A high-resolution remote sensing-based transmission line house removal identification method

Abstract

The application relates to a high-resolution remote sensing-based power transmission line house removal identification method, which comprises the following steps: 1) high-resolution image acquisition and processing; 2) high-resolution image cutting; 3) establishment of a house removal interpretation database; 4) acquisition of house removal positions; 5) house removal state identification; and 6) interpretation archiving and preservation. The application breaks the limitations of traditional manual monitoring, such as poor timeliness, low data accuracy, high cost and the like, improves work efficiency, has high house removal identification efficiency and high identification accuracy, can guarantee accurate identification of houses removed within the range of boundary lines of power transmission line construction, is convenient for subsequent house removal work arrangement, guarantees the safety of residents, simultaneously ensures the smooth and safe implementation of power transmission line construction, and provides a new monitoring method for subsequent power transmission line house removal identification.

Description

Technical Field

[0001] This invention belongs to the field of environmental remote sensing, specifically relating to a method for identifying buildings demolished along power transmission lines based on high-resolution remote sensing. Background Technology

[0002] During the construction of power transmission lines, due to the influence of electromagnetic radiation, houses within a certain range of the transmission line conductors are not suitable for daily living, and the location of the houses may affect the construction of the transmission line towers.

[0003] Due to the long length of transmission lines, the large number of towers and their dispersed distribution, and the wide range of electromagnetic influence, it is necessary to monitor houses within a certain range of the transmission line conductors in a timely manner in order to protect the safety of residents' living environment and ensure the smooth construction of transmission lines. This will provide data support for house demolition, avoid subsequent economic disputes, and ensure the safety of transmission line construction and residents.

[0004] With the development of the national economy, the scope of power transmission line construction is constantly expanding. Currently, the main method for monitoring the demolition of houses along power transmission lines is manual on-site monitoring. This involves using a measuring tape to check whether the vertical distance between the side conductor and the house meets the requirements of the water conservation plan. If the distance is less than the specified distance, the house is designated for demolition. However, as the subsequent construction of the power transmission line progresses, manual on-site monitoring cannot obtain real-time information on the demolition status of houses, affecting the line construction and exhibiting disadvantages such as poor monitoring timeliness and inaccurate data. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method for identifying houses demolished for power transmission lines based on high-resolution remote sensing. This method breaks through the limitations of traditional manual monitoring, such as poor timeliness, low data accuracy, and high cost, thereby improving work efficiency, ensuring the safety of residents, and ensuring the smooth and safe construction of power transmission lines. It also provides a new monitoring means for the subsequent identification of houses demolished for power transmission lines.

[0006] The technical problem solved by this invention is achieved through the following technical solution:

[0007] A method for identifying demolished buildings along power transmission lines based on high-resolution remote sensing, characterized by the following steps:

[0008] S1. Obtain high-resolution images P before construction and B during construction by using the location information of transmission line towers. Then, through radiometric calibration, atmospheric correction, geometric correction and image fusion processing, eliminate weather, sensor and terrain errors so that high-resolution images P and B can obtain projection coordinates.

[0009] 1) Radiometric calibration is the process of converting the original DN values ​​of an image, i.e., pixel brightness values, into radiometric values;

[0010] L = gain * DN + bias

[0011] Where: L is the radiance value;

[0012] gain is the image gain;

[0013] bias refers to image offset;

[0014] 2) Atmospheric correction converts radiance into apparent reflectance, which is equal to the ratio of irradiance M to incident irradiance E at the top of the atmosphere.

[0015]

[0016] Where: η is the apparent reflectance of the top of the atmosphere;

[0017] M and E are the exit and incident radii of the top of the atmosphere, respectively.

[0018] L represents the radiance of the satellite sensor at the top of the atmosphere;

[0019] D is the average distance from the Sun to the Earth. This value is related to the sensor's imaging time. The value of ESUN varies depending on the band and the sensor. Therefore, it is called the Band Mean Solar Irradiance (BMSI). β is the solar zenith angle.

[0020] 3) Geometric correction utilizes ground control points (GCPs) to correct geometric distortions in remote sensing images caused by various factors. Using one image as a reference f(x,y), the original image g(x',y') is corrected. The geometric distortion relationship between the two images is assumed to be...

[0021]

[0022] Typically, k1(x,y) and k2(x,y) can be approximated by polynomials.

[0023] By using the coordinates of several points on the reference image, the unknown parameters are solved, and then the correction coordinates of the corresponding pixels are calculated from the original image. At the same time, the pixel grayscale value is assigned to the corresponding pixel, thus completing the geometric correction of the original image.

[0024] 4) Image fusion is the process of fusing low-resolution images with high-resolution images to obtain a new image. Image fusion adopts an area-based image fusion method. The area-based fusion method usually uses a fixed-size window to filter the high-resolution image, and the filtered pixel value is used as a measure of the intensity of detail information at that point.

[0025] S2. Using the cropping function of GIS software, high-resolution images within a 2km radius outside the transmission line are obtained. The cropped high-resolution image before construction is PC, and the cropped high-resolution image during construction is BC.

[0026] S3. Based on the spectral, shape, and texture features of demolished houses in high-resolution images, and combined with the actual house shapes, establish an interpretation library of demolished houses; use the Canny edge detection algorithm to extract the texture features of houses at different stages.

[0027] 1) Use Gaussian kernels for convolutional noise reduction;

[0028] 2) The gradient and direction of the denoised image are calculated using the Sobel operator;

[0029]

[0030]

[0031] 3) After obtaining the gradient and direction, remove all points that are not on the boundary, gradually traverse the pixels, and determine whether the current pixel has the maximum value of the gradient in the same direction as the surrounding pixels. If it does, keep it; otherwise, set it to 0.

[0032] 4) Based on 3), the gradient is thresholded twice, T1 and T2, 0.4*T1 = T2. The grayscale of pixels with gradient values ​​less than T1 (generally, the value should ensure that the pixels before the largest gradient are retained) is set to 0, resulting in image 1. Then, the grayscale of pixels with gradient values ​​less than T2 is set to 0, resulting in image 2. Since the threshold of image 1 is higher, most of the noise is removed, but some useful edge information is also lost. The threshold of image 2 is lower, and more information is retained. We can use image 1 as the basis and image 2 as a supplement to connect the edges of the images.

[0033] If the edges of the features are regular rectangular shapes, then the houses have not been built ahead of schedule; if the edges of the features are numerous and fragmented, it is determined that the houses have been demolished but not restored; if the edges of the features and the surrounding environment cannot be obtained, then the houses have been demolished and restored.

[0034] S4. In accordance with the provisions of the soil and water conservation plan, verify the houses within a certain range outside the conductor of the transmission line, and use GIS and vectorization technology to obtain the location of the houses to be demolished for the transmission line.

[0035] The support vector machine algorithm of GIS is used for automatic building identification. Combined with vectorization technology, the automatic identification results are checked, and erroneous results are removed and corrected to improve the accuracy of building identification. Then, GIS tools are used to obtain the latitude and longitude coordinates of the buildings.

[0036] Support Vector Machine (SVM) is a tool that constructs an optimal decision hyperplane that maximizes the distance between the two classes of samples that are closest to the hyperplane on either side, thus providing good generalization ability for classification problems.

[0037] Let the linearly separable sample set be (x i ,y i ), i = 1, ..., n, x ∈ R d y∈{+1,-1} is the class symbol. The general form of the linear discriminant function in d-dimensional space is T(x)=w·x+b, and the equation of the classification line is w·x+b=0. Normalizing the discriminant function ensures that all samples from both classes satisfy |T(x)|≥1, even if the sample closest to the classification line has T(x)=1. In this case, the classification margin is equal to 2 / ||w||. Therefore, maximizing the margin is equivalent to minimizing ||w||. For the classification line to correctly classify all samples, it needs to satisfy y i (w·x i +b)-1≥0,i=1,…,n;

[0038] S5. Identify the demolition status of the extracted demolished houses;

[0039] S6. Perform statistical analysis on the interpretation results, archive and save them, and complete the identification of houses to be demolished for power transmission lines.

[0040] Furthermore, the specific steps for identifying the demolition status of the demolished houses in step S5 are as follows:

[0041] 1) The status of demolished houses is divided into three categories: houses not demolished, houses demolished but not restored, and houses demolished and restored, which are represented by η1, η2, and η3 respectively. The probability (prior probability) of each category is P(η1), P(η2), and P(η3) respectively.

[0042] 2) Suppose there is an unknown class sample Q, whose class conditional probabilities are P(Q|η1), P(Q|η2), and P(Q|η3), respectively;

[0043] 3) According to Bayes' theorem, the posterior probability of sample Q occurring is:

[0044]

[0045] 4) The posterior probability of sample Q is used as the discriminant function to determine the category of sample Q. The classification criterion is as follows:

[0046] if Then Q∈η i .

[0047] Furthermore, the specific steps of area-based image fusion in step S1 are as follows:

[0048] 1) In image d Δ In (x,y)(Δ=A,B), the energy (or variance) within the window region centered at point (x,y) is calculated as a measure of the intensity of detail information at that point, p. Δ (x,y);

[0049] 2) Calculate d A and d B The local, normalized cross-correlation coefficient Q between them AB (x,y);

[0050] 3) Different fusion methods are adopted based on the size of the cross-correlation coefficient: when Q AB When (x,y)≤α (α is generally set to 0.8), it indicates that the correlation between pixels in the source image is relatively low. Therefore, selecting pixels with large local variance as the fused pixels is more reasonable.

[0051]

[0052] When Q AB When (x,y)>α, it indicates a relatively high correlation between the coefficients, making a weighted average method more reasonable, i.e., d F (x,y)=w(x,y)d A (x,y)+[T(x,y)-w(x,y)d B (x,y)]

[0053] Where T(x,y) is the identity matrix, and the weight coefficients w(x,y) are determined by the following formula:

[0054]

[0055] The advantages and beneficial effects of this invention are as follows:

[0056] 1. The present invention is a method for identifying buildings demolished for power transmission lines based on high-resolution remote sensing, which breaks through the limitations of traditional manual monitoring, such as poor timeliness, low data accuracy, and high cost, and improves work efficiency.

[0057] 2. The present invention provides a method for identifying demolished houses along power transmission lines based on high-resolution remote sensing. This method is highly efficient and accurate in identifying demolished houses, ensuring accurate identification of houses within the boundary conductor range of power transmission line construction. This facilitates subsequent demolition work arrangements, ensures the safety of residents, and also ensures the smooth and safe progress of power transmission line construction.

[0058] 3. The present invention provides a new monitoring method for identifying buildings demolished along power transmission lines based on high-resolution remote sensing. Attached Figure Description

[0059] Figure 1 This is a flowchart of the present invention;

[0060] Figure 2 This is a schematic diagram illustrating the establishment of the demolition house interpretation library of the present invention. Detailed Implementation

[0061] The present invention will be further described in detail below through specific embodiments. The following embodiments are merely descriptive and not limiting, and should not be used to limit the scope of protection of the present invention.

[0062] like Figure 1 As shown, a method for identifying buildings demolished along power transmission lines based on high-resolution remote sensing is innovative in that the method comprises the following steps:

[0063] S1. Obtain high-resolution images P before construction and B during construction by using the location information of transmission line towers. Then, through radiometric calibration, atmospheric correction, geometric correction and image fusion processing, eliminate weather, sensor and terrain errors so that high-resolution images P and B can obtain projection coordinates.

[0064] 1) Radiometric calibration is the process of converting the original DN values ​​of an image, i.e., pixel brightness values, into radiometric values;

[0065] L = gain * DN + bias

[0066] Where: L is the radiance value;

[0067] gain is the image gain;

[0068] bias refers to image offset;

[0069] 2) Atmospheric correction converts radiance into apparent reflectance, which is equal to the ratio of irradiance M to incident irradiance E at the top of the atmosphere.

[0070]

[0071] Where: η is the apparent reflectance of the top of the atmosphere;

[0072] M and E are the exit and incident radii of the top of the atmosphere, respectively.

[0073] L represents the radiance of the satellite sensor at the top of the atmosphere;

[0074] D is the average distance from the Sun to the Earth. This value is related to the sensor's imaging time. The value of ESUN varies depending on the band and the sensor. Therefore, it is called the Band Mean Solar Irradiance (BMSI). β is the solar zenith angle.

[0075] 3) Geometric correction utilizes ground control points (GCPs) to correct geometric distortions in remote sensing images caused by various factors. Using one image as a reference f(x,y), the original image g(x',y') is corrected. The geometric distortion relationship between the two images is assumed to be...

[0076]

[0077] Typically, k1(x,y) and k2(x,y) can be approximated by polynomials.

[0078] By using the coordinates of several points on the reference image, the unknown parameters are solved, and then the correction coordinates of the corresponding pixels are calculated from the original image. At the same time, the pixel grayscale value is assigned to the corresponding pixel, thus completing the geometric correction of the original image.

[0079] 4) Image fusion is the process of fusing low-resolution images with high-resolution images to obtain a new image. Image fusion adopts an area-based image fusion method. The area-based fusion method usually uses a fixed-size window to filter the high-resolution image, and the filtered pixel value is used as a measure of the intensity of detail information at that point.

[0080] The specific steps of area-based image fusion are as follows:

[0081] 1) In image d Δ In (x,y)(Δ=A,B), the energy (or variance) within the window region centered at point (x,y) is calculated as a measure of the intensity of detail information at that point, p. Δ (x,y);

[0082] 2) Calculate d A and d B The local, normalized cross-correlation coefficient Q between them AB (x,y);

[0083] 3) Different fusion methods are adopted based on the size of the cross-correlation coefficient: when Q AB When (x,y)≤α (α is generally set to 0.8), it indicates that the correlation between pixels in the source image is relatively low. Therefore, selecting pixels with large local variance as the fused pixels is more reasonable.

[0084]

[0085] When Q AB When (x,y)>α, it indicates a relatively high correlation between the coefficients, making a weighted average method more reasonable, i.e., d F (x,y)=w(x,y)d A (x,y)+[T(x,y)-w(x,y)d B (x,y)]

[0086] Where T(x,y) is the identity matrix, and the weight coefficients w(x,y) are determined by the following formula:

[0087]

[0088] S2. Using the cropping function of GIS software, high-resolution images within a 2km radius outside the transmission line are obtained. The cropped high-resolution image before construction is PC, and the cropped high-resolution image during construction is BC.

[0089] S3. Based on the spectral, shape, and texture features of demolished houses in high-resolution images, and combined with the actual house shapes, establish an interpretation library of demolished houses; use the Canny edge detection algorithm to extract the texture features of houses at different stages.

[0090] 1) Use Gaussian kernels for convolutional noise reduction;

[0091] 2) The gradient and direction of the denoised image are calculated using the Sobel operator;

[0092]

[0093]

[0094] 3) After obtaining the gradient and direction, remove all points that are not on the boundary, gradually traverse the pixels, and determine whether the current pixel has the maximum value of the gradient in the same direction as the surrounding pixels. If it does, keep it; otherwise, set it to 0.

[0095] 4) Based on 3), the gradient is thresholded twice, T1 and T2, 0.4*T1 = T2. The grayscale of pixels with gradient values ​​less than T1 (generally, the value should ensure that the pixels before the largest gradient are retained) is set to 0, resulting in image 1. Then, the grayscale of pixels with gradient values ​​less than T2 is set to 0, resulting in image 2. Since the threshold of image 1 is higher, most of the noise is removed, but some useful edge information is also lost. The threshold of image 2 is lower, and more information is retained. We can use image 1 as the basis and image 2 as a supplement to connect the edges of the images.

[0096] If the edges of the features are regular rectangular shapes, then the houses have not been built ahead of schedule; if the edges of the features are numerous and fragmented, it is determined that the houses have been demolished but not restored; if the edges of the features and the surrounding environment cannot be obtained, then the houses have been demolished and restored.

[0097] S4. In accordance with the provisions of the soil and water conservation plan, verify the houses within a certain range outside the conductor of the transmission line, and use GIS and vectorization technology to obtain the location of the houses to be demolished for the transmission line.

[0098] The support vector machine algorithm of GIS is used for automatic building identification. Combined with vectorization technology, the automatic identification results are checked, and erroneous results are removed and corrected to improve the accuracy of building identification. Then, GIS tools are used to obtain the latitude and longitude coordinates of the buildings.

[0099] Support Vector Machine (SVM) is a tool that constructs an optimal decision hyperplane that maximizes the distance between the two classes of samples that are closest to the hyperplane on either side, thereby providing good generalization ability for classification problems.

[0100] Let the linearly separable sample set be (x i ,y i ), i = 1, ..., n, x ∈ R d y∈{+1,-1} is the class symbol. The general form of the linear discriminant function in d-dimensional space is T(x)=w·x+b, and the equation of the classification line is w·x+b=0. Normalizing the discriminant function ensures that all samples from both classes satisfy |T(x)|≥1, even if the sample closest to the classification line has T(x)=1. In this case, the classification margin is equal to 2 / ||w||. Therefore, maximizing the margin is equivalent to minimizing ||w||. For the classification line to correctly classify all samples, it needs to satisfy y i (w·x i +b)-1≥0,i=1,…,n;

[0101] S5. Identify the demolition status of the extracted demolished houses;

[0102] like Figure 2 As shown, (a) is a schematic diagram of a house that has not been demolished η1; (b) is a schematic diagram of a house that has been demolished but not restored η2; and (c) is a schematic diagram of a house that has been demolished and restored η3.

[0103] The specific steps for identifying the demolition status of houses under demolition are as follows:

[0104] 1) The status of demolished houses is divided into three categories: houses not demolished, houses demolished but not restored, and houses demolished and restored, which are represented by η1, η2, and η3 respectively. The probability (prior probability) of each category is P(η1), P(η2), and P(η3) respectively.

[0105] 2) Suppose there are samples Q of unknown class, with class-conditional probabilities P(Q|η1), P(Q|η2), and P(Q|η3):

[0106] 3) According to Bayes' theorem, the posterior probability of sample Q occurring is:

[0107]

[0108] 4) The posterior probability of sample Q is used as the discriminant function to determine the category of sample Q. The classification criterion is as follows:

[0109] if Then Q∈η i .

[0110] S6. Perform statistical analysis on the interpretation results, archive and save them, and complete the identification of houses to be demolished for power transmission lines.

[0111] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for identifying buildings demolished along power transmission lines based on high-resolution remote sensing, characterized in that: The steps of the method are as follows: S1. Obtain high-resolution images P before construction and B during construction by using the location information of transmission line towers. Then, through radiometric calibration, atmospheric correction, geometric correction and image fusion processing, eliminate weather, sensor and terrain errors so that high-resolution images P and B can obtain projection coordinates. 1) Radiometric calibration is the process of converting the original DN values ​​of an image, i.e., pixel brightness values, into radiometric values; L = gain * DN + bias; Where: L is the radiance value; gain is the image gain; bias refers to image offset; 2) Atmospheric correction converts radiance into apparent reflectance, which is equal to the ratio of irradiance M to incident irradiance E at the top of the atmosphere. ； in: The apparent reflectance at the top of the atmosphere; M and E are the exit and incident radii of the top of the atmosphere, respectively. L represents the radiance of the satellite sensor at the top of the atmosphere; D is the average distance from the Sun to the Earth. This value is related to the sensor's imaging time, and the ESUN value varies depending on the band and sensor. Therefore, it is called the Band Mean Solar Irradiance (BMSI). The solar zenith angle; 3) Geometric correction utilizes ground control points (GCPs) to correct geometric distortions in remote sensing images caused by various factors. Using one image as a reference f(x,y), the original image g(x',y') is corrected. The geometric distortion relationship between the two images is assumed to be... ； generally It can be approximated by a polynomial, then ; By using the coordinates of several points on the reference image, the unknown parameters are solved, and then the correction coordinates of the corresponding pixels are calculated from the original image. At the same time, the pixel grayscale value is assigned to the corresponding pixel, thus completing the geometric correction of the original image. 4) Image fusion is the process of fusing low-resolution images with high-resolution images to obtain new images. Image fusion adopts an area-based image fusion method. The area-based fusion method usually uses a fixed-size window to filter the high-resolution image, and the filtered pixel value is used as a measure of the intensity of detail information at that point. S2. Using the cropping function of GIS software, high-resolution images within a 2km radius outside the transmission line are obtained. The cropped high-resolution image before construction is PC, and the cropped high-resolution image during construction is BC. S3. Based on the spectral, shape, and texture features of demolished houses in high-resolution images, and combined with the actual house shapes, establish an interpretation library of demolished houses; use the Canny edge detection algorithm to extract the texture features of houses at different stages. 1) Use Gaussian kernels for convolutional noise reduction; 2) The gradient and direction of the denoised image are calculated using the Sobel operator; ； In the formula, g x Represents the vertical gradient kernel, g y Represents the horizontal gradient kernel, E g Represents the gradient magnitude, A δ Represents the gradient direction; 3) After obtaining the gradient and direction, remove all points that are not on the boundary, gradually traverse the pixels, and determine whether the current pixel has the maximum value of the gradient in the same direction as the surrounding pixels. If it does, keep it; otherwise, set it to 0. 4) Based on 3), the gradient is thresholded twice, T1 and T2, 0.4*T1=T2. The grayscale of pixels with gradient values ​​less than T1 is set to 0 to obtain image 1. Then the grayscale of pixels with gradient values ​​less than T2 is set to 0 to obtain image 2. Since the threshold of image 1 is higher, most of the noise is removed, but some useful edge information is also lost. The threshold of image 2 is lower and more information is retained. The edges of the images are connected by using image 1 as the base and image 2 as a supplement. If the edges of the features are regular rectangular shapes, then the houses have not been built ahead of schedule; if the edges of the features are numerous and fragmented, it is determined that the houses have been demolished but not restored; if the edges of the features and the surrounding environment cannot be obtained, then the houses have been demolished and restored. S4. In accordance with the provisions of the soil and water conservation plan, verify the houses within a certain range outside the conductor of the transmission line, and use GIS and vectorization technology to obtain the location of the houses to be demolished for the transmission line. The support vector machine algorithm of GIS is used for automatic building identification. Combined with vectorization technology, the automatic identification results are checked, and erroneous results are removed and corrected to improve the accuracy of building identification. Then, GIS tools are used to obtain the latitude and longitude coordinates of the buildings. Support Vector Machine (SVM) is a tool that constructs an optimal decision hyperplane that maximizes the distance between the two classes of samples that are closest to the hyperplane on either side, thereby providing good generalization ability for classification problems. Let the linearly separable sample set be It is a category symbol, and the general form of the linear discriminant function in d-dimensional space is: The classification line equation is The discriminant function is normalized so that all samples from both classes satisfy the following conditions: Even the sample closest to the classification surface At this time, the classification interval is equal to Therefore, maximizing the interval is equivalent to maximizing the interval. To minimize the requirement that the classification line correctly classifies all samples, it must satisfy the following condition: ; S5. Identify the demolition status of the extracted demolished houses; S6. Perform statistical analysis on the interpretation results, archive and save them, and complete the identification of houses to be demolished for power transmission lines.

2. The method for identifying demolished buildings along power transmission lines based on high-resolution remote sensing according to claim 1, characterized in that: The specific steps of S5 are as follows: 1) The status of houses slated for demolition is divided into three categories: houses not yet demolished, houses demolished but not yet restored, and houses demolished and restored. These are respectively represented by... This indicates that the prior probabilities of each category are respectively... ; 2) Suppose there is an unknown class sample Q, whose class conditional probabilities are respectively ; 3) According to Bayes' theorem, the posterior probability of sample Q occurring is: ； 4) The posterior probability of sample Q is used as the discriminant function to determine the category of sample Q. The classification criterion is as follows: if ,but .

3. The method for identifying demolished buildings along power transmission lines based on high-resolution remote sensing according to claim 1, characterized in that: The specific steps of area-based image fusion in step S1 are as follows: 1) In the image In the middle, calculation is based on The energy within the window area surrounding a point is used as a measure of the intensity of detailed information at that point. ; 2) Calculation and Local, normalized cross-correlation coefficients between ; 3) Different fusion methods are adopted based on the size of the cross-correlation coefficient: when hour, A value of 0.8 indicates that the correlation between pixels in the source image is relatively low, making it reasonable to select pixels with large local variance as the fused pixels. ； when When the correlation between the coefficients is high, a weighted average method is more reasonable. ; in For the identity matrix, the weight coefficients Determined by the following formula: 。

Images