A sporadic house identification method and device and a storage medium

Abstract

The application discloses a sporadic house identification method and device and a storage medium, and the method comprises the following steps: acquiring SAR data; performing threshold segmentation on the acquired SAR data to extract highlight pixels; performing screening processing on the extracted highlight pixels to obtain a highlight pixel point set corresponding to sporadic houses; performing buffer analysis on the highlight pixel point set to eliminate the influence of mixed pixels, and obtaining a highlight pixel point set after the mixed pixels are eliminated; generating a slope graph by using DEM data, calculating a slope average of the highlight pixel point set range, setting a slope threshold, eliminating a point set group with a slope greater than the slope threshold, and obtaining a highlight pixel point set after the slope is eliminated; and taking an intersection between the highlight pixel point set after the mixed pixels are eliminated and the highlight pixel point set after the slope is eliminated as a final sporadic house extraction result. The application fully excavates the application value of a medium-low spatial resolution SAR image, and improves the sporadic house extraction accuracy based on the medium-low resolution SAR image.

Description

[0001] Technology Neighborhood

[0002] This invention relates to the fields of surveying, remote sensing and land resource management technology, and in particular to a method, apparatus and storage medium for identifying scattered houses in medium and low resolution polarimetric SAR images based on prominent small pixels. Background Technology

[0003] Synthetic Aperture Radar (SAR) is an active side-looking radar system. Its imaging geometry is based on slant-range projection, and SAR images differ significantly from optical images in many aspects, such as imaging mechanisms, geometric features, and radiation characteristics. The information in a SAR image reflects the radar beam's response to ground objects, primarily recording the echo signals formed by backscattering from these objects. Factors affecting backscattering can be broadly categorized into two types based on their source: the first is the radar system's operating parameters, including the radar sensor's incident angle, operating wavelength, and polarization; the second is the unique properties of the ground objects, including surface roughness and complex permittivity. Microwaves possess penetrating power, generally increasing with longer wavelengths. Depending on the wavelength and polarization, SAR systems can penetrate vegetation, snow, and bare soil. The electric field vector of the energy pulse emitted by the radar can be polarized in either the vertical (V) or horizontal (H) plane. By controlling the polarization of the signals in the transmission and reception paths, structural information about different imaging surfaces can be obtained. Polarization modes are divided into co-polarization (HH or VV) and cross-polarization (VH / HV). Cross-polarization reflects the electric field vectors in two directions, and its image information is richer than that of co-polarization.

[0004] During SAR imaging, the backscattering effect of bright ground features on the radar beam is very complex. The scattering characteristics of SAR images can be simply summarized as follows:

[0005] 1. Image brightness represents the intensity of backscattering;

[0006] 2. The rougher the surface of a pixel, the stronger the backscattering, and the brighter the image (e.g., a house);

[0007] 3. Smooth surfaces exhibit specular reflection, resulting in weak backscattering and a dark image (such as water bodies);

[0008] 4. It is related to the complex permittivity of the scattering body; the greater the water content, the stronger the backscattering.

[0009] The resolution of a SAR image is divided into azimuth and range. The resolution of a SAR image is usually greater than the size of each pixel. The display of the tone on the SAR image mainly depends on the backscattering cross section of the highlight of the ground object. The echo received by the radar sensor is converted into an electrical signal and recorded on an optical film with a certain gray tone or converted into a digital pixel with a specific value representing brightness. The tone of the radar image is mainly affected by three factors, namely, terrain undulation, surface roughness and complex permittivity (water content), among which the surface roughness plays a key role in determining the strength of the radar echo signal. The cross-polarized echo intensity is usually weaker than the co-polarized echo intensity because the cross-polarized design usually has a higher receiving bandwidth to compensate for the weakened echo signal.

[0010] The penetration feature of the SAR image can effectively make up for the problem of insufficient optical image in the area with complex weather conditions such as cloudy and rainy weather. The SAR image has the ability of all-weather and all-time data acquisition. The high-resolution SAR image can reach centimeter level and is widely used in the military field, but the cost of data acquisition is high. The low-cost medium and low-resolution SAR image (such as Sentinel data) has been applied in various investigation and monitoring work in the identification of macro objects due to its wide coverage and short revisit period, but the value of the data needs to be fully tapped.

[0011] The resolution of a SAR image is divided into azimuth and range. The resolution of a SAR image is usually greater than the size of each pixel. The display of the tone on the SAR image mainly depends on the backscattering cross section of the highlight of the ground object. The echo received by the radar sensor is converted into an electrical signal and recorded on an optical film with a certain gray tone or converted into a digital pixel with a specific value representing brightness. The tone of the radar image is mainly affected by three factors, namely, terrain undulation, surface roughness and complex permittivity (water content), among which the surface roughness plays a key role in determining the strength of the radar echo signal. The cross-polarized echo intensity is usually weaker than the co-polarized echo intensity because the cross-polarized design usually has a higher receiving bandwidth to compensate for the weakened echo signal.

[0012] The penetration feature of the SAR image can effectively make up for the problem of insufficient optical image in the area with complex weather conditions such as cloudy and rainy weather. The SAR image has the ability of all-weather and all-time data acquisition. The high-resolution SAR image can reach centimeter level and is widely used in the military field, but the cost of data acquisition is high. The low-cost medium and low-resolution SAR image (such as Sentinel data) has been applied in various investigation and monitoring work in the identification of macro objects due to its wide coverage and short revisit period, but the value of the data needs to be fully tapped.

[0013] The existing SAR image highlight detection usually carries out filtering pretreatment before edge extraction and image segmentation after the coherent speckle noise filtering. The commonly used filtering methods include low-pass filtering, structure filtering, adaptive filtering and the like. The speckle noise in the filtered image is suppressed to a certain extent, but at the same time, a lot of key image information may be lost, which brings difficulties for the further processing and analysis of the SAR image.

[0014] (1) Point highlight detection. In the point highlight detection of the SAR image, the interference of the background mainly manifests as the coherent speckle noise. Generally, the speckle noise is removed by adopting a proper filtering method, and the contrast between the highlight and the background is increased by adopting a certain image processing method, so as to highlight the point highlight. Or the point highlight on the background is equated to a central prominent function, and a sharpening filter is used for processing, to detect the point highlight.

[0015] (2) Line highlight detection. The line highlight in the SAR image usually has two meanings. One is to represent the boundary line (boundary linear highlight) of different types of highlights, such as the water-land boundary, which is only a linear body serving as the boundary line of the two types of highlights. The other is to represent the highlight itself (self-linear highlight), such as the road and the river. The width of the self-linear highlight is usually small relative to the medium and low resolution SAR image, and is equivalent to the resolution unit size. The boundary linear highlight is caused by the difference in the backscattering coefficients of two types of adjacent ground highlights, and the greater the difference, the easier to identify. The scattering intensity difference between the land and the water causes the water-land boundary to be relatively obvious, while the boundary between different tree species or between different crops in the same land is relatively difficult to identify due to the small difference. The image boundary point is formed at the gray mutation position of the adjacent ground, and the boundary point is connected to form the ground boundary in the image. However, on the SAR image, the noise point also has the gray mutation feature, and attention should be paid to distinguish that the noise point is an isolated random point and is irrelevant to the surrounding points.

[0016] (3) Face highlight detection. In the SAR image, the detection of the face highlight mainly depends on the mean value and the texture. When the mean value is used to distinguish different ground highlights, the contrast (difference in the mean value) between the highlights needs to be large enough. The texture does not depend on the absolute value of the gray scale, and represents the relative change of the color tone in the space. The texture information can be used to extract the ground highlight. The extraction of the face highlight is actually the image segmentation. The coherent speckle noise of the face highlight on the SAR image is obvious, but the homogeneous area affected by the noise generally corresponds to the same gray area on the texture measurement image, so the texture information can be introduced to improve the accuracy of the image segmentation.

[0017] The detection of sporadic houses in non-building areas is an important work of natural resource management. The houses in the scope of farmland, garden land, forest land and grassland are most likely to be illegal buildings, and timely and effective monitoring is of great significance. The existing SAR image building highlight detection mainly uses high-resolution SAR images as data sources. Although the combination of machine learning or deep learning methods for building detection can obtain higher accuracy, the quality and quantity of SAR data sets are required. The cost of medium and low resolution SAR images is lower and easier to obtain, and its role in practical application cannot be ignored. However, there are few related researches on building detection using medium and low resolution SAR images as data sources.

[0018] Sporadic houses are often reflected as highlighted small pixels on medium and low resolution SAR images. The idea of point highlight detection can be used, but on the one hand, the highlight of sporadic houses is small, and the existing technology is difficult to distinguish them from noise points. Therefore, the existing medium and low resolution SAR image recognition is more inclined to the recognition of more macro built-up areas, and there is little research on highlight detection of sporadic houses. On the other hand, the existing technology does not comprehensively consider the difference between the scattering coefficients of sporadic houses and the surrounding objects under different polarization characteristics. The simple use of highlight center function and sharpening filter processing lacks constraints for extracting sporadic houses, and the accuracy of sporadic house extraction is difficult to meet the expected value. SUMMARY

[0019] The purpose of the present application is to overcome the shortcomings of the prior art, provide a medium and low resolution polarimetric SAR image sporadic house recognition method, device and storage medium based on highlighted small pixels, fully exploit the application value of medium and low spatial resolution SAR images, improve the accuracy of sporadic house extraction based on medium and low resolution SAR images, shorten the detection period, reduce the detection cost, and improve the sporadic house detection technology level.

[0020] To achieve the above purpose, the technical scheme of the present application is:

[0021] In a first aspect, the present application provides a sporadic house recognition method, comprising:

[0022] Obtaining SAR data, using threshold segmentation on the obtained SAR data to extract highlighted pixels;

[0023] Screening the extracted highlighted pixels to obtain a set of highlighted pixel points corresponding to sporadic houses;

[0024] Buffer analysis is performed on the set of highlighted pixel points to eliminate the influence of mixed pixels, and a set of highlighted pixel points after eliminating mixed pixels is obtained;

[0025] generating a slope map using DEM data, calculating a slope mean value in a range of the highlighted pixel point set, setting a slope threshold value, removing a point set set whose slope is greater than the slope threshold value to obtain a highlighted pixel point set after removing the slope;

[0026] Taking an intersection between the highlighted pixel point set after removing the mixed pixel and the highlighted pixel point set after removing the slope as a final sporadic house extraction result.

[0027] Further, the SAR data is cross-polarization SAR image.

[0028] Further, before threshold segmentation is performed on the acquired SAR data, the SAR data is preprocessed to obtain backscattering intensity information under the post-geographic coordinates, so as to serve as basic data for extracting sporadic houses.

[0029] Further, the preprocessing includes thermal noise elimination, track correction, radiation scaling, refined Lee filtering, range-Doppler terrain correction, and conversion of backscattering coefficient σ0 linear scale into logarithmic scale dB:

[0030] σ0 (dB) = 10 x log 10 σ0.

[0031] Further, the threshold segmentation performed on the acquired SAR data to extract highlighted pixels includes:

[0032] A threshold value of -10 is set, the SAR data image is binarized, the highlighted pixels are assigned a value of 1, and the non-highlighted pixels are assigned a value of 0 to obtain a binary image.

[0033] Further, the screening processing performed on the extracted highlighted pixels to obtain the highlighted pixel point set corresponding to the sporadic houses includes:

[0034] Calculating neighborhood pixel information: using a 3*3 gray value rectangular window to perform convolution operation on the binary image, summing the highlighted pixels in the window; using a 7*7 rectangular window to sum the outermost highlighted pixels;

[0035] Screening highlighted point set center points: if the pixel satisfies that all the highlighted pixels in the 3*3 gray value window and the non-highlighted pixels in the 7*7 gray value window are in the outermost periphery, it is determined as a highlighted point set center point and is assigned a value of 1, otherwise, it is assigned a value of 0 to obtain a binary image;

[0036] Creating a mask: using a 5*5 gray value rectangular window to perform convolution operation on the binary image obtained in the step of screening the highlighted point set center points, summing the highlighted pixels in the window, if the value is 0, it means that the pixel does not contain a highlighted point set center point, that is, it does not belong to the highlighted point set and needs to be removed by the mask;

[0037] Finally, the high-light pixel point set corresponding to the 3*3 to 5*5 gray value range of the sporadic house is obtained.

[0038] Further, the high-light pixel point set is subjected to buffer analysis to eliminate the influence of mixed pixels, and a high-light pixel point set after elimination of mixed pixels is obtained, comprising:

[0039] The high-light pixel point set is subjected to buffer analysis according to the range and two pixel sizes, the average brightness of the inner circle of the high-light pixel point set and the outer circle of the buffer zone is calculated, the difference is obtained, the threshold is set to 5, the high-light pixel point set with small difference from the adjacent pixels is eliminated, and the high-light pixel point set after elimination of mixed pixels is obtained.

[0040] Further, the slope threshold is 4.

[0041] In the second aspect, the present application provides a sporadic house identification device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize the steps of any one of the above methods.

[0042] In the third aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of any one of the above methods.

[0043] Compared with the prior art, the present application has the following beneficial effects:

[0044] The medium and low resolution SAR image with short return period, such as Sentinel-1A, is suitable for early detection of illegal buildings, but most of the illegal buildings exist in the form of sporadic houses, and the existing medium and low resolution SAR image ground object identification technology cannot well realize the detection of sporadic houses. According to the distribution characteristics of sporadic houses, the present application fully excavates the sporadic house building information contained in the prominent small pixels in the SAR image, effectively filters the information of various non-sporadic house buildings on the basis of fully considering the noise and other ground object influences, realizes the accurate labeling of sporadic houses on single medium and low resolution SAR image, and the identification accuracy can be as high as 90% or more. Compared with the prior art, the present application utilizes the corresponding relationship between sporadic houses and prominent small pixels in the medium and low resolution SAR image, effectively eliminates the influence of speckle noise, and at the same time, the present technology also comprehensively considers the difference between the scattering coefficients of the sporadic house and the surrounding ground objects under the polarization feature and the scattering coefficient of the sporadic house itself, and more accurately labels the sporadic house.

[0045] The application utilizes the corresponding relationship between the prominent micro-pixels and sporadic houses on the medium and low resolution SAR image and optimizes the sporadic house extraction technology on the medium and low resolution SAR image, and only needs the single scene medium and low resolution SAR image, and can be automatically realized by computer programming. The fast and effective identification of the sporadic houses on the single scene medium and low resolution SAR image is realized, and the efficiency of the sporadic house identification can be greatly improved due to the low acquisition cost and short revisit period of the medium and low resolution SAR image, a more convenient and effective method for early discovery of illegal construction in the cloudy and rainy area is provided, and the method has important application value for developing geographic national condition monitoring and land satellite image law enforcement. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The prominent micro-pixels for the sporadic houses on the medium and low resolution SAR image;

[0047] Figure 2 The 3*3 to 5*5 high light point set diagram;

[0048] Figure 3 The overall technical route map of the sporadic house identification method based on the prominent micro-pixels of the medium and low resolution polarimetric SAR image provided by the embodiment of the application;

[0049] Figure 4 The specific implementation route map of the sporadic house identification method based on the prominent micro-pixels of the medium and low resolution polarimetric SAR image provided by the embodiment of the application.

[0050] Figure 5 The composition diagram of the sporadic house identification device provided by the embodiment of the application. DETAILED DESCRIPTION

[0051] The technical scheme of the application will be further described below in combination with the drawings and the embodiments.

[0052] Embodiment 1:

[0053] The sporadic house identification method based on the prominent micro-pixels of the medium and low resolution polarimetric SAR image provided by the embodiment mainly focuses on the prominent micro-pixels, and the prominent micro-pixels mainly refer to a small number of pixel point sets (i.e. high light pixel set) with the gray value mutation on the image, such as Figure 1 For the sporadic houses, the prominent pixels reflected on the medium and low resolution SAR image are easily mistaken for spot noise. The ships in the water area and the slope surfaces of the undulating terrain in the mountain area have the similar local high light characteristics as the houses on the SAR image, and are easy to cause interference.

[0054] Because cross polarization reflects two directions of electric field vector, its image information is more abundant than that of same polarization. According to experiments, the brightness of house highlights in cross polarization SAR images is not much different from that of house highlights in same polarization SAR images, but the number of highlights is more, and the reflected sporadic house detection rate using cross polarization SAR images is higher, therefore, the method uses cross polarization SAR images for sporadic house detection.

[0055] Because cross polarization reflects two directions of electric field vector, its image information is more abundant than that of same polarization. According to experiments, the brightness of house highlights in cross polarization SAR images is not much different from that of house highlights in same polarization SAR images, but the number of highlights is more, and the reflected sporadic house detection rate using cross polarization SAR images is higher, therefore, the method uses cross polarization SAR images for sporadic house detection.

[0056] The scattering intensity of sporadic houses on SAR images is high, and the micro-pixel highlighting is used as a candidate point set, on the one hand, considering that the periphery of sporadic houses is mainly non-built-up area ground, the micro-pixel highlighting is subjected to grid buffering, and the average brightness difference of the micro-pixel highlighting and the buffer zone is subjected to threshold segmentation, and the influence of mixed pixels is removed in the buffer zone, so that the extraction accuracy of sporadic houses can be further improved. On the other hand, houses are usually located on ground with gentle slope, and the slope is calculated using DEM data, and the average slope of the preliminarily extracted house range is calculated, and a threshold is set, and the influence of ships and slope is removed. After screening by the above two conditions, the final sporadic house extraction result is obtained.

[0057] After selecting the cross polarization SAR image with medium and low resolution, the scattering intensity information under the post-geographic coordinates is obtained through preprocessing, and is used as the basic data for extracting sporadic houses.

[0058] Referring to FIG. 1, Figure 3 , 4 The method for identifying sporadic houses based on micro-pixel highlighting of medium and low resolution polarimetric SAR images provided by the embodiment mainly includes the following steps:

[0059] (1) Data selection

[0060] The SAR data is selected from the Sentinel-1A image downloaded from the official website of the European Space Agency, the monitoring uses an L1-level ground-range multi-view product, the imaging mode is an interferometric wide swath mode, the polarization mode is VH, the width is 250 kilometers, the range resolution is 5 meters, the azimuth resolution is 20 meters, and the spatial resolution of the image after preprocessing is 10 meters.

[0061] The DEM data of SRTM 3Sec is used, the resolution is 90 meters, is automatically downloaded from the SNAP software network, and is used for extracting the slope and terrain correction of the radar satellite image.

[0062] (2) SAR image preprocessing

[0063] The SNAP software developed by the European Space Agency was used to preprocess the radar data. The data preprocessing mainly included six aspects: thermal noise elimination, orbit correction, radiation calibration, improved Lee filtering, range-Doppler terrain correction, and conversion of the linear scale of backscattering coefficient σ0 to the logarithmic scale dB.

[0064] σ0(dB) = 10 x log 10 σ0

[0065] (3) Extraction of protruding micro-pixels

[0066] 1) Extract high-light pixels. The gray value of the house pixel on the SAR image is high, set the threshold value to -10, binarize the preprocessed SAR image (Image1), and assign the high-light pixels to 1 and the non-high-light pixels to 0.

[0067] 2) Calculate the neighborhood pixel information. Use a 3x3 rectangular window to perform convolution operation on the binary image, and sum the pixels in the window, which is the number of high-light pixels (Image2); use a 7*7 rectangular window to sum the outermost pixels, which is the number of high-light pixels belonging to the high-light pixel set (Image3).

[0068] 3) Select high-light point set center points. If the pixel satisfies that all the 3x3 window is high-light pixel and the 7*7 window is non-high-light pixel (Image2=9, Image3=0), it is determined as the center point of the high-light point set, and is assigned to 1, otherwise, it is assigned to 0 (Image4).

[0069] 4) Create a mask. Use a 5x5 rectangular window to perform convolution operation on the binary image (Image4) of the previous step, sum the pixels in the window (Image5), if the value is 0, it does not contain the high-light point set center point, that is, it does not belong to the high-light point set, and needs to be masked and removed.

[0070] Extract protruding micro-pixels. Apply the mask to the high-light pixel binary image (Image=1, Image5≠0) to obtain the protruding micro-pixel set with a size range of 3x3 to 5x5, as shown in Figure 2 .

[0071] That is, in the highlight micro-pixel extraction link, first, the threshold segmentation is used to extract highlight pixels, and a 3*3 to 5*5 gray value mutation highlight pixel point set corresponding to the sporadic house is obtained, the adjacent pixel information is calculated through the rectangular window, and the point set is screened out, then the point set is generated according to the range and two pixel sizes, the average brightness difference of the inner ring of the micro-pixel and the outer ring of the buffer zone is calculated, the extraction of the highlight micro-pixel point set is avoided to avoid the influence of the spot noise, and the effective differentiation of the sporadic house on the medium and low resolution SAR image is realized.

[0072] (4) Extraction result optimization

[0073] 1) Buffer analysis:

[0074] The buffer zone is generated according to the range and two pixel sizes (20 meters) of the highlight micro-pixel set in the range of 3*3 to 5*5 sizes, the average brightness of the inner ring of the micro-pixel and the outer ring of the buffer zone is calculated, and the difference is obtained, the threshold is set to 5, and the point set with small difference from the adjacent pixel is removed.

[0075] 2) Slope analysis:

[0076] The slope graph is generated using DEM data, the slope average in the range of the highlight micro-pixel set in the range of 3*3 to 5*5 sizes is calculated, the threshold is set to 4, and the point set with large slope is removed.

[0077] The intersection of the highlight micro-pixel point sets screened out in the above two times is taken as the extraction result of the sporadic house, that is, the highlight micro-pixel point set after removing the point set not meeting the above two conditions is the final extraction result of the sporadic house.

[0078] That is, in the extraction result optimization link, first, the threshold segmentation is performed on the average brightness difference of the micro-pixel and the buffer zone, the micro-pixel point set with a gray value mutation exceeding the threshold is screened out, then the slope is calculated using DEM data, the slope average in the range of the micro-pixel point set is calculated, and the threshold is set to remove the influence of ships and slopes, and the extraction result of the sporadic house is obtained. By performing threshold segmentation on the difference between the scattering coefficient of the sporadic house and the scattering coefficient of the surrounding ground object under the polarization feature, the gray mutation feature of the pixel where the sporadic house is located and the surrounding pixel is grasped, and through the terrain analysis, the influence of water area and terrain is removed, and the high-accuracy automatic extraction of the sporadic house is realized.

[0079] In summary, the present application identifies the sporadic house based on the highlight micro-pixel of the medium and low resolution SAR image, fully excavates the micro-independent pixel information of the medium and low resolution SAR image, simultaneously utilizes the gray mutation feature of the sporadic house and the surrounding ground object, and combines terrain analysis to optimize the sporadic house extraction technology on the medium and low resolution SAR image, and greatly improves the extraction accuracy of the sporadic house on the medium and low resolution SAR image.

[0080] Embodiment 2:

[0081] Referring to Figure 5 The sporadic house identifying apparatus provided by the embodiment includes a processor 51, a memory 52, and a computer program 53, such as a sporadic house identifying program, stored in the memory 52 and executable on the processor 51. The processor 51 implements the steps of the above-mentioned embodiment 1 when executing the computer program 53, such as Figure 3 The steps are shown in the figure.

[0082] For example, the computer program 53 can be divided into one or more modules / units, which are stored in the memory 52 and executed by the processor 51 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 53 in the sporadic house identifying apparatus.

[0083] The sporadic house identifying apparatus can be a desktop computer, a notebook, a palm computer, a cloud server, and other computing devices. The sporadic house identifying apparatus can include, but is not limited to, a processor 51 and a memory 52. Those skilled in the art can understand that Figure 5 The figure is only an example of the sporadic house identifying apparatus and does not constitute a limitation on the sporadic house identifying apparatus. The sporadic house identifying apparatus can include more or fewer components than those shown in the figure, or combine certain components, or different components, such as the sporadic house identifying apparatus can also include input / output devices, network access devices, buses, etc.

[0084] The processor 51 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0085] The memory 52 can be an internal storage unit of the sporadic house identifying apparatus, such as a hard disk or a memory of the sporadic house identifying apparatus. The memory 52 can also be an external storage device of the sporadic house identifying apparatus, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, and the like. Further, the memory 52 can include both an internal storage unit and an external storage device of the sporadic house identifying apparatus. The memory 52 is used to store the computer program and other programs and data required by the sporadic house identifying apparatus. The memory 52 can also be used to temporarily store data that has been output or is to be output.

[0086] Embodiment 3

[0087] The embodiment provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the method in the embodiment 1.

[0088] The computer readable medium can be any means that can contain, store, communicate, propagate or transport the program for use by or in connection with the instruction execution system, apparatus or device. More specific examples (non-exhaustive list) of the computer readable medium include the following: electrical connections having one or more wires (electronic devices), portable computer disks (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), fiber optic devices, and portable compact disc read-only memories (CDROMs). In addition, the computer readable medium can even be paper or other suitable medium on which the program is printed, such as by bar code, two-dimensional code or other suitable code, and the like, for example, by optical scanning of the paper or other medium, followed by editing, interpreting or otherwise processing the program in electronic form, and then storing it in a computer memory. The program can be downloaded over a network from a remote computer or a server. From the above description, it can be seen that the embodiments of the present application provide a method and device for identifying sporadic houses.

[0089] The above embodiments are only for illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the essence of the present application should be covered within the protection scope of the present application.

Claims

1. A method for identifying scattered houses, characterized in that, include: Acquire SAR data and use threshold segmentation on the acquired SAR data to extract highlight pixels; The extracted bright pixels are filtered to obtain the set of bright pixels corresponding to scattered houses; A buffer analysis is performed on the set of bright pixels to remove the influence of mixed pixels, resulting in a set of bright pixels after removing mixed pixels. A slope map is generated using DEM data. The average slope is calculated for the range of the bright pixel point set, and a slope threshold is set to remove point sets with slopes greater than the slope threshold, thus obtaining the bright pixel point set after slope removal. The intersection of the set of bright pixels after removing mixed pixels and the set of bright pixels after removing slope is taken as the final result of the scattered houses extraction. A buffer analysis is performed on the set of bright pixels to remove the influence of mixed pixels, resulting in a set of bright pixels after removing mixed pixels, including: For the set of bright pixels, a buffer of two pixels in size is generated according to the range. The average brightness of the inner circle of the set of bright pixels and the outer circle of the buffer is calculated and the difference is obtained. The threshold is set to 5, and the set of bright pixels with small differences from neighboring pixels is removed to obtain the set of bright pixels after removing mixed pixels.

2. The method for identifying scattered houses as described in claim 1, characterized in that, The SAR data is a cross-polarization SAR image.

3. The method for identifying scattered houses as described in claim 1, characterized in that, Before applying threshold segmentation to the acquired SAR data, the SAR data is preprocessed to obtain the scattering intensity information under the geographic coordinates, which serves as the basis for extracting scattered houses.

4. The method for identifying scattered houses as described in claim 3, characterized in that, The preprocessing includes: thermal noise cancellation, orbit correction, radiometric calibration, refined Lee filtering, range-Doppler terrain correction, and backscattering coefficient adjustment. Converting linear scale to logarithmic scale in dB: 。 5. The method for identifying scattered houses as described in claim 1, characterized in that, The step of using threshold segmentation on the acquired SAR data to extract highlight pixels includes: Set a threshold to binarize the SAR data image, assigning a value of 1 to bright pixels and a value of 0 to non-bright pixels, thus obtaining a binary image.

6. The method for identifying scattered houses as described in claim 5, characterized in that, The process of filtering the extracted bright pixels to obtain the set of bright pixel points corresponding to scattered houses includes: Calculate neighborhood pixel information: Perform convolution operation on the binary image using a 3×3 grayscale rectangular window and sum the highlighted pixels within the window; use a 7×7 grayscale rectangular window to sum the outermost highlighted pixels; Filtering the center point of the high-brightness set: If a pixel satisfies the condition that all pixels in the 3×3 grayscale value pane are high-brightness pixels and the outermost part of the 7×7 grayscale value pane are non-high-brightness pixels, then it is determined to be the center point of the high-brightness set and assigned a value of 1; otherwise, it is assigned a value of 0, and a binary image is obtained. Create a mask: Use a 5×5 grayscale rectangular window to perform a convolution operation on the binary image obtained from the step of filtering the center point of the high-brightness set. Sum the high-brightness pixels in the window. If the sum is 0, then there is no center point of the high-brightness set, that is, the pixels do not belong to the high-brightness set and need to be removed by masking. Finally, a set of bright pixels with gray values ​​ranging from 3×3 to 5×5 corresponding to the scattered houses was obtained.

7. The method for identifying scattered houses as described in claim 1, characterized in that, The slope threshold is 4.

8. A scattered housing identification device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images