A method and system for processing urban building images based on drone inspection

Abstract

The present invention belongs to the field of image processing technology, and specifically relates to a method and system for processing urban building images based on drone inspection, the method comprising: decomposing urban building images to obtain detail sub-images of different sizes in multiple directions, obtaining the detail expression degree of the target pixel in each direction according to the expression degree of detail features of different sizes and the pixel values ​​of the corresponding pixels in the detail sub-images of different sizes in each direction of the target pixel in the urban building image, obtaining a comprehensive detail vector by integrating the detail expression degrees in multiple directions, and obtaining a Gaussian filter of the target pixel according to the comprehensive detail vector; filtering each pixel according to the Gaussian filter, subtracting the urban building image from the obtained background image, and obtaining an urban building image with uniform light. The urban building image with uniform light obtained by the present invention has both good clarity and good uniform light effect.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a method and system for processing urban building images based on drone inspection. Background Art

[0002] In the process of drone-based inspection and collection of urban building images, due to the influence of factors such as lighting conditions, shooting angles, and atmospheric conditions, uneven brightness distribution will appear in the obtained urban building images, increasing the difficulty of subsequent accurate analysis of urban buildings.

[0003] In the related technology, for example, the Chinese patent application document with application publication number CN115588087A discloses a UAV surveying and mapping data processing system and method for modeling, including an image acquisition module, an image processing module, an elevation accuracy detection module, an image homogenization processing module, an aerial triangulation processing module, a joint adjustment module, a texture mapping module and a film punctate module. The image acquisition module uses the UAV oblique photography to acquire images and transmits the acquired image data to the image processing module; the application document uses the oblique photography technology in the image acquisition module to add a method of simultaneously exposing oblique perspective lenses at four angles, and simultaneously obtains downward and oblique perspective images, which can access more characteristic surface information. In the process of oblique photography measurement, relative positioning and absolute positioning are required to directly measure the height of buildings and other objects; the image contains rich real environment information.

[0004] In the image uniformity processing module in the related art, The light uniformity algorithm performs light uniformity processing on images with uneven illumination to obtain images with uniform illumination. The homogenization algorithm performs low-pass filtering on the original image through a Gaussian filter to obtain a background image with simulated brightness distribution, and then subtracts the original image from the background image to obtain an image with uniform illumination. Among them, the choice of Gaussian filter determines the homogenization effect and clarity of the image with uniform illumination, which affects the accurate analysis of urban buildings. Summary of the invention

[0005] To solve the above The selection of Gaussian filter in the homogenization algorithm determines the homogenization effect and clarity of the uniformly illuminated image, thereby affecting the technical problem of accurate analysis of urban buildings. The present invention provides solutions in the following aspects.

[0006] In the first aspect, the present invention provides a method for processing urban building images based on drone inspection, including: inspecting and collecting urban building images through drones; decomposing urban building images through wavelet transform to obtain detail sub-images of different sizes in multiple directions; calculating the degree of expression of detail features of different sizes according to the ratio of different sizes to the size of the urban building image; taking any pixel point in the urban building image as the target pixel point; according to the degree of expression of detail features of different sizes, weighted summing up the pixel values ​​of the corresponding pixel points in the detail sub-images of different sizes in each direction of the target pixel point to obtain the degree of detail expression of the target pixel point in each direction; taking the degree of detail expression of the target pixel point in each direction as the modulus of the detail vector of the target pixel point in each direction, calculating the sum of the detail vectors of the target pixel point in all directions as the comprehensive detail vector of the target pixel point; obtaining a two-dimensional Gaussian model based on the comprehensive detail vector. , The directions of the two dimensions are and , the standard deviations of the two dimensions are equal to and ,in, is the direction of the comprehensive detail vector, , are the lengths of the projections of the comprehensive detail vector in the first and second dimensions respectively; according to the two-dimensional Gaussian model A Gaussian filter of a target pixel is obtained; each pixel in the urban building image is filtered according to the Gaussian filter of each pixel to obtain a background image; and the urban building image is subtracted from the background image to obtain an urban building image with uniform light.

[0007] The present invention obtains the degree of detail expression of the target pixel in each direction according to the degree of expression of detail features at different sizes and the pixel values ​​of the corresponding pixel points in the detail sub-images of different sizes in each direction of the target pixel point in the urban building image, obtains the comprehensive detail vector by comprehensively analyzing the detail expression degrees in multiple directions, and obtains the Gaussian filter of the target pixel point according to the comprehensive detail vector. For the pixel point with a greater degree of detail expression, the larger the standard deviation of the two-dimensional Gaussian model, the fewer detail features are contained in the local area of ​​the pixel point in the final background image, thereby enhancing the clarity of the obtained urban building image with uniform light exposure; for the pixel point with a smaller degree of detail expression, the smaller the standard deviation of the two-dimensional Gaussian model, the better the effect of simulating the brightness distribution of the local area of ​​the pixel point in the final background image, thereby enhancing the uniform light effect of the obtained urban building image with uniform light exposure; therefore, the present invention ensures that the obtained urban building image with uniform light exposure has both good clarity and good uniform light effect, thereby ensuring the accuracy of subsequent analysis of urban buildings.

[0008] Preferably, the process of decomposing the urban building image by wavelet transform to obtain detail sub-images of different sizes in multiple directions comprises: when decomposing the urban building image by wavelet transform, a total of decomposition, then the number of types of all sizes is equal to ;in, , No. The size of the detail sub-images in each direction obtained by the decomposition is ,Will Recorded as Size, then For the The length of the size, For the The width of the size, and , , is the length of the urban building image, is the width of the urban building image; wherein the multiple directions include a vertical direction, a horizontal direction, and a diagonal direction, and the vertical direction, the horizontal direction, and the diagonal direction are respectively recorded as the first Direction, Direction and direction.

[0009] The present invention utilizes the multi-resolution characteristics of wavelet transform to obtain detail sub-images of urban building images in multiple scales and directions, thereby analyzing the intensity of detail features of the local area of ​​each pixel point.

[0010] Preferably, the degree of expression of detail features of different sizes satisfies the expression: ; In the formula, For the The degree of detail characteristics expressed by size, is the number of decompositions, and the number of types of all sizes is equal to , is the natural exponential function, For the size, For the The length of the size, For the The width of the size, and , , is the size of the city building image, is the length of the urban building image, is the width of the city building image.

[0011] In the present invention, the pixel values ​​of the pixel points in the obtained detail sub-image can all express the detail features in the urban building image, and the pixel values ​​of the pixel points in the detail sub-images of different sizes have different expressions of the detail features in the urban building image. Therefore, the present invention calculates the expression of the detail features of different sizes based on the ratio of different sizes to the size of the urban building image.

[0012] Preferably, the detail expression degree of the target pixel in each direction satisfies the expression: ; In the formula, is the coordinate of the target pixel in the urban building image, , are the horizontal and vertical coordinates of the target pixel, respectively. The target pixel is The degree of detail in each direction, is the number of decompositions, and the number of types of all sizes is equal to , For the The degree of detail characteristics expressed by size, For the Direction The coordinates in the detail sub-graph of the size are The pixel value of the pixel point, is the ceiling function.

[0013] The present invention weights the pixel values ​​of the pixel points corresponding to the target pixel points in the detail sub-images of different sizes according to the expression degree of detail features of different sizes, so as to obtain the detail expression degree of the target pixel points in different directions.

[0014] Preferably, the Direction The coordinates in the detail sub-graph of the size are The pixel point is the target pixel point in Direction The corresponding pixel in the detail sub-image of the size.

[0015] Preferably, the two-dimensional Gaussian model Each dimension of has two parameters, the mean and the standard deviation, and the two-dimensional Gaussian model The mean of both dimensions of is 0.

[0016] Preferably, the Gaussian filter is essentially a matrix, wherein the number of rows of the Gaussian filter in the first dimension is , the number of rows in the second dimension is , , They are two-dimensional Gaussian models The standard deviation of the two dimensions of is the ceiling function, is the expansion factor.

[0017] In the present invention, for pixels with a greater degree of detail expression, a Gaussian filter with a larger size is set, so that the blurring degree of the pixel after filtering is greater, and the local area of ​​the pixel in the final background image contains fewer detail features, thereby enhancing the clarity of the obtained urban building image with uniform light; for pixels with a smaller degree of detail expression, a Gaussian filter with a smaller size is set, so that the effect of simulating the brightness distribution of the local area of ​​the pixel in the final background image is better, thereby enhancing the uniform light effect of the obtained urban building image with uniform light.

[0018] Preferably, the method of obtaining a uniformly illuminated image of urban buildings comprises: performing corresponding stretching and displaying on the image after subtraction processing, and finally obtaining a uniformly illuminated image of urban buildings; the stretching and displaying methods comprise 2% linear stretching, contrast parameter stretching and gradient stretching display.

[0019] Preferably, the method comprises: constructing a rectangular coordinate system with the pixel point at the upper left corner of the urban building image as the origin, the horizontal rightward direction of the origin as the positive direction of the horizontal axis, and the vertical downward direction of the origin as the positive direction of the vertical axis; then the value range of the horizontal coordinate of the pixel point in the urban building image is , the vertical axis value range is ,and is the length of the urban building image, is the width of the city building image.

[0020] In a second aspect, the present invention provides an urban building image processing system based on drone inspection, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned urban building image processing method based on drone inspection is implemented.

[0021] By adopting the above technical solution, the above-mentioned urban building image processing method based on drone inspection is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is made according to the memory and the processor for easy use.

[0022] The beneficial effects of the present invention are:

[0023] The present invention obtains the degree of detail expression of the target pixel in each direction according to the degree of expression of detail features at different sizes and the pixel values ​​of the corresponding pixel points in the detail sub-images of different sizes in each direction of the target pixel point in the urban building image, obtains the comprehensive detail vector by comprehensively analyzing the detail expression degrees in multiple directions, and obtains the Gaussian filter of the target pixel point according to the comprehensive detail vector. For the pixel point with a greater degree of detail expression, the larger the standard deviation of the two-dimensional Gaussian model, the fewer detail features are contained in the local area of ​​the pixel point in the final background image, thereby enhancing the clarity of the obtained urban building image with uniform light exposure; for the pixel point with a smaller degree of detail expression, the smaller the standard deviation of the two-dimensional Gaussian model, the better the effect of simulating the brightness distribution of the local area of ​​the pixel point in the final background image, thereby enhancing the uniform light effect of the obtained urban building image with uniform light exposure; therefore, the present invention ensures that the obtained urban building image with uniform light exposure has both good clarity and good uniform light effect, thereby ensuring the accuracy of subsequent analysis of urban buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flow chart schematically illustrating a method for processing urban building images based on drone inspection in the present invention;

[0026] Figure 2 is a schematic diagram schematically showing collected urban building images;

[0027] Figure 3 is a flowchart schematically showing the process of obtaining the Gaussian filter of each pixel in the urban building image in step S2;

[0028] Figure 4 is a schematic diagram schematically illustrating a two-dimensional Gaussian model in which the standard deviations of the first dimension and the second dimension are both 0.5;

[0029] Figure 5 It is a schematic diagram showing the Figure 4 Schematic diagram of the Gaussian filter obtained from the two-dimensional Gaussian model shown;

[0030] Figure 6 is a schematic diagram schematically illustrating a two-dimensional Gaussian model in which the standard deviation of the first dimension and the second dimension are both 0.25;

[0031] Figure 7 It is a schematic diagram showing the Figure 6Schematic diagram of the Gaussian filter obtained from the two-dimensional Gaussian model shown;

[0032] Figure 8 It is a schematic diagram showing Figure 2 A schematic diagram of a background image obtained after filtering the urban building image shown;

[0033] Fig. 9 It is schematically shown Figure 2 The city building images shown are similar to Fig. 9 Schematic diagram of an image of a city building with uniform illumination obtained after subtracting the background image shown. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0035] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0036] The embodiment of the present invention discloses a method for processing urban building images based on drone inspection, referring to Figure 1 , comprising steps S1 to S3:

[0037] In the process of drone-based inspection and collection of urban building images, due to the influence of factors such as lighting conditions, shooting angles, and atmospheric conditions, uneven brightness distribution will appear in the obtained urban building images, increasing the difficulty of subsequent accurate analysis of urban buildings.

[0038] In order to eliminate the uneven brightness distribution in the urban building images collected by drone inspection, it is necessary to perform light uniformity processing on the urban building images to obtain urban building images with uniform light.

[0039] The light uniformity algorithm is a common method for uniforming urban building images with uneven brightness distribution. This method enhances the detail contrast of the image and suppresses the abnormal brightness changes of the image by enhancing high-frequency information and suppressing low-frequency information, thereby achieving uniform brightness distribution. Specifically, The uniform light algorithm performs low-pass filtering on the original image through a Gaussian filter to obtain a background image with simulated brightness distribution, and then subtracts the original image from the background image to obtain an image with uniform light.

[0040] Among them, the selection of Gaussian filter determines the uniform light effect and clarity of the image with uniform light, thus affecting the accurate analysis of urban buildings. The Gaussian filter is obtained by a two-dimensional Gaussian model: the larger the standard deviation of the two-dimensional Gaussian model, the larger the size of the Gaussian filter, the more pixels participating in the simulation of brightness distribution but the more discrete the distribution, the greater the blurring degree of the filtered pixels, and the final background image contains fewer detail features, but the simulation effect of the brightness distribution in the urban building image is not good. The uniform light image of the urban building obtained by subtracting the original image from the background image retains a large number of detail features, but the uniform light effect is poor; the smaller the standard deviation of the two-dimensional Gaussian model, the smaller the size of the Gaussian filter, the fewer pixels participating in the simulation of brightness distribution but the more concentrated the distribution, and the final background image can better simulate the brightness distribution in the urban building image, but also contains more detail features, resulting in the loss of a large number of detail features in the uniform light image of the urban building obtained by subtracting the original image from the background image, that is, the clarity of the uniform light image of the urban building obtained is not high.

[0041] Therefore, the present invention decomposes the collected urban building images to obtain detail sub-images of different sizes in multiple directions, obtains the detail expression degree of the target pixel in each direction according to the expression degree of detail features of different sizes and the pixel values ​​of the corresponding pixel points in the detail sub-images of different sizes in each direction of the target pixel in the urban building image, takes the detail expression degree of each direction as the modulus length of the detail vector in each direction, takes the sum of the detail vectors in all directions as the comprehensive detail vector, and obtains the Gaussian filter of the target pixel according to the comprehensive detail vector; performs filtering processing on each pixel according to the Gaussian filter, subtracts the urban building image from the obtained background image, and obtains a uniformly illuminated urban building image; thereby ensuring that the obtained uniformly illuminated urban building image has both good clarity and good uniform lighting effect, thereby ensuring the accuracy of subsequent analysis of urban buildings.

[0042] S1. Use drones to inspect and collect images of urban buildings.

[0043] Specifically, when conducting inspections by drones, the drones carry multiple cameras on a flying platform to collect images of urban buildings from multiple angles, thereby obtaining complete and accurate information about urban buildings.

[0044] For the collected urban building images, a rectangular coordinate system is constructed with the pixel point in the upper left corner of the urban building image as the origin, the horizontal rightward direction of the origin as the positive direction of the horizontal axis, and the vertical downward direction of the origin as the positive direction of the vertical axis; the value range of the horizontal coordinate of the pixel point in the urban building image is , the vertical axis value range is ,and is the length of the urban building image, is the width of the city building image.

[0045] For example, Figure 2 The following is a schematic diagram of the collected urban building images.

[0046] S2. Obtain the Gaussian filter of each pixel in the urban building image.

[0047] Specifically, by analyzing the urban building image, a Gaussian filter of each pixel in the urban building image is obtained. The flowchart of obtaining the Gaussian filter of each pixel in the urban building image in step S2 is as follows: Figure 3 , including steps S201 to S205, specifically:

[0048] S201, decomposing the urban building image by wavelet transform to obtain detail sub-images of different sizes in all directions.

[0049] It should be noted that the present invention utilizes the multi-resolution characteristics of wavelet transform to obtain detail sub-images of urban building images at multiple scales and directions, thereby analyzing the intensity of detail features of the local area of ​​each pixel.

[0050] Specifically, through Wavelet transform is used to decompose the urban building image. Decompose once, and obtain detailed sub-images of different sizes in all directions, and the number of types of all sizes is equal to In this embodiment, the number of decompositions In other embodiments, the number of decompositions can be set according to actual application scenarios and requirements; and Wavelet transform is a well-known technology and will not be described in detail here.

[0051] The directions include vertical, horizontal and diagonal directions. Therefore, the obtained detail sub-images in each direction include detail sub-images in the vertical direction, detail sub-images in the horizontal direction and detail sub-images in the diagonal direction. The vertical direction, horizontal direction and diagonal direction are respectively recorded as Direction, Direction and direction.

[0052] Among them, The size of the detail sub-images in each direction obtained by the decomposition is ,Will Recorded as Size, then For the The length of the dimension, and , For the The width of the size, and , is the length of the urban building image, is the width of the city building image.

[0053] Among them, the pixel points in the detail sub-image have pixel values, and a higher pixel value generally indicates the existence of an edge or texture, while a lower pixel value indicates that the edge or texture is weak or non-existent. Therefore, the pixel value of the pixel point reflects the detail features of the local area of ​​the pixel point in the detail sub-image, and the detail features include edge features and texture features, and the larger the pixel value, the greater the intensity of the detail features of the local area of ​​the pixel point.

[0054] Among them, the detail sub-image in the vertical direction contains the low-frequency components of the urban building image in the horizontal direction and the high-frequency components in the vertical direction. Therefore, the detail sub-image in the vertical direction contains the detail features of the urban building image in the vertical direction; the detail sub-image in the horizontal direction contains the high-frequency components of the urban building image in the horizontal direction and the low-frequency components in the vertical direction. Therefore, the detail sub-image in the horizontal direction contains the detail features of the urban building image in the horizontal direction; the detail sub-image in the diagonal direction contains the high-frequency components of the urban building image in the horizontal direction and the high-frequency components in the vertical direction. Therefore, the detail sub-image in the diagonal direction contains the detail features of the urban building image in the diagonal direction.

[0055] S202, calculating the expression degree of detail features at different sizes.

[0056] It should be noted that the pixel values ​​of the pixels in the detail sub-image can all express the detail features in the urban building image, and the pixel values ​​of the pixels in detail sub-images of different sizes have different expressions of the detail features in the urban building image: the pixel values ​​of the pixels in the larger detail sub-image have smaller expressions of the detail features in the urban building image, and the pixel values ​​of the pixels in the smaller detail sub-image have greater expressions of the detail features in the urban building image; therefore, the expressions of the detail features of different sizes can be calculated based on the ratios of different sizes to the size of the urban building image.

[0057] Specifically, according to the ratio of different sizes to the size of the urban building image, the expression of detail features of different sizes is calculated. The expression of size to detail features satisfies the expression:

[0058] ;

[0059] In the formula, For the The degree of detail characteristics expressed by size, is the number of decompositions, and the number of types of all sizes is equal to , is the natural exponential function, For the size, For the The length of the dimension, and , For the The width of the size, and , is the size of the city building image, is the length of the urban building image, is the width of the city building image.

[0060] S203, calculating the detail expression levels of the pixels in the urban building image in different directions.

[0061] Specifically, for the urban building image with coordinates The pixel point in each direction The coordinates of the corresponding pixel points in the detail sub-image of size are ,in, is the ceiling function; where , are the horizontal and vertical coordinates of the pixel points respectively.

[0062] Furthermore, according to the degree of detail feature expression of different sizes, the coordinates of the urban building image are The pixel values ​​of the corresponding pixels in the detail sub-images of different sizes in each direction are weighted to obtain the detail expression degree of the pixel in each direction; then the coordinates in the urban building image are The pixel point is The degree of detail in each direction satisfies the expression:

[0063] ;

[0064] In the formula, The coordinates of the city building image are The pixel point is The degree of detail in each direction, is the number of decompositions, and the number of types of all sizes is equal to , For the The degree of detail characteristics expressed by size, For the Direction The coordinates in the detail sub-graph of the size are The pixel value of the pixel point, The coordinates of the city building image are The pixel points in each direction The coordinates of the corresponding pixel points in the detail sub-image of the size, is the ceiling function.

[0065] It should be noted that the expression of detail features by different sizes The pixel values ​​of the corresponding pixels in the detail sub-images of different sizes Weighted, the coordinates of the urban building image are obtained as The pixel point is The degree of detail in each direction , where the pixel value of a pixel reflects the detail features of the local area of ​​the pixel in the detail sub-image, and the larger the pixel value, the greater the intensity of the detail features of the local area of ​​the pixel. The greater the intensity of the detail features of the local area of ​​the pixel, the greater the coordinates in the urban building image are The pixel point is The degree of detail in each direction The bigger; Size’s ability to express detailed features The larger the value is, the coordinates in the urban building image are The pixel point is The degree of detail in each direction The bigger.

[0066] S204, synthesizing the detail expression levels of the pixels in the urban building image in all directions to obtain a comprehensive detail vector of the pixels.

[0067] Specifically, the coordinates of the urban building image are The pixel point is The degree of detail in each direction , as the pixel point in Detail vector in each direction The module length, at the same time, direction, as the pixel point in the Detail vector in each direction The direction of the city building image is obtained by The pixel point is Detail vector in each direction .

[0068] Further, the coordinates of the urban building image are calculated as The sum of the detail vectors of the pixel in all directions is used as the comprehensive detail vector of the pixel ; Among them, the result of summing up multiple vectors is still a vector, and the summation of multiple vectors is a well-known technology and will not be described in detail here.

[0069] It should be noted that, by integrating the detail vectors of the pixel in all directions, a comprehensive detail vector of the pixel is obtained, and the comprehensive detail vector comprehensively reflects the distribution of detail features of the local area of ​​the pixel.

[0070] S205 . Determine a Gaussian filter for the pixel point according to the comprehensive detail vector of the pixel point in the urban building image.

[0071] Specifically, for pixel points in an urban building image, a Gaussian filter of the pixel points is obtained based on a two-dimensional Gaussian model, wherein the two-dimensional Gaussian model has two dimensions, which are respectively the first dimension and the second dimension, and the directions of the two dimensions are perpendicular to each other; each dimension has two parameters, which are respectively the mean and the standard deviation.

[0072] Furthermore, for the urban building image with coordinates The pixel point, the comprehensive detail vector of the pixel point The direction of ;According to the comprehensive detail vector of the pixel point, obtain the two-dimensional Gaussian model .

[0073] Among them, the two-dimensional Gaussian model The directions of the two dimensions are and , and the direction and direction In perpendicular directions to each other.

[0074] It should be noted that in the Gaussian filter, the larger the weight along the two-dimensional direction, the greater the influence of the pixel value of the filtered pixel point on the pixel value of other pixels at the position with the larger weight, and the smaller the influence of the pixel value of other pixels at the position with the smaller weight. The comprehensive detail vector of the pixel point comprehensively reflects the distribution of the detail features of the local area of ​​the pixel point, so if the direction of the comprehensive detail vector of the pixel point is as well as As the direction of the two dimensions of the Gaussian filter, the pixel value of the filtered pixel retains more detail features, resulting in a large number of detail features lost in the uniformly illuminated urban building image obtained by subtracting the original image from the background image, that is, the clarity of the uniformly illuminated urban building image is not high; In summary, the present invention uses the direction of the comprehensive detail vector of the pixel point Plus And its vertical direction, as the two-dimensional direction of the Gaussian filter, can make the blurred degree of the filtered pixel points greater, and the final background image contains fewer detailed features, thereby enhancing the clarity of the obtained urban building image with uniform lighting.

[0075] Among them, the two-dimensional Gaussian model Each dimension of has two parameters, namely the mean and standard deviation, the two-dimensional Gaussian model The mean of both dimensions is 0 and the standard deviation is and , satisfying the expression:

[0076] ;

[0077] ;

[0078] In the formula, , are the comprehensive detail vectors of the pixel respectively The length of the projection in the first and second dimensions.

[0079] It should be noted that for pixels with greater detail, their comprehensive detail vector The longer the modulus of The longer the projection length in the first and second dimensions, the better the two-dimensional Gaussian model. The larger the standard deviation of each dimension, the larger the size of the Gaussian filter, the more pixels involved in the simulated brightness distribution, but the more discrete the distribution, the greater the blur of the filtered pixels, and the fewer detail features contained in the local area of ​​the pixel in the final background image, thereby enhancing the clarity of the uniformly illuminated urban building image; for pixels with smaller detail expression, their comprehensive detail vector The shorter the modulus of The shorter the projection length in the first and second dimensions, the smaller the two-dimensional Gaussian model. The smaller the standard deviation of each dimension, the smaller the size of the Gaussian filter, the fewer pixels participating in the simulated brightness distribution but the more concentrated the distribution, and the better the effect of simulating the brightness distribution of the local area of ​​the pixel in the final background image, thereby enhancing the uniform lighting effect of the uniformly illuminated urban building image.

[0080] Further, according to the two-dimensional Gaussian model Obtain the Gaussian filter of the target pixel. The essence of the Gaussian filter is a matrix, where the number of rows of the Gaussian filter in the first dimension is , the number of rows in the second dimension is , , They are two-dimensional Gaussian models The standard deviation of the two dimensions of is the ceiling function, is the expansion factor; in this embodiment, the expansion factor In other embodiments, the number of decompositions can be set according to actual application scenarios and requirements.

[0081] For example, Figure 4 As shown, it is a schematic diagram of a two-dimensional Gaussian model with the standard deviation of the first dimension and the second dimension both being 0.5; Figure 5 As shown, according to Figure 4 Schematic diagram of the Gaussian filter obtained from the two-dimensional Gaussian model shown in FIG. Figure 6 As shown, it is a schematic diagram of a two-dimensional Gaussian model with the standard deviation of the first dimension and the second dimension both being 0.25; Figure 7 As shown, according to Figure 6 Schematic diagram of the Gaussian filter obtained from the two-dimensional Gaussian model shown.

[0082] S3, according to the Gaussian filter of each pixel in the urban building image, The uniform light algorithm obtains images of urban buildings with uniform light.

[0083] Specifically, the Gaussian filter of each pixel in the urban building image is used to The uniform light algorithm obtains a uniformly illuminated urban building image, including: filtering each pixel in the urban building image according to a Gaussian filter of each pixel, obtaining a background image composed of filtered pixels, wherein the brightness distribution is simulated in the background image, subtracting the urban building image from the obtained background image, and stretching and displaying the subtracted image accordingly, finally obtaining a uniformly illuminated urban building image.

[0084] It should be noted that the present invention obtains the degree of detail expression of the target pixel in each direction according to the degree of expression of detail features at different sizes and the pixel values ​​of the corresponding pixel points in the detail sub-images of different sizes in each direction of the target pixel point in the urban building image, obtains a comprehensive detail vector by comprehensively analyzing the degree of detail expression in multiple directions, and obtains a Gaussian filter of the target pixel point according to the comprehensive detail vector. For pixels with greater detail expression, the greater the standard deviation of the two-dimensional Gaussian model, the fewer detail features are contained in the local area of ​​the pixel point in the final background image, thereby enhancing the clarity of the obtained uniformly illuminated urban building image; for pixels with smaller detail expression, the smaller the standard deviation of the two-dimensional Gaussian model, the better the effect of simulating brightness distribution in the local area of ​​the pixel point in the final background image, thereby enhancing the uniform lighting effect of the obtained uniformly illuminated urban building image; therefore, the present invention ensures that the obtained uniformly illuminated urban building image has both good clarity and good uniform lighting effect, thereby ensuring the accuracy of subsequent analysis of urban buildings.

[0085] For example, Figure 8 As shown, for Figure 2 Schematic diagram of the background image obtained after filtering the urban building image shown in FIG. Fig. 9 As shown, Figure 2 The city building images shown are similar to Fig. 9 Schematic diagram of an image of a city building with uniform illumination obtained after subtracting the background image shown.

[0086] Among them, commonly used stretching display methods include but are not limited to 2% linear stretching, contrast parameter stretching and gradient stretching display; The uniform light algorithm, 2% linear stretching, contrast parameter stretching and gradient stretching display are all well-known technologies and will not be described in detail here.

[0087] It should be noted that after the subtraction process, the overall contrast of the obtained subtraction processed image will be reduced, especially in the darker areas. In order to improve the overall contrast and maintain the consistency of the overall contrast of the subtraction processed image, highlight the details in the image, and maintain the clarity of the image, the subtraction processed image needs to be stretched and displayed.

[0088] An embodiment of the present invention also discloses an urban building image processing system based on drone inspection, including a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, an urban building image processing method based on drone inspection according to the present invention is implemented.

[0089] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0090] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0091] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for processing urban building images based on drone inspection, characterized in that: include: Inspect and collect images of urban buildings through drones; The urban building image is decomposed by wavelet transform to obtain detail sub-images of different sizes in multiple directions; According to the ratio of different sizes to the size of the urban building image, the degree of expression of detail features of different sizes is calculated; Any pixel in the urban building image is taken as the target pixel; according to the degree of detail feature expression of different sizes, the pixel values ​​of the corresponding pixels in the detail sub-images of different sizes in each direction of the target pixel are weighted summed to obtain the degree of detail expression of the target pixel in each direction, satisfying the expression: ； In the formula, is the coordinate of the target pixel in the urban building image, , are the horizontal and vertical coordinates of the target pixel, respectively. The target pixel is The degree of detail in each direction, is the number of decompositions, and the number of types of all sizes is equal to , For the The degree of detail characteristics expressed by size, For the Direction The coordinates in the detail sub-graph of the size are The pixel value of the pixel point, is the ceiling rounding function; The degree of detail of the target pixel in each direction is taken as the modulus of the detail vector of the target pixel in each direction, and the sum of the detail vectors of the target pixel in all directions is calculated as the comprehensive detail vector of the target pixel; according to the comprehensive detail vector, a two-dimensional Gaussian model is obtained. , The directions of the two dimensions are and , the standard deviations of the two dimensions are equal to and ,in, is the direction of the comprehensive detail vector, , are the lengths of the projections of the comprehensive detail vector in the first dimension and the second dimension respectively; according to the two-dimensional Gaussian model Obtain the Gaussian filter of the target pixel; Perform low-pass filtering on each pixel point in the urban building image according to the Gaussian filter of each pixel point to obtain a background image; subtract the urban building image from the background image to obtain an urban building image with uniform light, including: The image after the subtraction process is stretched and displayed accordingly, and finally an image of a city building with uniform light is obtained; the stretching display method includes 2% linear stretching, contrast parameter stretching and gradient stretching display.

2. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: Decomposing the urban building image by wavelet transform to obtain detail sub-images of different sizes in multiple directions includes: When decomposing the urban building image by wavelet transform, decomposition, then the number of types of all sizes is equal to ;in, , No. The size of the detail sub-images in each direction obtained by the decomposition is ,Will Recorded as Size, then For the The length of the size, For the The width of the size, and , , is the length of the urban building image, is the width of the urban building image; The multiple directions include a vertical direction, a horizontal direction, and a diagonal direction, and the vertical direction, the horizontal direction, and the diagonal direction are respectively recorded as Direction, Direction and direction.

3. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: The expression of detail features of different sizes satisfies the expression: ； In the formula, For the The degree of detail characteristics expressed by size, is the number of decompositions, and the number of types of all sizes is equal to , is the natural exponential function, For the size, For the The length of the size, For the The width of the size, and , , is the size of the city building image, is the length of the urban building image, is the width of the city building image.

4. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: The said Direction The coordinates in the detail sub-graph of the size are The pixel point is the target pixel point in Direction The corresponding pixel in the detail sub-image of the size.

5. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: The two-dimensional Gaussian model Each dimension of has two parameters, the mean and the standard deviation, and the two-dimensional Gaussian model The mean of both dimensions of is 0.

6. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: The essence of the Gaussian filter is a matrix, where the number of rows in the first dimension is , the number of rows in the second dimension is , , They are two-dimensional Gaussian models The standard deviation of the two dimensions, is the ceiling function, is the expansion factor.

7. The urban building image processing method based on drone inspection according to claim 1 is characterized in that: The method comprises: A rectangular coordinate system is constructed with the pixel point in the upper left corner of the urban building image as the origin, the horizontal direction to the right of the origin as the positive direction of the horizontal axis, and the vertical direction downward from the origin as the positive direction of the vertical axis; the value range of the horizontal coordinate of the pixel point in the urban building image is , the vertical axis value range is ,and is the length of the urban building image, is the width of the city building image.

8. An urban building image processing system based on drone inspection, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an urban building image processing method based on drone inspection according to any one of claims 1 to 7 is implemented.

Images