Unmanned aerial vehicle remote sensing surveying and mapping image processing method

Abstract

The invention discloses an unmanned aerial vehicle remote sensing surveying and mapping image processing method, and relates to the technical field of image processing, and the method specifically comprises the steps: S1, carrying out the image data collection of a target region, and obtaining a to-be-processed surveying and mapping image set; s2, obtaining an initial gray level image set from the to-be-processed surveying and mapping image set, and denoising the initial gray level image set through median filtering introducing a window weight factor to process a denoised gray level image; s3, performing calculation processing on the de-noised grayscale image through inverse grayscale transformation and a color channel reduction algorithm to obtain a color enhanced surveying and mapping image; s4, performing deblurring processing on the color-enhanced surveying and mapping image, calculating a Dew gray scale main value region representative, a fixed segmentation threshold value and an ambient light interference amount based on the de-noised gray scale image and the color-enhanced surveying and mapping image, and performing calculation through a dynamic segmentation threshold value to obtain a deblurred surveying and mapping image; and S5, performing registration and splicing processing based on the deblurred remote sensing surveying and mapping image to obtain a panoramic registration image.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for processing images obtained by remote sensing of unmanned aerial vehicles. Background Art

[0002] In recent years, with the rapid development of sensor technology, communication technology and automatic control technology, the performance of UAVs has been continuously improved, and they can now undertake a large amount of remote sensing mapping work. Compared with manual mapping and satellite mapping, UAVs have outstanding advantages such as low cost, high speed and flexible data acquisition. With the increasing application of UAV remote sensing mapping, the optimization and development of UAV remote sensing mapping image processing is a promising research direction.

[0003] At present, the Chinese invention patent application with application number CN202411472333.4 discloses a method for processing remote sensing mapping images taken by drones, which includes: S1, collecting remote sensing mapping images taken by drones, generating a step function for each row of the remote sensing mapping images, and determining the corresponding change curve; S2, using the change curve to generate a sharpening kernel; S3, using the sharpening kernel to enhance the pixel points of the remote sensing mapping image. The present invention can enhance the edges and contours in the remote sensing mapping image so that the enhanced remote sensing mapping image looks clearer, which is convenient for users to further analyze and process the image. However, the application sharpens the image contours through the kernel function, and lacks the enhancement processing of the image color direction. Summary of the invention

[0004] The technical problem solved by the present invention is that the conventional method processes all images uniformly when removing noise, and does not adopt adaptive adjustment. Meanwhile, when enhancing remote sensing images, the sharpening process is mainly performed on the edges and contours, and the enhancement process on the color direction is lacking.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: A method for processing images obtained by remote sensing of unmanned aerial vehicles, comprising: Step S1: collecting image data of the target area to obtain a mapping image set to be processed; Step S2: obtaining an initial grayscale image set from the mapping image set to be processed, and denoising the initial grayscale image set by median filtering with a window weight factor; Step S3: calculating and processing the denoised grayscale image by inverse grayscale transformation and color channel restoration algorithm to obtain a color enhanced mapping image; Step S4: Deblurring the color enhanced surveying image, calculating the representative of the main grayscale value area of ​​the German and Austrian images, the fixed segmentation threshold and the amount of ambient light interference based on the denoised grayscale image and the color enhanced surveying image, and obtaining the deblurred surveying image through dynamic segmentation threshold calculation; Step S5: performing registration and stitching processing based on the deblurred remote sensing mapping image to obtain a panoramic registered image.

[0006] As a preferred solution of the method for processing UAV remote sensing mapping images described in the present invention, the UAV is controlled to fly based on a planned inspection route, and image data of the target mapping area is collected through the remote sensing optical sensor on the UAV platform to obtain a set of mapping images to be processed.

[0007] As a preferred solution of the processing method of a UAV remote sensing mapping image described in the present invention, wherein: grayscale mapping is performed on each image in the mapping image set to be processed to obtain an initial grayscale image set, the initial grayscale image set is partitioned by a sliding window of preset size and step length to obtain window sub-images, the window grayscale median corresponding to each window sub-image is obtained by sorting calculation, and the window weight factor is calculated based on the window grayscale median corresponding to each window sub-image.

[0008] As a preferred solution of the method for processing UAV remote sensing mapping images described in the present invention, the processing logic for calculating the window weight factor includes: The grayscale average value of all pixels in each window sub-image is obtained by an averaging calculation. The grayscale average value corresponding to each window sub-image and the window grayscale median are respectively subtracted and the absolute value is taken to obtain the pixel discrete data. The discrete data corresponding to each window sub-image is averaged twice to obtain the window discrete data. The window weight factor is obtained based on the pixel discrete data and the window discrete data. The corresponding sliding window is weighted based on the window weight factor to obtain the denoised grayscale image. The calculation expression is: ; ; ; Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and G(x,y) represents the grayscale value of the pixel point at the coordinate (x,y). represents the grayscale average value of the window sub-image, DAV(x,y) represents the discrete data of the pixel points, AAV(x,y) represents the discrete data of the window, and N represents the total number of pixels of the window sub-image. represents the window weight factor, c represents the window weight factor constant, and max{·} represents the maximum value function.

[0009] As a preferred solution of the processing method of the UAV remote sensing mapping image described in the present invention, the denoised grayscale image is subjected to inverse grayscale transformation to obtain the denoised original color mapping image, and the denoised original color mapping image is computationally processed by an improved color channel restoration algorithm to obtain a color enhanced mapping image.

[0010] As a preferred solution of the method for processing UAV remote sensing mapping images described in the present invention, the processing logic of the improved color channel restoration algorithm includes: Based on the preset first color depth value, the denoised original color mapping image is subjected to color quantization reduction processing to obtain a low-bit color depth image, the low-bit color depth image is subjected to color enhancement calculation to obtain a transition-enhanced image, and based on the preset second color depth value, the transition-enhanced image is subjected to color quantization enhancement processing to obtain a color-enhanced mapping image, and the calculation expression is: ; ; Among them, x represents the horizontal coordinate of the pixel, y represents the vertical coordinate of the pixel, PT(x,y) represents the transition enhanced image, PI(x,y) represents the low-bit color depth image, e represents the natural constant, MF(x,y) represents the color enhancement function, Represents a mapping constant, represents the gain constant, BC(x,y) represents the color band in the low-bit color depth image, and M represents the total number of color bands in the low-bit color depth image.

[0011] As a preferred solution of the method for processing a UAV remote sensing mapping image described in the present invention, wherein: the color enhanced mapping image is deblurred to obtain a deblurred mapping image, and the processing logic includes: Compare the R channel, G channel and B channel values ​​of each pixel in the color enhanced mapping image and take the minimum value to obtain the dark channel value, traverse all the pixels in the color enhanced mapping image to obtain the dark channel mapping image, sort the pixel values ​​in the dark channel mapping image from high to low, select the largest U% pixel value in the dark channel mapping image based on the preset ambient light ratio coefficient U% and sum them to obtain the ambient light interference amount; A grayscale histogram is calculated based on the denoised grayscale image, and a grayscale probability distribution is obtained by linearly mapping the grayscale histogram through the Min-Max normalization algorithm. The mutual correlation coefficient between the grayscale probability distribution and the standard normal distribution function is calculated, and the grayscale probability distribution is threshold-screened through a preset mutual correlation parameter. The grayscale values ​​in the grayscale probability distribution that are greater than or equal to the preset mutual correlation parameter are selected as the grayscale main value area, and the grayscale main value area is median-processed to obtain the grayscale main value area representative; The grayscale histogram is calculated by Otsu's algorithm to obtain a fixed segmentation threshold, and the dynamic segmentation threshold is calculated based on the grayscale main value area representative, the fixed segmentation threshold and the ambient light interference amount; Its calculation expression is: ; Among them, DST represents the dynamic segmentation threshold, max(·) represents the maximum value function, ALI represents the ambient light interference, PGR represents the grayscale main value region representative, and BST represents the fixed segmentation threshold.

[0012] As a preferred solution of the processing method of the UAV remote sensing mapping image described in the present invention, wherein: the denoised grayscale image is segmented based on the dynamic segmentation threshold to obtain a first fuzzy area image and a first non-fuzzy area image, and the dark channel mapping image is subjected to mask segmentation processing based on the pixel coordinates of the first fuzzy area image and the first non-fuzzy area image to obtain a second fuzzy area image and a second non-fuzzy area image; A deblurring dynamic coefficient is calculated based on the first fuzzy area image and the dynamic segmentation threshold, and the deblurring dynamic coefficient and the second fuzzy area image are multiplied to obtain a fuzzy area enhanced image; Based on the blurred area enhanced image and the second non-blurred area image, an image stitching process is performed to obtain a deblurred dark channel image, and based on the atmospheric light scattering model and the deblurred dark channel image, a restoration process is performed to obtain a deblurred remote sensing mapping image; The calculation expression of the defuzzification dynamic coefficient is: ; Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and DEV(x,y) represents the deblurring dynamic coefficient. represents the limiting constant, FGR(x,y) represents the grayscale value of the pixel in the first fuzzy area image, and DST represents the dynamic segmentation threshold.

[0013] As a preferred solution of the method for processing UAV remote sensing mapping images described in the present invention, the reference image and the image to be registered are selected based on the deblurred remote sensing mapping image, and the reference image and the image to be registered are registered and spliced ​​to obtain a panoramic registered image.

[0014] As a preferred solution of the processing method of the UAV remote sensing mapping image described in the present invention, the processing logic of performing registration processing on the image to be registered to obtain the panoramic registration image includes: The first corner point feature vector group and the second corner point feature vector group are obtained by calculating the reference image and the image to be registered respectively through the Harris matrix; Performing FFT calculation on the reference image and the image to be registered by fast Fourier transform to obtain a first phase angle data set and a second phase angle data set, extracting frequency domain phase data from the first phase angle data set and connecting them in series with the corresponding vectors in the first corner point feature vector group to obtain a first descriptor vector set, extracting frequency domain phase data from the second phase angle data set and connecting them in series with the corresponding vectors in the second corner point feature vector group to obtain a second descriptor vector set; Perform a registration determination based on the first descriptor vector set and the second descriptor vector set, calculate the Euclidean distance value between the elements in the first descriptor vector set and the second descriptor vector set, and determine that the feature points are pre-paired when the Euclidean distance is greater than or equal to a preset first registration threshold; Based on the pre-pairing of feature points, a secondary registration judgment is performed, and the number of pixel points that form the pre-pairing of feature points in the image to be registered is counted through a registration sliding window of size Q*Q. When the number of pixel points that form the pre-pairing of feature points in the image to be registered is greater than or equal to a preset second registration threshold, the feature points are retained to obtain the final pairing; Based on the final alignment of the feature points of the reference image and the image to be registered, the registration calculation is performed to the panoramic registration image through the affine transformation model.

[0015] The beneficial effects of the present invention are as follows: the introduction of window weight factors can be adaptively adjusted according to the grayscale distribution of pixels in the window, which is beneficial to improving the pertinence and accuracy of denoising. An improved color channel restoration algorithm is established by using color depth transformation, and color quantization and reduction processing is actively performed to form a color band in the image, which is beneficial to highlighting the color distribution information of the image and enhancing the color expression and image resolution of the remote sensing image. The dynamic segmentation threshold is calculated by combining the dark channel and grayscale histogram, and the traditional atmospheric light deblurring model is optimized. The feature descriptor is formed by Harris matrix and FFT, and a two-stage threshold determination is performed, which is beneficial to accurately align and obtain a panoramic image. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the basic flow of a method for processing UAV remote sensing mapping images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0018] Reference Figure 1 , as an embodiment of the present invention, provides a method for processing images obtained by remote sensing of unmanned aerial vehicles, comprising: Step S1: collecting image data of the target area to obtain a mapping image set to be processed; Step S2: obtaining an initial grayscale image set from the mapping image set to be processed, and denoising the initial grayscale image set by median filtering with a window weight factor; Step S3: calculating and processing the denoised grayscale image by inverse grayscale transformation and color channel restoration algorithm to obtain a color enhanced mapping image; Step S4: Deblurring the color enhanced surveying image, calculating the representative of the main grayscale value area of ​​the German and Austrian images, the fixed segmentation threshold and the amount of ambient light interference based on the denoised grayscale image and the color enhanced surveying image, and obtaining the deblurred surveying image through dynamic segmentation threshold calculation; Step S5: performing registration and stitching processing based on the deblurred remote sensing mapping image to obtain a panoramic registered image.

[0019] In this embodiment, the UAV is controlled to fly based on the planned inspection route, and the remote sensing optical sensor on the UAV platform collects image data of the target mapping area to obtain a mapping image set to be processed.

[0020] In this embodiment, grayscale mapping is performed on each image in the mapped image set to be processed to obtain an initial grayscale image set, the initial grayscale image set is partitioned by a sliding window of preset size and step length to obtain window sub-images, the window grayscale median corresponding to each window sub-image is obtained by sorting calculation, and the window weight factor is calculated based on the window grayscale median corresponding to each window sub-image.

[0021] In this embodiment, the processing logic for calculating the window weight factor includes: The grayscale average value of all pixels in each window sub-image is obtained by an averaging calculation. The grayscale average value corresponding to each window sub-image and the window grayscale median are respectively subtracted and the absolute value is taken to obtain the pixel discrete data. The discrete data corresponding to each window sub-image is averaged twice to obtain the window discrete data. The window weight factor is obtained based on the pixel discrete data and the window discrete data. The corresponding sliding window is weighted based on the window weight factor to obtain the denoised grayscale image. The calculation expression is: ; ; ; Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and G(x,y) represents the grayscale value of the pixel point at the coordinate (x,y). represents the grayscale average value of the window sub-image, DAV(x,y) represents the discrete data of the pixel points, AAV(x,y) represents the discrete data of the window, and N represents the total number of pixels of the window sub-image. represents the window weight factor, c represents the window weight factor constant, and max{·} represents the maximum value function.

[0022] Among them, the collected remote sensing images have system interference such as electromagnetic noise. Grayscale processing is conducive to single-channel calculation, improving calculation efficiency, and providing grayscale images for subsequent deblurring calculations. The innovative introduction of window weight factors can be adaptively adjusted according to the grayscale distribution of pixels in the window, which is conducive to improving the pertinence and accuracy of denoising.

[0023] In this embodiment, the denoised grayscale image is subjected to inverse grayscale transformation to obtain a denoised original color mapping image, and the denoised original color mapping image is subjected to computational processing by an improved color channel restoration algorithm to obtain a color enhanced mapping image.

[0024] In this embodiment, the processing logic of the improved color channel restoration algorithm includes: Based on the preset first color depth value, the denoised original color mapping image is subjected to color quantization reduction processing to obtain a low-bit color depth image, the low-bit color depth image is subjected to color enhancement calculation to obtain a transition-enhanced image, and based on the preset second color depth value, the transition-enhanced image is subjected to color quantization enhancement processing to obtain a color-enhanced mapping image, and the calculation expression is: ; ; Among them, x represents the horizontal coordinate of the pixel, y represents the vertical coordinate of the pixel, PT(x,y) represents the transition enhanced image, PI(x,y) represents the low-bit color depth image, e represents the natural constant, MF(x,y) represents the color enhancement function, Represents a mapping constant, represents the gain constant, BC(x,y) represents the color band in the low-bit color depth image, and M represents the total number of color bands in the low-bit color depth image.

[0025] Among them, actively performing color quantization and reduction processing to form color bands in the image is beneficial to highlighting the color distribution information of the image, enhancing the color contrast, and enhancing the image recognition of remote sensing mapping.

[0026] In this embodiment, the color enhanced mapping image is deblurred to obtain a deblurred mapping image, and the processing logic includes: Compare the R channel, G channel and B channel values ​​of each pixel in the color enhanced mapping image and take the minimum value to obtain the dark channel value, traverse all the pixels in the color enhanced mapping image to obtain the dark channel mapping image, sort the pixel values ​​in the dark channel mapping image from high to low, select the largest U% pixel value in the dark channel mapping image based on the preset ambient light ratio coefficient U% and sum them to obtain the ambient light interference amount; A grayscale histogram is calculated based on the denoised grayscale image, and a grayscale probability distribution is obtained by linearly mapping the grayscale histogram through the Min-Max normalization algorithm. The mutual correlation coefficient between the grayscale probability distribution and the standard normal distribution function is calculated, and the grayscale probability distribution is threshold-screened through a preset mutual correlation parameter. The grayscale values ​​in the grayscale probability distribution that are greater than or equal to the preset mutual correlation parameter are selected as the grayscale main value area, and the grayscale main value area is median-processed to obtain the grayscale main value area representative; The grayscale histogram is calculated by Otsu's algorithm to obtain a fixed segmentation threshold, and the dynamic segmentation threshold is calculated based on the grayscale main value area representative, the fixed segmentation threshold and the ambient light interference amount; Its calculation expression is: ; Among them, DST represents the dynamic segmentation threshold, max(·) represents the maximum value function, ALI represents the ambient light interference, PGR represents the grayscale main value region representative, and BST represents the fixed segmentation threshold.

[0027] In this embodiment, the denoised grayscale image is segmented based on the dynamic segmentation threshold to obtain a first blurred area image and a first non-blurred area image, and the dark channel mapping image is subjected to mask segmentation processing based on the pixel coordinates of the first blurred area image and the first non-blurred area image to obtain a second blurred area image and a second non-blurred area image; A deblurring dynamic coefficient is calculated based on the first fuzzy area image and the dynamic segmentation threshold, and the deblurring dynamic coefficient and the second fuzzy area image are multiplied to obtain a fuzzy area enhanced image; Based on the blurred area enhanced image and the second non-blurred area image, an image stitching process is performed to obtain a deblurred dark channel image, and based on the atmospheric light scattering model and the deblurred dark channel image, a restoration process is performed to obtain a deblurred remote sensing mapping image; The calculation expression of the defuzzification dynamic coefficient is: ; Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and DEV(x,y) represents the deblurring dynamic coefficient. represents the limiting constant, FGR(x,y) represents the grayscale value of the pixel in the first fuzzy area image, and DST represents the dynamic segmentation threshold.

[0028] In this embodiment, a reference image and an image to be registered are selected based on the deblurred remote sensing mapping image, and the reference image and the image to be registered are registered and spliced ​​to obtain a panoramic registered image.

[0029] Among them, the target area of ​​remote sensing mapping is often large, and the single-frame image area collected by drones is small, unlike satellites. There is overlap between the deblurred remote sensing mapping images, and multiple images need to be merged to obtain a panoramic image of the target area. Registration is helpful to obtain an accurate panoramic large image. In this example, L non-overlapping deblurred remote sensing mapping images are used as reference images, and the remaining (total number - L) overlapping deblurred remote sensing mapping images are used as images to be registered.

[0030] In this embodiment, the processing logic of performing registration processing on the image to be registered to obtain the panoramic registered image includes: The first corner point feature vector group and the second corner point feature vector group are obtained by calculating the reference image and the image to be registered respectively through the Harris matrix; Performing FFT calculation on the reference image and the image to be registered by fast Fourier transform to obtain a first phase angle data set and a second phase angle data set, extracting frequency domain phase data from the first phase angle data set and connecting them in series with the corresponding vectors in the first corner point feature vector group to obtain a first descriptor vector set, extracting frequency domain phase data from the second phase angle data set and connecting them in series with the corresponding vectors in the second corner point feature vector group to obtain a second descriptor vector set; Perform a registration determination based on the first descriptor vector set and the second descriptor vector set, calculate the Euclidean distance value between the elements in the first descriptor vector set and the second descriptor vector set, and determine that the feature points are pre-paired when the Euclidean distance is greater than or equal to a preset first registration threshold; Based on the pre-pairing of feature points, a secondary registration judgment is performed, and the number of pixel points that form the pre-pairing of feature points in the image to be registered is counted through a registration sliding window of size Q*Q. When the number of pixel points that form the pre-pairing of feature points in the image to be registered is greater than or equal to a preset second registration threshold, the feature points are retained to obtain the final pairing; Based on the final alignment of the feature points of the reference image and the image to be registered, the registration calculation is performed to the panoramic registration image through the affine transformation model.

[0031] Among them, the Harris matrix and fast Fourier transform are combined to construct a descriptor vector set to accurately extract image features. The first registration judgment uses Euclidean distance to screen out possible feature points for pre-pairing. The second registration judgment counts the number of pixels through a sliding window. The two-stage screening is conducive to improving the accuracy of registration.

[0032] It should be understood by those skilled in the art that the embodiments of the present invention may provide methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium may be based on any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for processing UAV remote sensing mapping images, characterized in that: include: Step S1: collecting image data of the target area to obtain a mapping image set to be processed; Step S2: obtaining an initial grayscale image set from the mapping image set to be processed, and denoising the initial grayscale image set by median filtering with a window weight factor; Step S3: calculating and processing the denoised grayscale image by inverse grayscale transformation and color channel restoration algorithm to obtain a color enhanced mapping image; Step S4: Deblurring the color enhanced surveying image, calculating the representative of the main grayscale value area of ​​the German and Austrian images, the fixed segmentation threshold and the amount of ambient light interference based on the denoised grayscale image and the color enhanced surveying image, and obtaining the deblurred surveying image through dynamic segmentation threshold calculation; Step S5: performing registration and stitching processing based on the deblurred remote sensing mapping image to obtain a panoramic registered image.

2. The method for processing UAV remote sensing images as claimed in claim 1, characterized in that: The UAV is controlled to fly based on the planned inspection route, and the remote sensing optical sensor on the UAV platform collects image data of the target mapping area to obtain the mapping image set to be processed.

3. The method for processing UAV remote sensing imaging images as claimed in claim 1, characterized in that: Grayscale mapping is performed on each image in the mapped image set to be processed to obtain an initial grayscale image set, and the initial grayscale image set is partitioned by a sliding window of preset size and step length to obtain window sub-images. The window grayscale median corresponding to each window sub-image is obtained by sorting calculation, and the window weight factor is calculated based on the window grayscale median corresponding to each window sub-image.

4. The method for processing UAV remote sensing imaging images as claimed in claim 3, characterized in that: The processing logic for calculating the window weight factor includes: The grayscale average value of all pixels in each window sub-image is obtained by an averaging calculation. The grayscale average value corresponding to each window sub-image and the window grayscale median are respectively subtracted and the absolute value is taken to obtain the pixel discrete data. The discrete data corresponding to each window sub-image is averaged twice to obtain the window discrete data. The window weight factor is obtained based on the pixel discrete data and the window discrete data. The corresponding sliding window is weighted based on the window weight factor to obtain the denoised grayscale image. The calculation expression is: ； ； ； Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and G(x,y) represents the grayscale value of the pixel point at the coordinate (x,y). represents the grayscale average value of the window sub-image, DAV(x,y) represents the discrete data of the pixel points, AAV(x,y) represents the discrete data of the window, and N represents the total number of pixels of the window sub-image. represents the window weight factor, c represents the window weight factor constant, and max{·} represents the maximum value function.

5. The method for processing UAV remote sensing imaging images as claimed in claim 1, characterized in that: The denoised grayscale image is inversely transformed to obtain a denoised original color mapping image, and the denoised original color mapping image is computationally processed using an improved color channel restoration algorithm to obtain a color enhanced mapping image.

6. The method for processing UAV remote sensing imaging images as claimed in claim 5, characterized in that: The processing logic of the improved color channel restoration algorithm includes: Based on the preset first color depth value, the denoised original color mapping image is subjected to color quantization reduction processing to obtain a low-bit color depth image, the low-bit color depth image is subjected to color enhancement calculation to obtain a transition-enhanced image, and based on the preset second color depth value, the transition-enhanced image is subjected to color quantization enhancement processing to obtain a color-enhanced mapping image, and the calculation expression is: ； ； Among them, x represents the horizontal coordinate of the pixel, y represents the vertical coordinate of the pixel, PT(x,y) represents the transition enhanced image, PI(x,y) represents the low-bit color depth image, e represents the natural constant, MF(x,y) represents the color enhancement function, Represents a mapping constant, represents the gain constant, BC(x,y) represents the color band in the low-bit color depth image, and M represents the total number of color bands in the low-bit color depth image.

7. The method for processing UAV remote sensing imaging images as claimed in claim 1, characterized in that: The color enhanced mapping image is deblurred to obtain a deblurred mapping image. The processing logic includes: Compare the R channel, G channel and B channel values ​​of each pixel in the color enhanced mapping image and take the minimum value to obtain the dark channel value, traverse all the pixels in the color enhanced mapping image to obtain the dark channel mapping image, sort the pixel values ​​in the dark channel mapping image from high to low, select the largest U% pixel value in the dark channel mapping image based on the preset ambient light ratio coefficient U% and sum them to obtain the ambient light interference amount; A grayscale histogram is calculated based on the denoised grayscale image, and a grayscale probability distribution is obtained by linearly mapping the grayscale histogram through the Min-Max normalization algorithm. The mutual correlation coefficient between the grayscale probability distribution and the standard normal distribution function is calculated, and the grayscale probability distribution is threshold-screened through a preset mutual correlation parameter. The grayscale values ​​in the grayscale probability distribution that are greater than or equal to the preset mutual correlation parameter are selected as the grayscale main value area, and the grayscale main value area is median-processed to obtain the grayscale main value area representative; The grayscale histogram is calculated by Otsu's algorithm to obtain a fixed segmentation threshold, and the dynamic segmentation threshold is calculated based on the grayscale main value area representative, the fixed segmentation threshold and the ambient light interference amount; Its calculation expression is: ； Among them, DST represents the dynamic segmentation threshold, max(·) represents the maximum value function, ALI represents the ambient light interference, PGR represents the grayscale main value region representative, and BST represents the fixed segmentation threshold.

8. The method for processing UAV remote sensing imaging images as claimed in claim 7, characterized in that: The denoised grayscale image is segmented based on a dynamic segmentation threshold to obtain a first blurred area image and a first non-blurred area image, and the dark channel mapping image is subjected to mask segmentation processing based on the pixel coordinates of the first blurred area image and the first non-blurred area image to obtain a second blurred area image and a second non-blurred area image; A deblurring dynamic coefficient is calculated based on the first fuzzy area image and the dynamic segmentation threshold, and the deblurring dynamic coefficient and the second fuzzy area image are multiplied to obtain a fuzzy area enhanced image; Based on the blurred area enhanced image and the second non-blurred area image, an image stitching process is performed to obtain a deblurred dark channel image, and based on the atmospheric light scattering model and the deblurred dark channel image, a restoration process is performed to obtain a deblurred remote sensing mapping image; The calculation expression of the defuzzification dynamic coefficient is: ； Among them, x represents the horizontal coordinate of the pixel point, y represents the vertical coordinate of the pixel point, and DEV(x,y) represents the deblurring dynamic coefficient. represents the limiting constant, FGR(x,y) represents the grayscale value of the pixel in the first fuzzy area image, and DST represents the dynamic segmentation threshold.

9. The method for processing UAV remote sensing imaging images as claimed in claim 1, characterized in that: Based on the deblurred remote sensing mapping image, a reference image and an image to be registered are selected, and the reference image and the image to be registered are registered and spliced ​​to obtain a panoramic registered image.

10. The method for processing UAV remote sensing imaging images according to claim 1, characterized in that: The processing logic of performing registration processing on the image to be registered to obtain the panoramic registration image includes: The first corner point feature vector group and the second corner point feature vector group are obtained by calculating the reference image and the image to be registered respectively through the Harris matrix; Performing FFT calculation on the reference image and the image to be registered by fast Fourier transform to obtain a first phase angle data set and a second phase angle data set, extracting frequency domain phase data from the first phase angle data set and connecting them in series with the corresponding vectors in the first corner point feature vector group to obtain a first descriptor vector set, extracting frequency domain phase data from the second phase angle data set and connecting them in series with the corresponding vectors in the second corner point feature vector group to obtain a second descriptor vector set; Perform a registration determination based on the first descriptor vector set and the second descriptor vector set, calculate the Euclidean distance value between the elements in the first descriptor vector set and the second descriptor vector set, and determine that the feature points are pre-paired when the Euclidean distance is greater than or equal to a preset first registration threshold; Based on the pre-pairing of feature points, a secondary registration judgment is performed, and the number of pixel points that form the pre-pairing of feature points in the image to be registered is counted through a registration sliding window of size Q*Q. When the number of pixel points that form the pre-pairing of feature points in the image to be registered is greater than or equal to a preset second registration threshold, the feature points are retained to obtain the final pairing; Based on the final alignment of the feature points of the reference image and the image to be registered, the registration calculation is performed to the panoramic registration image through the affine transformation model.

Images