CN117036222B - Water body detection method, device and medium based on fusion of multi-scale polarimetric SAR images - Google Patents

Abstract

The present application discloses a water body detection method, device and medium for fusing multi-scale polarization SAR images. The method includes obtaining multiple polarization SAR images of the same SAR image data, obtaining a water body marker image corresponding to each polarization SAR image through multi-scale image segmentation operations and image multi-scale fusion operations, performing image denoising processing using Markov random fields and simulated annealing denoising strategies, obtaining multiple single-polarization water body extraction results, and obtaining multi-polarization water body detection results based on the multiple single-polarization water body extraction results. The present application does not require a complex model training process, can quickly identify water body information in SAR images over a large area, is suitable for water body information extraction in complex environments, overcomes noise interference problems through hydrological constraints and elevation constraints, and uses Markov random fields and simulated annealing strategies to achieve global optimization for image denoising. The present application is widely used in the field of SAR image processing technology.

Description

Water body detection method, device and medium based on fusion of multi-scale polarimetric SAR images

Technical Field

The present application relates to the technical field of SAR image processing, and in particular to a water body detection method, device and medium for fusing multi-scale polarization SAR images.

Background technique

Floods are one of the most serious natural disasters in the world. They occur quickly, affect a wide range, and are often accompanied by extreme weather. The visible light band where traditional optical sensors are located has weak penetration ability and is easily affected by climatic conditions, resulting in detection blind spots.

At present, SAR technology is usually used to detect flood information, but the use of SAR technology for flood detection in complex environments such as vegetation and urban areas still needs to be further improved. On the one hand, the radar signals in these places may be quite complex and difficult to predict. Smooth surfaces may produce specular scattering, and extreme weather such as heavy rainfall and strong winds may cause the water surface to be rough, resulting in a decrease in the backscattering contrast between water and non-water bodies, affecting classification accuracy. On the other hand, due to the influence of the blind spot of the ground slope and the sudden change of radar wave backscattering, noise spots similar to water body characteristics are likely to appear on the image, resulting in a large number of false detections in water body detection results.

Summary of the invention

In order to solve at least one technical problem existing in the above-mentioned related technologies, the embodiments of the present application propose a water body detection method, device and medium by fusing multi-scale polarization SAR images.

A first aspect of an embodiment of the present application proposes a water body detection method by fusing multi-scale polarization SAR images, comprising:

Acquire multiple polarimetric SAR images of the same SAR image data;

For each of the polarimetric SAR images, a multi-scale image segmentation operation is performed to obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and a plurality of first-scale pixel grayscale histograms and a plurality of second-scale pixel grayscale histograms are determined; the first-scale pixel grayscale histogram is a pixel grayscale histogram of the first-scale image block; the second-scale pixel grayscale histogram is a pixel grayscale histogram of the second-scale image block;

Determine a first-scale water body portion and a second-scale water body portion according to the first-scale pixel grayscale histogram and the second-scale pixel grayscale histogram; the first-scale water body portion is a water body area included in the first-scale image block; the second-scale water body portion is a water body area included in the second-scale image block;

According to the spatial relationship of the image blocks, construct an image block parent-child relationship between the first-scale image block and the second-scale image block;

Performing an image multi-scale fusion operation according to the parent-child relationship of the image blocks, the first-scale water body part, and the second-scale water body part to obtain a water body marker image corresponding to each of the polarimetric SAR images;

Performing raster graphic run-length encoding on the water body marker image, and obtaining a water body spot marker image corresponding to each polarimetric SAR image through hydrological constraints and elevation constraints;

Using Markov random field and simulated annealing denoising strategy, image denoising is performed on the water body spot marking image to obtain a single polarization water body extraction result corresponding to each polarization SAR image;

The multipolarized water body detection result is determined according to the plurality of single-polarized water body extraction results.

In some embodiments, the method further comprises:

A water body detection image is generated according to the multi-polarization water body detection result and the SAR image data.

In some embodiments, the step of acquiring multiple polarization SAR images of the same SAR image data specifically includes:

Acquire SAR image data;

The SAR image data is preprocessed and decibelized to obtain a plurality of polarized SAR images; the preprocessing includes focusing processing, multi-view processing, filtering processing, geocoding processing and radiation correction processing.

In some embodiments, the step of performing a multi-scale image segmentation operation on each of the polarimetric SAR images to obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determining a plurality of first-scale pixel grayscale histograms and a plurality of second-scale pixel grayscale histograms specifically includes:

Obtaining a first scale image segmentation size and a second scale image segmentation size;

Segmenting the polarimetric SAR image according to the first-scale image segmentation size to obtain a plurality of first-scale image blocks;

Segmenting the polarimetric SAR image according to the second-scale image segmentation size to obtain a plurality of second-scale image blocks;

For each first-scale image block, counting pixel grayscale histograms to obtain the first-scale pixel grayscale histogram;

For each second-scale image block, pixel grayscale histogram is counted to obtain the second-scale pixel grayscale histogram.

In some embodiments, the step of performing raster graphics run-length encoding on the water body marker image and obtaining the water body spot marker image corresponding to each polarimetric SAR image through hydrological constraints and elevation constraints specifically includes:

Performing raster graphics run-length encoding on the water body marker image to determine a plurality of predetermined water body patches and a plurality of land patches;

Correcting the predetermined water body spots and the land spots through a digital elevation model to obtain a plurality of water body corrected spots;

Using the finite difference method, the average gradient of each water body correction patch is calculated;

According to a preset gradient threshold and an average gradient of the water body correction pattern, determining whether the water body correction pattern is a water body pattern;

The water body spots of each of the polarization SAR images are marked to obtain the water body spot marked image.

In some embodiments, the step of using the finite difference method to calculate the average gradient of each water body correction patch is specifically expressed by the following formula:

Formula 1:

Formula 2:

Formula 3:

Among them, G represents the average gradient of the water correction pattern, dx represents the gradient of the water correction pattern in the x-axis direction, dy represents the gradient of the water correction pattern in the y-axis direction, and f(i,j) represents the elevation value of the pixel coordinates (i,j) on the digital elevation model.

In some embodiments, the step of using Markov random field and simulated annealing denoising strategy to perform image denoising on the water body spot marked image to obtain a single polarization water body extraction result corresponding to each polarization SAR image specifically includes:

Constructing the second-order field energy field of the water body spot mark image through the Markov random field, and calculating the initial energy of the water body spot mark image;

The initial energy is minimized by a simulated annealing noise reduction strategy and an iterative conditional mode strategy to obtain the single polarization water body extraction result.

In some embodiments, the step of constructing the second-order field energy field of the water body spot mark image through the Markov random field and calculating the initial energy of the water body spot mark image is specifically expressed by the following formula:

Among them, E is the initial energy of the water body spot marked image, represents the local energy of the energy group { _xi , _yi , _xj }, h, β, η represent non-negative weights, _xi is the pixel in the water body spot marked image, _xj is the second-order neighboring pixel of _xi , and _yi is the pixel in the noise image corresponding to _xi .

A second aspect of the embodiments of the present application provides a water body detection device that fuses multi-scale polarization SAR images, including:

The first module is used to obtain multiple polarization SAR images of the same SAR image data;

The second module is used to perform a multi-scale image segmentation operation for each of the polarimetric SAR images, obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determine a plurality of first-scale pixel grayscale histograms and a plurality of second-scale pixel grayscale histograms; the first-scale pixel grayscale histogram is a pixel grayscale histogram of the first-scale image block; the second-scale pixel grayscale histogram is a pixel grayscale histogram of the second-scale image block;

The third module is used to determine a first-scale water body part and a second-scale water body part according to the first-scale pixel grayscale histogram and the second-scale pixel grayscale histogram; the first-scale water body part is the water body area included in the first-scale image block; the second-scale water body part is the water body area included in the second image block;

A fourth module is used to construct an image block parent-child relationship between the first-scale image block and the second-scale image block according to the spatial relationship of the image blocks;

A fifth module is used to perform an image multi-scale fusion operation according to the parent-child relationship of the image blocks, the first-scale water body part and the second-scale water body part, to obtain a water body marker image corresponding to each of the polarized SAR images;

The sixth module is used to perform raster graphic run-length encoding on the water body marker image, and obtain a water body spot marker image corresponding to each polarimetric SAR image through hydrological constraints and elevation constraints;

The seventh module is used to perform image denoising on the water body spot marked image by using Markov random field and simulated annealing denoising strategy to obtain a single polarization water body extraction result corresponding to each polarization SAR image;

The eighth module is used to determine the multi-polarization water body detection result according to the plurality of single-polarization water body extraction results.

A third aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a computer program, and when the computer program is executed by a processor, the computer program implements the water body detection method of fusing multi-scale polarimetric SAR images described in the first aspect.

The present application provides a water body detection method, device and medium that fuses multi-scale polarization SAR images. It obtains multiple polarization SAR images of the same SAR image data, obtains a water body marker image corresponding to each polarization SAR image through multi-scale image segmentation operations and image multi-scale fusion operations, uses Markov random fields and simulated annealing noise reduction strategies to perform image denoising, obtains multiple single-polarization water body extraction results, and obtains multi-polarization water body detection results based on the multiple single-polarization water body extraction results. The present application does not require a complex model training process and a large amount of manual participation. It is simple and efficient, can quickly identify water body information in SAR images over a large area, is suitable for water body information extraction in complex environments, reduces noise interference caused by slope blind spots and backscattering mutations through hydrological constraints and elevation constraints, overcomes the noise interference problem, and uses Markov random fields and simulated annealing strategies to achieve global optimization of image denoising.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a water body detection method for fusing multi-scale polarimetric SAR images provided in an embodiment of the present application;

FIG2 is a schematic diagram of image processing effects of a water body detection method for fusing multi-scale polarimetric SAR images provided in an embodiment of the present application;

FIG3 is a schematic diagram of the structure of a water body detection device for fusing multi-scale polarization SAR images provided in an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

Referring to FIG. 1 , FIG. 1 is an optional flow chart of a method for water body detection by fusing multi-scale polarimetric SAR images provided in an embodiment of the present application, the method including but not limited to steps S101 to S108:

Step S101, acquiring multiple polarization SAR images of the same SAR image data;

Step S102, performing a multi-scale image segmentation operation on each polarimetric SAR image, obtaining a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determining a plurality of first-scale pixel grayscale histograms and a plurality of second-scale pixel grayscale histograms;

Step S103, determining a first-scale water body portion and a second-scale water body portion according to the first-scale pixel grayscale histogram and the second-scale pixel grayscale histogram;

Step S104, constructing an image block parent-child relationship between the first-scale image block and the second-scale image block according to the spatial relationship of the image blocks;

Step S105, performing an image multi-scale fusion operation according to the parent-child relationship of the image blocks, the first-scale water body part, and the second-scale water body part, to obtain a water body marker image corresponding to each polarimetric SAR image;

Step S106, performing raster graphics run-length encoding on the water body marker image, and obtaining a water body spot marker image corresponding to each polarimetric SAR image through hydrological constraints and elevation constraints;

Step S107, using Markov random field and simulated annealing denoising strategy to perform image denoising on the water body spot marking image, and obtain a single polarization water body extraction result corresponding to each polarization SAR image;

Step S108, determining a multi-polarized water body detection result according to a plurality of single-polarized water body extraction results.

In some embodiments, the method further comprises step S109:

Step S109: generating a water body detection image according to the multi-polarization water body detection result and the SAR image data.

In step S101 of some embodiments, SAR image data of the area to be detected is obtained, and the SAR image data is preprocessed using the software ENVI, the preprocessing includes focusing processing, multi-view processing, filtering processing, geocoding processing and radiation correction processing. In the preprocessing operation, the filter uses the Refind Lee filtering algorithm to process the image, the window size is set to 5×5, and the other preprocessing operations use the default parameters of SARscape. The preprocessed SAR image data is decibelized to obtain multiple polarization SAR images of the SAR image data under multiple polarization channels.

In step S102 of some embodiments, the first scaled pixel grayscale histogram is a pixel grayscale histogram of the first scaled image block; and the second scaled pixel grayscale histogram is a pixel grayscale histogram of the second scaled image block.

In some embodiments, step S102 may include but is not limited to steps S201 to S205:

Step S201, obtaining a first scale image segmentation size and a second scale image segmentation size;

Step S202, segmenting the polarimetric SAR image according to the first scale image segmentation size to obtain a plurality of first scale image blocks;

Step S203, segmenting the polarimetric SAR image according to the second scale image segmentation size to obtain a plurality of second scale image blocks;

Step S204, for each first-scale image block, counting the pixel grayscale histogram to obtain a first-scale pixel grayscale histogram;

Step S205 : For each second-scale image block, count the pixel grayscale histogram to obtain a second-scale pixel grayscale histogram.

In steps S201 to S203 of some embodiments, optionally, the first-scale image segmentation size is 500×500, and the second-scale image segmentation size is 1000×1000. According to the two groups of image segmentation sizes, image block processing is performed on each polarimetric SAR image respectively, and insufficient parts of image pixels are filled with null values. Each polarimetric SAR image is divided into two groups of image blocks of different scales, the first group of image blocks includes a plurality of the first-scale image blocks, and the second group of image blocks includes a plurality of the second-scale image blocks.

In step S103 of some embodiments, the first-scale water body portion is the water body region included in the first-scale image block; and the second-scale water body portion is the water body region included in the second-scale image block.

Since radar waves are specularly scattered on the surface of water bodies and have a low backscattering coefficient, a normal fitting is performed on the pixel grayscale histogram of each image block to determine the area of water bodies in the image block (the first scale water body part and the second scale water body part). If there is only one peak, it means that it contains only one type of land object, water body or non-water body. It is classified as water body or non-water body according to the pixel value of the histogram peak; if there are two peaks, it means that it contains two types of land objects, water body and non-water body. The KI threshold method is used to segment the image block into two categories, water body and non-water body. The KI threshold discrimination criterion function under Gaussian distribution is specifically expressed as follows:

J(T)＝1+2[ _Pw (T) _lnσw (T)+ _Pn (T) _lnσn (T)]+2H(Ω,T)

H(Ω,T)＝-[ _Pw (T) _lnPw (T)+ _Pn (T) _lnPn (T)]

Among them, J(T) is the criterion function of KI threshold segmentation, which is used to describe the correct classification performance when the threshold is T, H(Ω,T) represents the entropy of the category set Ω∈{w,n}, w represents water body, n represents non-water body, _Pw (T) and _Pn (T) are the prior probabilities of water body and non-water body at threshold T, _σw (T) and _σn (T) are the standard deviations of water body and non-water body. By changing the threshold T and counting the size of J(T), the best segmentation effect is achieved when J(T) reaches the minimum value, and all pixel values less than T in the pixel grayscale histogram of the image block are classified as water body.

In step S104 of some embodiments, a parent-child relationship of image blocks at two groups of image scales is constructed according to the spatial relationship of the image blocks, wherein each child image block (image block at the first scale) corresponds to only one parent image block (image block at the second scale).

In step S105 of some embodiments, two groups of image blocks (first-scale image blocks and second-scale image blocks) are spatially superimposed according to the parent-child relationship of the image blocks, and the water body parts in the image blocks of different scales (image segmentation sizes) are fused in a union manner. After the fusion, the above-mentioned water body marked image is obtained, and the water body part and the non-water body part are marked separately in the water body marked image, thereby greatly reducing the probability of missed detection due to the small area of the water body area at a large scale.

In some embodiments, step S106 may include but is not limited to steps S301 to S305:

Step S301, performing raster graphics run-length coding on the water body mark image to determine a plurality of predetermined water body spots and a plurality of land spots;

Step S302, correcting predetermined water body patches and land patches through a digital elevation model to obtain a plurality of water body corrected patches;

Step S303, using the finite difference method to calculate the average gradient of each water body correction patch;

Step S304, judging whether the water body correction spot is a water body spot according to a preset gradient threshold and an average gradient of the water body correction spot;

Step S305 , marking the water body spots in each polarization SAR image to obtain a water body spot marked image.

In step S301 of some embodiments, raster graphics run-length encoding is used to determine the spots of the water body marking image. According to the spatial connectivity, each piece of land surrounded by water on the water body marking image is marked as a land spot, and each piece of water surrounded by land is marked as a predetermined water body spot, thereby obtaining a plurality of predetermined water body spots.

In step S302 of some embodiments, DEM (digital elevation model) data is introduced, and the image of the predetermined water body patches and land patches determined in step S301 is superimposed on the digital elevation model. According to the fluidity of water, the elevation of the water body should be lower than the elevation of the surrounding land. Therefore, the land patches with elevations lower than the surrounding water bodies are corrected to water body patches, and the predetermined water body patches with elevations higher than the surrounding land are corrected to land patches. This can effectively reduce false detections caused by sudden changes in radar wave backscattering, thereby achieving hydrological constraints.

In step S303 and step S304 of some embodiments, the finite difference method is used to calculate the average gradient of each water body correction patch for secondary determination, and a gradient threshold is set. In theory, the elevation gradient of the water surface should be close to 0. Taking into account the existence of resolution error, a certain tolerance is pre-set according to the DEM resolution. If the average gradient of the water body correction patch is greater than the gradient threshold, it means that this is a false detection caused by the slope blind spot, and it is modified to a land patch. If the average gradient of the water body correction patch is less than the gradient threshold, the water body correction patch is determined to be a water body patch, thereby achieving elevation constraint.

Using the finite difference method, the formula for calculating the water body correction pattern is specifically expressed as follows:

Formula 1:

Formula 2:

Formula 3:

Among them, G represents the average gradient of the water correction pattern, dx represents the gradient of the water correction pattern in the x-axis direction, dy represents the gradient of the water correction pattern in the y-axis direction, and f(i,j) represents the elevation value of the pixel coordinates (i,j) on the digital elevation model.

In step S305 of some embodiments, after the corresponding water body spots in the polarization SAR image are determined, the water body spots are marked to obtain a water body spot marked image.

In some embodiments, step S107 may include but is not limited to steps S401 to S402:

Step S401, constructing a second-order neighborhood energy field of the water body spot marked image through a Markov random field, and calculating the initial energy of the water body spot marked image;

Step S402, the initial energy is minimized by using a simulated annealing denoising strategy and an iterative conditional mode strategy to obtain a water body extraction result.

In steps S401 to S402 of some embodiments, the second-order neighborhood energy field of the image is constructed using the Markov random field, the initial energy of the entire image is calculated, and the ICM (iterative conditional mode) strategy and the simulated annealing denoising strategy are used to achieve energy minimization, so that the denoising effect reaches the global optimum, and the single-polarization water body extraction result corresponding to the polarization SAR image under each single-polarization channel is obtained.

In step S401 of some embodiments, the second-order field energy field of the water body spot mark image is constructed by using the Markov random field, and the initial energy of the water body spot mark image is calculated, which is specifically expressed by the following formula:

Among them, E is the initial energy of the water body spot marked image, represents the local energy of the energy group { _xi , _yi , _xj }, h, β, η represent non-negative weights, _xi is the pixel in the water body spot marked image, _xj is the second-order neighboring pixel of _xi , and _yi is the pixel in the noise image corresponding to _xi .

In step S402 of some embodiments, the ICM strategy calculates the local energy change of the water body spot mark image by changing the current pixel and fixing other pixels. In simulated annealing, the local acceptable probability q is calculated according to the current temperature t and the local energy change, and the calculation formula is as follows:

Wherein, exp(*) represents an exponential function with the natural constant e as the base, which is equivalent to e ^* . E(x ^new ) and E(x _k ) represent the local energy of the current pixel of the water body spot mark image before and after the change, respectively. The temperature t controls the entire algorithm process. As the iterative calculation proceeds, the temperature t eventually approaches 0. The calculation formula is as follows:

Among them, s is a constant that controls the temperature drop rate, k and kmax are the current iteration number and the maximum iteration number. According to the Metropolis criterion, the probability of whether to keep the new state is determined by the following formula:

Among them, q represents the above-mentioned local acceptable probability, _xk represents the current pixel value of the image at the kth iteration, ^xnew represents the new value of the current pixel of the image after the change, _xk+1 represents the current pixel value entering the k+1th iteration after the kth iteration is completed, and ξ is a random number uniformly distributed in the interval [0,1). When q is greater than ξ or q is greater than or equal to 1, the current pixel change is accepted, otherwise it is not accepted.

In step S108 of some embodiments, multiple single-polarization water body extraction results are combined in an intersection manner, that is, when and only when a pixel is a water body under different polarization channels (all single-polarization water body extraction results show that the pixel is part of a water body), the pixel is finally marked as a water body, thereby determining a multi-polarization water body detection result. This multi-polarization fusion method can greatly improve the water body detection result.

Referring to Figure 2, Figure 2 is a schematic diagram of an optional image processing effect of a water body detection method for fusing multi-scale polarization SAR images provided by an embodiment of the present application, wherein, in the first step, different polarization SAR images of the same SAR image data are obtained (step S101); the second step is to preprocess the image (step S202); the third step is multi-scale fusion threshold segmentation (steps S102 to S105); the fourth step is elevation constraint and hydrological constraint (step S106); the fifth step is Markov random field and simulated annealing denoising (step S107); and the sixth step is multi-polarization channel fusion (step S108).

Referring to FIG. 3 , FIG. 3 is an optional structural schematic diagram of a water body detection device for fusing multi-scale polarimetric SAR images provided in an embodiment of the present application, the device comprising:

The first module is used to obtain multiple polarization SAR images of the same SAR image data;

The second module is used to perform a multi-scale image segmentation operation for each polarimetric SAR image, obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determine a plurality of first-scale pixel grayscale histograms and a plurality of second-scale pixel grayscale histograms; the first-scale pixel grayscale histogram is a pixel grayscale histogram of the first-scale image block; the second-scale pixel grayscale histogram is a pixel grayscale histogram of the second-scale image block;

The third module is used to determine the first-scale water body part and the second-scale water body part according to the first-scale pixel grayscale histogram and the second-scale pixel grayscale histogram; the first-scale water body part is the water body area included in the first-scale image block; the second-scale water body part is the water body area included in the second-scale image block;

A fourth module is used to construct an image block parent-child relationship between the first-scale image block and the second-scale image block according to the spatial relationship of the image blocks;

The fifth module is used to perform a multi-scale image fusion operation according to the parent-child relationship of the image blocks, the first-scale water body part and the second-scale water body part, so as to obtain a water body label image corresponding to each polarimetric SAR image;

The sixth module is used to perform raster graphics run-length encoding on the water body marker image, and obtain the water body spot marker image corresponding to each polarimetric SAR image through hydrological constraints and elevation constraints;

The seventh module is used to perform image denoising on the water body spot marked image using Markov random field and simulated annealing denoising strategy to obtain the single polarization water body extraction result corresponding to each polarization SAR image;

The eighth module is used to determine the multi-polarization water body detection result based on the multiple single-polarization water body extraction results.

The specific implementation of the water body detection device for fusing multi-scale polarimetric SAR images is basically the same as the specific implementation of the water body detection method for fusing multi-scale polarimetric SAR images described above, and will not be repeated here.

An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned water body detection method for fusing multi-scale polarization SAR images.

The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The embodiment of the present application provides a water body detection method, device and medium that fuses multi-scale polarization SAR images. It obtains multiple polarization SAR images of the same SAR image data, obtains a water body marker image corresponding to each polarization SAR image through multi-scale image segmentation operation and image multi-scale fusion operation, uses Markov random field and simulated annealing noise reduction strategy to perform image denoising, obtains multiple single-polarization water body extraction results, and obtains multi-polarization water body detection results based on multiple single-polarization water body extraction results. The present application does not require a complex model training process and a large amount of manual participation. It is simple and efficient, can quickly identify water body information in SAR images over a large area, is suitable for water body information extraction in complex environments, reduces noise interference caused by slope blind spots and backscattering mutations through hydrological constraints and elevation constraints, overcomes the noise interference problem, and uses Markov random fields and simulated annealing strategies to achieve global optimization of image denoising.

The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims (9)

1. The water body detection method for fusing the multi-scale polarized SAR image is characterized by comprising the following steps of:

acquiring a plurality of polarized SAR images of the same SAR image data;

Performing multi-scale image segmentation operation on each polarized SAR image to obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determining a plurality of first-scale pixel gray histograms and a plurality of second-scale pixel gray histograms; the first-scale pixel gray level histogram is a pixel gray level histogram of the first-scale image block; the second-scale pixel gray level histogram is the pixel gray level histogram of the second-scale image block;

determining a first-scale water body part and a second-scale water body part according to the first-scale pixel gray level histogram and the second-scale pixel gray level histogram; the first scale water body part is a water body area contained in the first scale image block; the second-scale water body part is a water body area contained in the second-scale image block;

Constructing an image block father-son relationship between the first-scale image block and the second-scale image block according to the spatial relationship of the image blocks;

performing image multi-scale fusion operation according to the image block father-son relationship, the first-scale water body part and the second-scale water body part to obtain a water body mark image corresponding to each polarized SAR image;

Performing grid pattern run-length coding on the water body mark images, and obtaining water body pattern spot mark images corresponding to each polarized SAR image through hydrologic constraint and high Cheng Yaoshu;

carrying out image noise reduction processing on the water body image spot marked image by adopting a Markov random field and a simulated annealing noise reduction strategy to obtain a single polarized water body extraction result corresponding to each polarized SAR image;

Determining a multi-polarization water body detection result according to the single-polarization water body extraction results;

The step of performing grid pattern run-length coding on the water body mark images and obtaining the water body pattern spot mark image corresponding to each polarized SAR image through hydrologic constraint and high Cheng Yaoshu specifically comprises the following steps:

Performing grid pattern run-length coding on the water body marker image to determine a plurality of predetermined water body map spots and a plurality of land map spots;

Correcting the predetermined water body map spots and the land map spots through a digital elevation model to obtain a plurality of water body correction map spots;

Calculating the average gradient of each water body correction map spot by adopting a finite difference method;

Judging whether the water body correction pattern spots are water body pattern spots or not according to a preset gradient threshold value and the average gradient of the water body correction pattern spots;

and marking the water body image spots of each polarized SAR image to obtain the water body image spot marking image.

2. The method for water detection fusing multiscale polarized SAR images according to claim 1, further comprising:

And generating a water body detection image according to the multi-polarization water body detection result and the SAR image data.

3. The method for detecting a water body by fusing multiscale polarized SAR images according to claim 1, wherein said step of acquiring a plurality of polarized SAR images of the same SAR image data comprises:

SAR image data are obtained;

Preprocessing and decibeling are carried out on the SAR image data to obtain a plurality of polarized SAR images; the preprocessing includes focusing processing, multiview processing, filtering processing, geocoding processing, and radiation correction processing.

4. The method for detecting a water body by fusing multi-scale polarized SAR images according to claim 1, wherein said step of performing multi-scale image segmentation operation to obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks and determining a plurality of first-scale pixel gray level histograms and a plurality of second-scale pixel gray level histograms for each of said polarized SAR images specifically comprises:

acquiring a first scale image segmentation size and a second scale image segmentation size;

Dividing the polarized SAR image according to the first scale image dividing size to obtain a plurality of first scale image blocks;

dividing the polarized SAR image according to the dividing size of the second scale image to obtain a plurality of second scale image blocks;

Counting pixel gray scales rectangularities for each first-scale image block to obtain a first-scale pixel gray scale histogram;

and counting the pixel gray level rectangularities for each second-scale image block to obtain the second-scale pixel gray level histogram.

5. The method for detecting a water body by fusing multiscale polarized SAR images according to claim 1, wherein said step of calculating the average gradient of each of said water body corrected patches by finite difference method is specifically represented by the following formula:

formula 1:

formula 2:

Formula 3:

wherein G represents the average gradient of the water body correction pattern spot, dx represents the x-axis gradient of the water body correction pattern spot, dy represents the y-axis gradient of the water body correction pattern spot, and f (i, j) represents the elevation value of the pixel coordinate (i, j) on the digital elevation model.

6. The method for detecting water body by fusing multi-scale polarized SAR images according to claim 1, wherein the step of performing image denoising processing on the water body map spot marker images by using a markov random field and a simulated annealing denoising strategy to obtain a single polarized water body extraction result corresponding to each polarized SAR image specifically comprises the steps of:

constructing a second-order field energy field of the water body image spot marking image through a Markov random field, and calculating initial energy of the water body image spot marking image;

and performing minimization treatment on the initial energy through a simulated annealing noise reduction strategy and an iteration condition mode strategy to obtain the single-polarized water body extraction result.

7. The method for detecting water body by fusing multiscale polarized SAR images according to claim 6, wherein said step of constructing a second-order domain energy field of said water body speckle marker image by markov random field and calculating the initial energy of said water body speckle marker image is specifically represented by the following formula:

Where E is the initial energy of the water body pattern spot marker image, and represents the local energy of the energy group { x _i,y_i,x_j }, h, β, η represents the non-negative weight, x _i is the pixel in the water body pattern spot marker image, x _j is the second-order neighborhood pixel of x _i, and y _i is the pixel in the noise map corresponding to x _i.

8. A water body detection device fusing multiscale polarized SAR images is characterized by comprising:

A first module for acquiring a plurality of polarized SAR images of the same SAR image data;

The second module is used for executing multi-scale image segmentation operation for each polarized SAR image to obtain a plurality of first-scale image blocks and a plurality of second-scale image blocks, and determining a plurality of first-scale pixel gray histograms and a plurality of second-scale pixel gray histograms; the first-scale pixel gray level histogram is a pixel gray level histogram of the first-scale image block; the second-scale pixel gray level histogram is the pixel gray level histogram of the second-scale image block;

A third module for determining a first scale water body portion and a second scale water body portion according to the first scale pixel gray level histogram and the second scale pixel gray level histogram; the first scale water body part is a water body area contained in the first scale image block; the second-scale water body part is a water body area contained in the second-scale image block;

A fourth module, configured to construct an image block parent-child relationship between the first-scale image block and the second-scale image block according to a spatial relationship of the image blocks;

A fifth module, configured to perform an image multi-scale fusion operation according to the image block father-son relationship, the first-scale water body portion, and the second-scale water body portion, and obtain a water body marker image corresponding to each polarized SAR image;

A sixth module, configured to perform raster graphics run-length encoding on the water body marker images, and obtain a water body map spot marker image corresponding to each polarized SAR image through hydrologic constraint and altitude Cheng Yaoshu; the step of performing grid pattern run-length coding on the water body mark images and obtaining the water body pattern spot mark image corresponding to each polarized SAR image through hydrologic constraint and high Cheng Yaoshu specifically comprises the following steps:

Performing grid pattern run-length coding on the water body marker image to determine a plurality of predetermined water body map spots and a plurality of land map spots;

Correcting the predetermined water body map spots and the land map spots through a digital elevation model to obtain a plurality of water body correction map spots;

Calculating the average gradient of each water body correction map spot by adopting a finite difference method;

Judging whether the water body correction pattern spots are water body pattern spots or not according to a preset gradient threshold value and the average gradient of the water body correction pattern spots;

marking the water body image spots of each polarized SAR image to obtain the water body image spot marking image;

A seventh module, configured to perform image noise reduction processing on the water body map spot marker image by using a markov random field and a simulated annealing noise reduction strategy, to obtain a single polarized water body extraction result corresponding to each polarized SAR image;

And an eighth module, configured to determine a multi-polarized water body detection result according to the multiple single-polarized water body extraction results.

9. A computer readable storage medium comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method for water detection of fusion of multiscale polarized SAR images according to any one of claims 1 to 7.