Method for extracting shallow water region of multispectral remote sensing image based on object adaptive ndwi threshold

Abstract

The present application relates to the technical field of remote sensing image processing, and particularly relates to a multispectral remote sensing image shallow water region extraction method based on object adaptive NDWI threshold, comprising the following steps: S1, receiving an image, segmenting the image to generate object regions, and extracting a shallow water region based on the object as a basic unit; S2, calculating NDWI values of each object, determining an optimal NDWI threshold of the shallow water region, and obtaining an initial shallow water region extraction result based on the optimal NDWI threshold; S3, performing optimization processing on the initial shallow water region extraction result in S2 based on an object-based region growing algorithm, and outputting a final remote sensing image shallow water region extraction result, which can effectively overcome the problems that the remote sensing image shallow water region extraction process is prone to be affected by image noise and the optimal NDWI threshold is difficult to accurately determine, and accurate shallow water region extraction results of nearshore and island reef regions under different environments are obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing image processing, and particularly relates to a multispectral remote sensing image shallow water region extraction method based on object adaptive NDWI threshold. BACKGROUND

[0002] At present, different researches have been carried out by domestic and foreign scholars for water body extraction of remote sensing images, which can be divided into two categories: traditional water body extraction method and water body extraction method based on deep learning. The traditional water body extraction method usually needs to design water body index features artificially, and realizes the water body extraction of remote sensing images based on the water body index features. Common water body index features include normalized difference water index (NDWI), modified NDWI (MNDWI), normalized difference multi-band water index (NDMBWI) and contrast difference water index (CDWI), etc. The water body extraction method based on traditional water body index features can obtain good water body extraction results of remote sensing images; but this kind of method is easily affected by image noise, and it is difficult to accurately set the appropriate water body index threshold, so the robustness and universality of the algorithm are weak. Convolutional neural network (CNN) is the most popular deep learning model at present, and the water body extraction method based on CNN has been widely applied to water body extraction of remote sensing images. The water body extraction method based on CNN can obtain good water body extraction effect in different scenes, and compared with the traditional water body extraction algorithm based on artificial design of water body features, this kind of algorithm has stronger robustness and universality. However, the existing water body extraction method based on CNN can only distinguish water body and non-water body, and cannot further distinguish shallow water area (water depth less than 20 meters) and deep water area (water depth greater than 20 meters), while satellite derived bathymetry (SDB) usually needs to obtain shallow water area from remote sensing images, and then performs water depth inversion on the shallow water area. In addition, there is no publicly available remote sensing image shallow water area extraction dataset at present, which greatly limits the research progress of remote sensing image shallow water area extraction based on CNN.

[0003] In the current related research of SDB, two methods of visual interpretation and shallow water extraction based on NDWI are usually used to obtain the shallow water area in remote sensing image. The visual interpretation method can obtain high-precision shallow water area extraction results, but it is low in efficiency and weak in automation. The shallow water extraction method based on NDWI is susceptible to image noise, and the extraction accuracy is highly related to the setting of NDWI threshold value, but the optimal NDWI threshold value is usually difficult to obtain.

[0004] In view of the above problems, the present application provides a multispectral remote sensing image shallow water area extraction method based on object adaptive NDWI threshold value. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a multispectral remote sensing image shallow water area extraction method based on object adaptive NDWI threshold value.

[0006] The present application is realized by the following technical scheme: the multispectral remote sensing image shallow water area extraction method based on object adaptive NDWI threshold value comprises the following steps:

[0007] S1, receiving an image, segmenting the image to generate an object area, and extracting a shallow water area based on the object as a basic unit to eliminate the influence of image noise on the extraction of the shallow water area;

[0008] S2, calculating the NDWI value of each object, determining the optimal NDWI threshold value of the shallow water area, and obtaining an initial shallow water area extraction result based on the optimal NDWI threshold value;

[0009] S3, optimizing the initial shallow water area extraction result in S2 based on the region growing algorithm of the object, and outputting the final remote sensing image shallow water area extraction result.

[0010] In S1, the image is segmented to generate an object area using a simple linear iterative clustering algorithm, which can effectively overcome the influence of image noise such as "pepper and salt" noise on image processing.

[0011] In the simple linear iterative clustering algorithm, the compactness coefficient is 10, and the number of superpixels generated is M / 25, where M is the number of pixels of the image.

[0012] S2 comprises the following sub-steps:

[0013] S2-1, calculating the NDWI value of each object;

[0014] S2-2, sorting the NDWI values of the objects in ascending order;

[0015] S2-3, determining the optimal NDWI threshold value of the remote sensing image shallow water area by using a two-stage maximum inter-class variance;

[0016] S2-4, classifying the objects to obtain the initial shallow water area extraction result of the multispectral remote sensing image.

[0017] The calculation formula of the NDWI value of each object in S2-1 is as follows:

[0018] ;

[0019] wherein, NDWI value of each object, NDWI value of the pixel labeled as in the object; number of pixels contained in the object.

[0020] The object NDWI values in the land, deep water area and shallow water area in S2-3 show a step distribution trend, and based on the step distribution trend, two stages are divided, the inflection points between different stages are obtained, and then the optimal NDWI threshold for the extraction of the shallow water area of the remote sensing image is obtained, and the two stages include stage one and stage two;

[0021] The range of stage one is , number of objects, and the first inflection point threshold is obtained by maximum inter-class variance calculation in Based on the threshold , the influence of the land area on the subsequent extraction of the shallow water area can be eliminated;

[0022] The range of stage two is , corresponding object serial number in ascending order, the second inflection point threshold is obtained by maximum inter-class variance calculation in , which is the optimal NDWI threshold of the shallow water area.

[0023] The objective function of the maximum inter-class variance is as follows:

[0024] ;

[0025] wherein, and are the proportions of the foreground object and the background object in respectively; and are the NDWI mean values of the foreground object and the background object respectively.

[0026] The classification of the objects in S2-4 is in accordance with the following formula: ​

[0027]

[0028] wherein, represents the object category, 1 represents that the object is a shallow water area, and 0 represents that the object is a non-shallow water area;

[0029] All objects determined as 1 are collected, that is, the initial shallow water area extraction result of the remote sensing image.

[0030] The object-based region growing algorithm is used in the S3 to optimize the initial shallow water area extraction result, and the optimization process is to perform region growing with the object as a basic unit, objects of the same category are merged into a region, and then the number of pixels contained in the region is classified and judged, so that the optimization of the initial shallow water area extraction result is realized.

[0031] The calculation formula for classifying and judging the number of pixels contained in the region in the S3 is as follows:

[0032]

[0033] In the formula, represents the region category; 1 and 0 represent the shallow water area and the non-shallow water area category labels respectively; represents the number of pixels contained in the region; and are the lower limit and upper limit thresholds of the number of pixels contained in the region, wherein and are set as and , is the number of image pixels.

[0034] Compared with the prior art, the present application has the following beneficial effects:

[0035] The object-based adaptive NDWI threshold multispectral remote sensing image shallow water area extraction method provided by the present application can effectively overcome the problems that the shallow water area extraction process of the remote sensing image is easily affected by image noise and the optimal NDWI threshold is difficult to accurately determine, and accurate shallow water area extraction results of the near-shore and island reef regions under different environments are obtained.

[0036] The present application divides the image through a simple linear iterative clustering algorithm, generates an object region and takes the object as a basic processing unit, overcomes the defect that the traditional pixel-level method is easily affected by noise interference, and makes the shallow water area extraction result more stable.

[0037] ​The application proposes two-stage maximum inter-class variance calculation, removes land area interference in stage 1, and distinguishes shallow water area from deep water area in stage 2, automatically adapts to NDWI value distribution characteristics of different scenes, does not need to manually set a threshold, and significantly improves the universality of the method in different environments.

[0038] The application optimizes the initial result based on the region growing algorithm of the object, forms a region by merging the same object, and judges the number of region pixels, so that false detection and missed detection caused by interference factors such as clouds, coastal spray, and sea ships can be effectively eliminated, and the final extraction result is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a flowchart of the application;

[0040] Figure 2 is an ascending trend graph of the object NDWI value;

[0041] Figure 3 is an optimization processing result graph of the initial shallow water area extraction result;

[0042] Figure 4 is a region growing diagram based on the object category. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0044] Referring to Figures 1-4 The multispectral remote sensing image shallow water area extraction method based on object adaptive NDWI threshold value includes the following steps:

[0045] S1, receiving an image, segmenting the image to generate an object region, and extracting a shallow water area based on the object as a basic unit to eliminate the influence of image noise on the extraction of the shallow water area;

[0046] The simple linear iterative clustering algorithm is used in S1 to segment the image to generate an object region. The influence of image noise such as “pepper and salt” noise on image processing can be effectively overcome.

[0047] Further, in the process of generating the object by the simple linear iterative clustering algorithm, the color of the image is first converted from the RGB color space to the CIELAB color space; then the color attribute and the spatial attribute of each pixel in the image are combined into a five-dimensional feature vector , all seed points are uniformly distributed in the image, and the pixels in the neighborhood of the seed points are clustered and segmented based on a five-dimensional feature vector to generate the objects. The similarity measure between the pixel and the seed point is calculated as follows.

[0048] ;

[0049] ;

[0050] ;

[0051] wherein, and respectively represent the spectral distance and the spatial distance between the pixel and the seed point; represents the similarity measure distance between the pixel and the seed point; is the five-dimensional feature vector of the pixel; is the five-dimensional feature vector of the seed point; is the distance between the seed points; is the compact coefficient in the simple linear iterative clustering algorithm, which is set to 10 in the present application. The number of generated superpixels is M / 25, and M is the number of pixels of the image.

[0052] S2, calculate the NDWI value of each object, determine the optimal NDWI threshold of the shallow water area, and obtain the initial shallow water area extraction result based on the optimal NDWI threshold;

[0053] The S2 includes the following sub-steps:

[0054] S2-1, calculate the NDWI value of each object;

[0055] The calculation formula of the NDWI value of each object in the S2-1 is as follows:

[0056] ;

[0057] wherein, is the NDWI value of each object, is the NDWI value of the pixel labeled in the object; is the number of pixels contained in the object.

[0058] wherein, , and are the reflectivity of the green band and the near-infrared band of the pixel, respectively.

[0059] S2-2, sort the NDWI values of the objects in ascending order; the sorting result is referred to Figure 2 .

[0060] S2-3, the two-stage maximum inter-class variance is used to determine the optimal NDWI threshold of the shallow water region of the remote sensing image;

[0061] Referring to Figure 2 , the object NDWI values in the land, deep water region and shallow water region in S2-3 have obvious stage ladder distribution trend, based on which two stages are divided, and the optimal NDWI threshold for extracting the shallow water region of the remote sensing image can be obtained by obtaining the inflection point between different stages, further, the two stages include stage one and stage two.

[0062] The range of stage one is , is the number of objects, and the first inflection point threshold is obtained by maximum inter-class variance calculation in , based on the threshold , the influence of the land region on the subsequent extraction of the shallow water region can be eliminated.

[0063] The range of stage two is , is the object serial number in ascending order, and the second inflection point threshold is obtained by maximum inter-class variance calculation in , that is, the optimal NDWI threshold of the shallow water region. The objective function of the maximum inter-class variance is expressed as follows:

[0064]

[0065] ;

[0066] wherein, and are the proportions of the foreground object and the background object in ; and and are the NDWI mean values of the foreground object and the background object.

[0067] S2-4, based on the optimal NDWI threshold , the objects are classified to obtain the initial shallow water region extraction result of the multi-spectral remote sensing image.

[0068] In S2-4, the objects are classified according to the following formula:

[0069]

[0070] wherein, represents the object category, 1 represents that the object is the shallow water region, and 0 represents that the object is the non-shallow water region. ​​​​

[0071] All objects determined as 1 are collected, which is the initial shallow water region extraction result of the remote sensing image.

[0072] S3, the object-based region growing algorithm optimizes the initial shallow water region extraction result in S2, and outputs the final shallow water region extraction result of the remote sensing image.

[0073] Due to the influence of objects such as clouds, coastal spray and sea vessels in the image, there are a small amount of missed detection and false detection in the initial shallow water region extraction result, such as Figure 3 indicated in the initial shallow water region extraction result. Therefore, the initial shallow water region extraction result needs to be further optimized to eliminate the influence of the above objects on the shallow water region extraction. The object-based region growing algorithm in S3 is used to optimize the initial shallow water region extraction result, and the optimization process is as shown in Figure 4 .

[0074] In Figure 4 , the region growing is performed based on objects, objects of the same category are merged into a region, and then the number of pixels contained in the region is classified and judged to realize the optimization of the initial shallow water region extraction result, and the result after optimization is as shown in Figure 3 .

[0075] The calculation formula of the classification judgment of the number of pixels contained in the region in S3 is as follows:

[0076] ;

[0077] In the formula: represents the region category; 1 and 0 represent the shallow water region and the non-shallow water region category label respectively; represents the number of pixels contained in the region; and are the lower limit and upper limit threshold values of the number of pixels contained in the region, wherein and are set to and , is the number of pixels of the image.

[0078] The above only describes the optional embodiments of the present application, and does not limit the patent scope of the present application, and any equivalent structural transformation made by using the content of the present application, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding, characterized in that, Includes the following steps: S1. Receive the image, segment the image to generate the object region, and extract the shallow water area using the object as the basic unit. S2. Calculate the NDWI value of each object, determine the optimal NDWI threshold for the shallow water area, and obtain the initial shallow water area extraction results based on the optimal NDWI threshold. S2 includes the following sub-steps: S2-1. Calculate the NDWI value for each object; S2-2. Sort the objects in ascending order based on their NDWI values. S2-3. The optimal NDWI threshold for shallow water areas in remote sensing images is determined by using two-stage maximum inter-class variance. The NDWI values ​​of objects in the land, deep water and shallow water areas in S2-3 show a step distribution trend. Based on the step distribution trend, it is divided into two stages to obtain the inflection point between different stages, and then obtain the optimal NDWI threshold for extracting the shallow water area of ​​the remote sensing image. The two stages include stage one and stage two. The scope of Phase One is , For the number of objects, in The first inflection point threshold is obtained by calculating the maximum inter-class variance. ; The scope of Phase Two is , In ascending order sorting The corresponding object serial number, in The second inflection point threshold is obtained by calculating the maximum inter-class variance. , This is the optimal NDWI threshold for shallow water areas; S2-4. Classify the objects to obtain the initial shallow water area extraction results from the remote sensing images; S3. The object-based region growing algorithm optimizes the initial shallow water region extraction results in S2 and outputs the final shallow water region extraction results of the remote sensing image.

2. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 1, characterized in that, In step S1, a simple linear iterative clustering algorithm is used to segment the image and generate object regions.

3. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 2, characterized in that, In the simple linear iterative clustering algorithm, the compaction coefficient is 10, and the number of superpixels generated is M / 25, where M is the number of pixels in the image.

4. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 1, characterized in that, The formula for calculating the NDWI value of each object in S2-1 is as follows: ； in, For each object, the NDWI value, The label is included in the object. The NDWI value of a pixel; The number of pixels contained in the object.

5. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 1, characterized in that, Objective function for maximizing inter-class variance It is expressed as follows: ； in, and Foreground and background objects respectively The proportion within; and These are the NDWI mean values ​​for the foreground and background objects, respectively.

6. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 5, characterized in that, The objects are classified in S2-4 according to the following formula: in, Indicates the object category: 1 represents the object as a shallow water area, and 0 represents the object as a non-shallow water area. The collection of all objects that are classified as 1 represents the initial shallow water area extraction result from the remote sensing image.

7. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 1, characterized in that, In step S3, an object-based region growing algorithm is used to optimize the initial shallow water region extraction results. The optimization process involves growing the region using objects as the basic unit, merging objects of the same category into one region, and then classifying and judging the number of pixels contained in the region, thereby optimizing the initial shallow water region extraction results.

8. The method for extracting shallow water areas from multispectral remote sensing images based on object-adaptive NDWI thresholding according to claim 7, characterized in that, The calculation formula for classifying the number of pixels contained in a region in S3 is as follows: ； In the formula: Indicates the area category; 1 and 0 represent shallow water area and non-shallow water area category labels, respectively; Indicates the number of pixels contained in the region; and These are the lower and upper thresholds for the number of pixels contained in the region, respectively. and They were set to and , This represents the number of pixels in the image.

Images