborhood method, searching for adjacent pixels around the target pixel to form a pixel set as a connected region;

[0035] S3.3: Judge each connected area obtained in step S3.2 to obtain the final water body extraction result of the water body study area.

[0036] Furthermore, in step S3.1, a four-neighborhood or eight-neighborhood method is used to connect the spatial regions.

[0037] Furthermore, in step S3.3, the judgment rule for water body extraction for each connected area is as follows:

[0038] ① If all pixels in the connected area are "final determined water body" pixels, the area is determined to be a water body;

[0039] ② If the connected area contains both "final confirmed water body" pixels and "final potential water body" pixels, and the number of "final potential water body" pixels accounts for more than 10% of the total number of pixels or the corresponding water body index mean is greater than the given global threshold T S , then the area is also determined to be a water body;

[0040] ③ If all pixels in the connected area are “final potential water body” pixels, the area is judged as non-water body.

[0041] The technical solution provided by the present invention has the beneficial effects of: the water body extraction method is simple and easy to operate, and the accuracy and completeness of water body extraction are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0043] Figure 1 1 is a flow chart of the water extraction method in an embodiment of the present invention.

[0044] Figure 2 Schematic diagram of the MNDWI image obtained in the embodiment of the present invention.

[0045] Figure 3 It is a SWIR band diagram obtained in an embodiment of the present invention.

[0046] Figure 4 Schematic diagram of the roughness image obtained in step 2 of the embodiment of the present invention.

[0047] Figure 5 Schematic diagram of an image of a region to be identified obtained in an embodiment of the present invention.

[0048] Figure 6 Schematic diagram of the final water body image obtained in the embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0050] An embodiment of the present invention provides a water body extraction method that integrates spectral and spatial information. The method first divides the study area into an initial determined water body, an initial potential water body and an initial non-water body with the help of spectral information; then, by reclassifying the "initial potential water body", potential water bodies and non-water bodies are obtained. In order to further extract the remaining water bodies from the "potential water bodies", a concept of water body index roughness based on local spatial information is proposed, so that part of the water bodies in the "potential water bodies" are further divided into "determined water bodies" to obtain final determined water bodies and final potential water bodies. Finally, based on the spatial relationship between the "final determined water bodies" and the "final potential water bodies", the final water body extraction result is obtained.

[0051] Please refer to Figure 1 , Figure 1 : is a flow chart of a water body extraction method integrating spectral and spatial information in an embodiment of the present invention, which specifically includes the following steps:

[0052] Step 1, such as Figure 2 As shown, Figure 1 is the MNDWI image obtained in this example. First, according to formulas 1-1 and 1-2, two thresholds are calculated: the larger threshold T pure and a smaller threshold T mixed ,

[0053] T pure =max{T s ,0.5×(M1+std1)} 1-1

[0054] T mixed = min{T s,0.5×(M2-std2)} 1-2

[0055] Where, T S is a given global threshold. In this embodiment, T S =0.1. Using this global threshold T S The water index MNDWI can be divided into two parts. The water index is greater than or equal to T S The part and water index are less than T S The first part contains most of the water bodies in the study area. M1 and std1 represent the median and standard deviation of the first part respectively. S The part is the second part, which contains a large amount of non-water bodies and a small amount of water bodies. M2 and std2 represent the median and standard deviation of the second part respectively. pure , T mixed -0.1 and 0.3 respectively.

[0056] Then, according to equations 1-3, 1-4, and 1-5, the acquired MNDWI images are divided into three categories, namely, the initial determined water body C1, the initial potential water body C2, and the initial non-water body N1:

[0057] C1={i∈V|WI i >T pure} 1-3

[0058] P1={i∈V|T mixed <WI i ≤T pure} 1-4

[0059] N1={i∈V|WI i ≤T mixed} 1-5

[0060] In the formula, C1 is the initial determined water body, P1 is the initial potential water body, N1 is the initial non-water body, V is the pixel set of the area to be classified, WI i is the MNDWI value of pixel i, that is, the water index value of the i-th pixel, T pure is a larger threshold, T mixed is a smaller threshold.

[0061] Step 2: Reclassify the initial potential water body P1. First, according to formula 2-1, use Figure 3 The SWIR bands shown divide the initial potential water body P1 into potential water body P2 and non-water body N2:

[0062] P2={i∈P1|ρ SWIR,i <T SWIR} 2-1

[0063] In the formula, ρ SWIR,i is the reflectivity of pixel i in the short-wave infrared (SWIR) band, T SWIR is a given empirical threshold, in this embodiment, T SWIR =0.1, and the part that does not meet the formula is classified as non-water body N2.

[0064] Then calculate the roughness of the potential water body P2 according to formula 2-2:

[0065]

[0066] Where WIR i is the roughness of pixel i, N i is the neighborhood of pixel i, in this example the neighborhood is 5×5 in size, j is the neighborhood N i The pixel j in WI j is the MNDWI value of pixel j, For the neighborhood N i The maximum MNDWI value within the range, the calculated roughness result is as follows Figure 4 shown.

[0067] Then, according to formula 2-3 and 2-4, the potential water body P2 is divided into the determined water body C2 and the final potential water body P:

[0068] C2={i∈P2|WIR i >T WIR} 2-3

[0069] P={i∈P2|WIR i ≤T WIR} 2-4

[0070] Where, T WIR is the roughness threshold, in this embodiment, T WIR =0.4.

[0071] Then the determined water bodies C1 and C2 are obtained by union to obtain the final determined water body C.

[0072] Step 3: First, find the union of the final potential water body P and the final determined water body C, and get Figure 5 The area to be determined shown in FIG. 1 includes pixels called target pixels.

[0073] Then, the target pixel is connected to the spatial region, that is, all target pixels are traversed by the neighborhood method, and the adjacent pixels around the target pixel are searched to form a pixel set as a connected region. In this embodiment, the 8-neighborhood method is used for traversal.

[0074] Finally, the following judgment is made for each connected area obtained in the previous step, that is, the reclassified "final potential water body" is judged, and the water body in the "final potential water body" is searched according to the spatial distribution of the "final determined water body". That is, if there is a "potential water body" around the "final determined water body" that is closely connected to it, then the "final potential water body" is also considered to be a water body, and the final result is as follows Figure 6 Final water bodies for the water body study area shown:

[0075] ① If all pixels in the connected area are "final determined water body" pixels, the area is determined to be a water body;

[0076] ② If the connected area contains both "final confirmed water body" pixels and "final potential water body" pixels, and the number of "final potential water body" pixels accounts for more than 10% of the total number of pixels or the corresponding water body index mean is greater than the given global threshold T S , then the area is also determined to be a water body;

[0077] ③ If all pixels in the connected area are “final potential water body” pixels, the area is judged as non-water body.

[0078] The beneficial effects of the present invention are: the water body extraction method is simple and easy to operate, and the accuracy and completeness of water body extraction are improved.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A water body extraction method integrating spectral and spatial information, characterized in that: The following steps are involved: S1: The water body study area is divided into three categories based on spectral information: initial determined water body, initial potential water body and initial non-water body; S2: reclassifying the initial potential water body to obtain potential water bodies and non-water bodies, and further screening out final determined water bodies and final potential water bodies according to the roughness of the potential water body; The specific implementation process of step S2 is: S2.1: Use the short-wave infrared band to re-judge the initial potential water body. The re-judgement criteria are: P2={i∈P1|ρ SWIR,i <T SWIR }2-1 In the formula, ρ SWIR,i is the reflectivity of the i-th pixel in the SWIR band, T SWIR is a given empirical threshold. The part that meets the formula is classified as potential water body P2, and the part that does not meet the formula is classified as non-water body N2; S2.2: Calculate the roughness of potential water body P2. The calculation formula is shown in Formula 2-2: Where WIR i is the roughness of the i-th pixel, N i is the neighborhood of pixel i, j is the neighborhood N i The pixel j in WI j is the water index value of pixel j, For the neighborhood N i The maximum water index value within the S2.3: Further screen out and determine water body C2 and potential water body P according to the roughness of potential water body P2. The specific discrimination formulas are shown in formulas 2-3 and 2-4: C2={i∈P2WIR i >T WIR }2-3 P={i∈P2WIR i ≤T WIR }2-4 Where, T WIR is the roughness threshold; S2.4: Calculate the union of the initially determined water bodies C1 and C2 to obtain the final determined water body C, and search for the final potential water body according to the spatial distribution of the final determined water body C; S3: According to the final determined water body and the final potential water body, an area to be determined is obtained, target pixels in the area to be determined are connected in a spatial region, and then each connected region is judged to obtain a final water body classification result of the water body study area; The specific implementation process in step S3 is: S3.1: Calculate the union of the final potential water body and the final determined water body to obtain the area to be judged, and the pixels contained in the area are called target pixels; S3.2: Connecting the target pixel to a spatial region, that is, traversing all target pixels by a neighborhood method, searching for adjacent pixels around the target pixel to form a pixel set as a connected region; S3.3: Judge each connected area obtained in step S3.2 to obtain the final water body extraction result of the water body study area.

2. A water body extraction method integrating spectral and spatial information as claimed in claim 1, characterized in that: The classification criteria in step S1 are: C1={i∈V|WI i >T pure }1-1 P1={i∈V|T mixed <WI i ≤T pure }1-2 N1={i∈V|WI i ≤T mixed }1-3 Where V is the pixel set of the water body research area, WI i is the water index value of the ith pixel, C1 is the initial determined water body, P1 is the initial potential water body, and N1 is the initial non-water body; Threshold T pure , T mixed The calculation method is shown in formula 1-4 and 1-5: T pure =max{T s ,0.5×(M1+std1)}1-4 T mixed =min{T s ,0.5×(M2-std2)}1-5 Where, T S is a given global threshold, using the global threshold T S The water index WI is divided into two parts. The water index is greater than or equal to T S The part and water index are less than T S The part, M1, std1 represent the water index greater than or equal to T S The median and standard deviation of the part, M2, std2 represent the water index less than T S The median and standard deviation of the parts.

3. The water body extraction method integrating spectral and spatial information as claimed in claim 1, characterized in that: In step S3.1, a four-neighborhood or eight-neighborhood method is used to connect the spatial regions.

4. The water body extraction method integrating spectral and spatial information as claimed in claim 1, characterized in that: In step S3.3, the judgment rule for water body extraction in each connected area is as follows: ① If all pixels in the connected area are "final determined water body" pixels, the area is determined to be a water body; ② If the connected area contains both "final confirmed water body" pixels and "final potential water body" pixels, and the number of "final potential water body" pixels accounts for more than 10% of the total number of pixels or the corresponding water body index mean is greater than the given global threshold T S , then the area is also determined to be a water body; ③ If all pixels in the connected area are "final potential water body" pixels, the area is judged as non-water body.