Unmanned aerial vehicle remote sensing image joint radiation correction method, device and equipment

Abstract

The invention relates to an unmanned aerial vehicle remote sensing image joint radiation correction method, device and equipment. According to the main technical scheme, the method comprises the following steps: obtaining ground feature spectral information, unmanned aerial vehicle high-altitude global image information and unmanned aerial vehicle low-altitude local image information, and carrying out relative radiation correction on the unmanned aerial vehicle low-altitude local image information based on the unmanned aerial vehicle high-altitude global image information to obtain relative radiation correction parameters; according to the relative radiation correction parameters and the unmanned aerial vehicle low-altitude local image information, unmanned aerial vehicle low-altitude global image information is generated, absolute radiation correction is carried out based on the ground feature spectral information and the unmanned aerial vehicle low-altitude global image information, the ground real reflectivity is obtained, and unmanned aerial vehicle remote sensing image joint radiation correction is completed. According to the method, the influence of weather, illumination, unmanned aerial vehicle attitude and the like can be eliminated, the radiation correction precision is improved, and more accurate quantitative application is realized.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and in particular to a method, apparatus and equipment for joint radiometric correction of UAV remote sensing images. Background Technology

[0002] Remote sensing is a way to observe and analyze objects on the Earth's surface in a remote and non-contact manner. It represents a qualitative change from traditional geoscientific methods. This new geographical approach has brought great convenience to human life and is becoming an irreplaceable tool in various applications such as agricultural production, ecological protection, transportation, and disaster prevention.

[0003] Traditional remote sensing uses artificial satellites or space shuttles to observe the Earth's surface and acquire data on ground features. However, this method has several drawbacks, including long satellite revisit cycles leading to short image timeliness, high costs due to the large amount of manpower and resources required, strict requirements for flight weather conditions, and the need for multiple processing steps for the acquired images. UAV remote sensing combines UAV technology with sensor technology, enabling the automation and specialization of ground information detection.

[0004] Current methods for radiometric correction of UAV remote sensing images utilize a large number of ground calibration points—natural or artificial features with uniform ground texture—to perform regression analysis with multiple low-altitude UAV images, obtaining sensor radiometric correction parameters. This method is simple and fast, and can suppress image distortion in the spectral dimension to some extent. However, during UAV flight, constantly changing weather conditions, lighting conditions, and the attitude of the UAV equipment can cause significant radiometric differences in the acquired images, making it difficult to achieve more accurate quantitative applications. Summary of the Invention

[0005] Based on this, this application provides a method, apparatus and equipment for joint radiometric correction of UAV remote sensing images to eliminate the influence of weather, lighting, UAV attitude and other factors, improve the accuracy of radiometric correction and enable more precise quantitative applications.

[0006] Firstly, a method for joint radiometric correction of UAV remote sensing images is provided, the method comprising: Acquire spectral information of ground features, global high-altitude imagery from UAVs, and local low-altitude imagery from UAVs; Based on the high-altitude global image information of the UAV, relative radiometric correction is performed on the low-altitude local image information of the UAV to obtain the relative radiometric correction parameters. Based on the relative radiometric correction parameters and the UAV low-altitude local image information, generate the UAV low-altitude global image information; Absolute radiometric correction is performed based on ground object spectral information and UAV low-altitude global image information to obtain the true ground reflectance, thus completing the joint radiometric correction of UAV remote sensing images.

[0007] According to one feasible method in an embodiment of this application, relative radiometric correction is performed on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information to obtain relative radiometric correction parameters, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, the UAV's high-altitude global imagery is cropped to obtain the high-altitude target area imagery. The high-altitude target area imagery is the same as the imagery containing the calibration blanket in the UAV's low-altitude local imagery. Using high-altitude target area images as reference images, histogram matching operations were performed on the low-altitude local image information of the UAV using remote sensing image processing software to obtain relative radiometric correction parameters.

[0008] According to one feasible method in an embodiment of this application, relative radiometric correction is performed on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information to obtain relative radiometric correction parameters, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, the initial matching feature points of the UAV's high-altitude global imagery and low-altitude local imagery are determined. Image matching points with pixel values ​​of maximum and minimum values ​​within a preset scale range are selected as candidate matching feature points; Determine the scale-invariant feature transformation feature similarity of candidate matching feature points, and select candidate matching feature points whose scale-invariant feature transformation feature similarity is greater than a preset threshold as matching feature points; The relative radiometric correction parameters are determined based on the matching feature points, the UAV's high-altitude global image information, and the UAV's low-altitude local image information.

[0009] According to one feasible method in an embodiment of this application, initial matching feature points of the UAV's high-altitude global imagery and low-altitude local imagery are determined based on the UAV's low-altitude local imagery, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, different Gaussian kernel standard deviations are used to perform continuous Gaussian blurring on the UAV's high-altitude global imagery and low-altitude local imagery to generate a series of images of the same size. Based on a series of images of the same size, the initial matching feature points of the UAV high-altitude global image and the UAV low-altitude local image are determined.

[0010] According to one feasible method in an embodiment of this application, relative radiometric correction is performed on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information to obtain relative radiometric correction parameters, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, the first and second location regions of the preset ground features in the UAV's high-altitude global imagery and low-altitude local imagery are determined respectively. Calculate the average pixel value for the first location region and the second location region respectively; The average pixel values ​​of the first and second location regions of each preset feature are linearly regressed using the least squares method to obtain the relative radiometric correction parameters.

[0011] According to one feasible method in an embodiment of this application, low-altitude global image information of a UAV is generated based on relative radiometric correction parameters and low-altitude local image information of the UAV, including: Based on the relative radiometric correction parameters and UAV low-altitude local image information, the pixel values ​​in each UAV low-altitude local image are subjected to relative radiometric correction to obtain the UAV low-altitude local corrected image. By stitching together locally corrected low-altitude images from the UAV, global low-altitude image information of the UAV can be obtained.

[0012] According to one achievable method in an embodiment of this application, the spectral information of ground features includes the spectral reflectance of ground features; based on the spectral information of ground features and the low-altitude global imagery information from a UAV, absolute radiometric correction is performed to obtain the ground reflectance, including: Based on the low-altitude global imagery information from the UAV, determine the calibration blanket image area in the low-altitude global imagery of the UAV. Calculate the average pixel value of the calibration blanket image area; Based on the average pixel value and spectral reflectance of the ground cover in the calibration blanket image area, the linear regression coefficients were determined by the least squares method. Absolute radiometric correction is performed on each pixel value in the UAV low-altitude global image based on the linear regression coefficients and preset formulas to obtain the true ground reflectance.

[0013] Secondly, a joint radiometric correction device for UAV remote sensing images is provided, the device comprising: The acquisition module is used to acquire spectral information of ground features, global high-altitude imagery information of UAVs, and local low-altitude imagery information of UAVs. The relative radiometric correction module is used to perform relative radiometric correction on the low-altitude local image information of the UAV based on the high-altitude global image information of the UAV, and obtain the relative radiometric correction parameters. The generation module is used to generate low-altitude global image information of the UAV based on the relative radiometric correction parameters and the low-altitude local image information of the UAV. The absolute radiometric correction module is used to perform absolute radiometric correction based on ground object spectral information and UAV low-altitude global image information to obtain the true ground reflectance and complete the joint radiometric correction of UAV remote sensing images.

[0014] Thirdly, a computer device is provided, comprising: At least one processor; and A memory that is communicatively connected to at least one processor; wherein, The memory stores computer instructions that can be executed by at least one processor to enable the at least one processor to perform the methods involved in the first aspect above.

[0015] Fourthly, a computer-readable storage medium is provided, having stored thereon computer instructions, characterized in that the computer instructions are used to cause a computer to perform the methods involved in the first aspect above.

[0016] According to the technical content provided in the embodiments of this application, by acquiring ground object spectral information, UAV high-altitude global image information, and UAV low-altitude local image information, relative radiometric correction is performed on the UAV low-altitude local image information based on the UAV high-altitude global image information to obtain relative radiometric correction parameters. Based on the relative radiometric correction parameters and the UAV low-altitude local image information, UAV low-altitude global image information is generated. Based on the ground object spectral information and the UAV low-altitude global image information, absolute radiometric correction is performed to obtain the true ground reflectance. This completes the joint radiometric correction of UAV remote sensing images, eliminating the influence of weather, illumination, UAV attitude, etc., improving the accuracy of radiometric correction, and enabling more precise quantitative applications. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for joint radiometric correction of UAV remote sensing images in one embodiment; Figure 2 This is an intermediate and result diagram of histogram matching in one embodiment; Figure 3 This is a structural block diagram of a UAV remote sensing image combined radiometric correction device in one embodiment; Figure 4 This is a schematic structural diagram of a computer device in one embodiment. Detailed Implementation

[0018] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the scope of the present application.

[0019] At the end of the 20th century, remote sensing, focusing on the dynamically changing global environment, adopted multi-angle and multi-platform remote sensing technology systems, continuously pursuing high spatial, temporal, and spectral resolution remote sensing applications. Remote sensing will become a rising star in the 21st century, not only achieving the integration of space-ground information services but also making significant contributions to sustainable social development, economic growth, and improved living standards.

[0020] Compared with traditional remote sensing methods, UAV remote sensing has advantages such as low risk, low cost, low environmental constraints, high acquisition speed, high reusability, high spatial resolution, and easy analysis and processing. It provides accurate data references for fields such as precision agriculture, land use, disaster emergency response, and high-precision mapping, demonstrating capabilities unmatched by traditional methods. Due to its strong operability and wide reach, UAV remote sensing has successfully enabled real-time remote sensing applications.

[0021] However, due to the long development history and high durability of satellite sensors, UAV data processing technology is still in its early stages compared to satellite remote sensing. Precision correction technologies, such as radiometric correction, urgently need systematic and comprehensive development, necessitating the establishment of corresponding data product standards and specifications to achieve large-scale, quantitative production and application. During UAV flight, constantly changing weather conditions, lighting conditions, and the attitude of the UAV equipment can cause significant radiometric differences in the acquired images, which is highly detrimental to the subsequent application of image data. Therefore, current UAV remote sensing image radiometric correction technology urgently needs improvement and refinement to overcome the constraints of radiometric differences among multiple remote sensing images, laying a solid foundation for more accurate quantitative applications.

[0022] To address the problems of existing technologies, embodiments of this application provide a method, apparatus, device, and computer storage medium for joint radiometric correction of UAV remote sensing images. The joint radiometric correction method for UAV remote sensing images provided in this application embodiment will be described first below. For example... Figure 1 As shown, the method may include the following steps: S110 acquires spectral information of ground features, global high-altitude imagery of UAVs, and local low-altitude imagery of UAVs.

[0023] Spectral information of ground features refers to the characteristics of various objects on the Earth's surface in reflecting, absorbing, and emitting electromagnetic waves (especially sunlight) of different wavelengths. Ground features can include vegetation, water bodies, soil, rocks, man-made structures, etc. Spectral information of ground features can include spectral reflectance, spectral feature location, spectral feature morphology, and spectral response range.

[0024] UAV high-altitude global imagery information refers to high-altitude global images covering a large area, acquired from an aerial perspective by sensors carried by the UAV, and the information contained therein. This may include the location information and pixel value of each pixel in the high-altitude global imagery, ground feature location information, etc.

[0025] Low-altitude local imagery information from drones refers to low-altitude local images acquired by drones at relatively low flight altitudes for specific, small targets or areas, and the information they contain. This may include the location information and pixel value of each pixel in the low-altitude local image, ground feature location information, etc.

[0026] The acquisition areas for ground feature spectral information, UAV high-altitude global imagery, and UAV low-altitude local imagery are defined based on user research needs. Ground calibration mats are deployed within the study area, ideally on open ground in the center. A UAV low-altitude flight path is planned to cover the study area, acquiring low-altitude flight path imagery, and UAV high-altitude imagery of the entire study area is also captured. Ground spectral information for the calibration mats and other ground features is obtained using an ASD spectrometer.

[0027] The ground-based spectral information acquired by the ASD spectrometer has a spectral resolution of 1 nm, while the UAV sensor used in the experiment only has five bands: blue, green, red, and red-edge bands with a spectral resolution of 16 nm within the center wavelength range, and near-infrared band with a spectral resolution of 26 nm within the center wavelength range. Therefore, the ASD-measured ground object spectral data needs to be resampled to ensure consistent resolution between the two. For the measured ground object points, the average of the center wavelength and reflectance values ​​within the five bands is extracted as the true reflectance value corresponding to the UAV's band. The resampled spectral curve shows almost no change in trend, but it transforms from a relatively smooth curve into a broken line segment, indicating a decrease in spectral resolution. However, this is helpful for subsequent band matching and regression operations.

[0028] For example, a certain brand's Phantom 4 multispectral drone includes one visible light lens and five multispectral lenses (blue, green, red, red-edge, and near-infrared) to acquire low-altitude drone imagery at an altitude of approximately 36-40 meters. During acquisition, the main flight path image overlap rate is approximately 80%, and the inter-flight overlap rate is approximately 60%. When acquiring high-altitude full-area imagery data, the flight altitude is controlled at approximately 450 meters. The drone's attitude is maintained stably throughout the flight.

[0029] Ground feature spectral information was collected using an ASD field spectrometer. The ASD spectrometer has a spectral range covering 350-2500 nm and a spectral resolution of up to 1 nm. Thirty-one ground feature locations were measured and marked on October 31st, and 16 ground feature locations were measured and marked on November 1st. This ground feature survey covered almost all types of ground features within the survey area, including various land types such as woodland, grassland, soil, sandy land, and dry and wet concrete, as well as common ground features such as water, bridge surfaces, rooftops, pebbles, and plastic sheeting.

[0030] S120 performs relative radiometric correction on the low-altitude local image information of the UAV based on the high-altitude global image information of the UAV, and obtains the relative radiometric correction parameters.

[0031] Relative radiometric correction methods include nonlinear relative radiometric correction methods and linear relative radiometric correction methods. Nonlinear relative radiometric correction methods include histogram matching methods, while linear relative radiometric correction methods include methods based on image feature points and methods based on the average value of ground features.

[0032] For histogram matching, the image information of five bands in each UAV low-altitude local image is first observed. Since the gray levels of vegetation and concrete are similar in the visible light band, the contrast between different ground features in the image is low and difficult to distinguish. Therefore, the near-infrared band is selected for histogram matching.

[0033] The high-altitude global imagery from the UAV is cropped to obtain areas that are relatively consistent with the calibration blanket in the low-altitude local imagery from the UAV. Histogram matching is then performed in ENVI, using the high-altitude global imagery as the histogram reference image to correct the low-altitude local imagery.

[0034] Linear relative radiometric correction assumes that the gray values ​​of the same feature in two images of the same area taken at different times or by different sensors have a linear relationship that can be derived from each other. Therefore, a linear expression can be used to relate and describe the two images. Let x represent the image to be corrected and y represent the original reference image. The relationship established between the reference image and the image to be corrected using the linear correction method can be expressed as: ; Where a and b represent gain and bias, respectively, and are relative radiometric correction parameters. Typically, the selected ground features can be targets with definite geographical significance, such as sample points whose reflectivity remains almost unchanged under different times or different sensors and whose spectral performance is stable, such as stable and homogeneous water bodies or concrete roads; or they can be points whose image properties are extremely similar over a short period of time or under the same sensor, such as ground features with the same geographical coordinates and no significant short-term changes.

[0035] For image feature point-based methods, the SIFT operator is used to extract and match image matching feature points from UAV high-altitude global images and UAV low-altitude local images. Specifically, multiple potential feature points are initially selected for each image. Then, a threshold is set based on the SIFT feature similarity of the potential feature points, and n mutually matching feature points are calculated. The corresponding coordinates of these n mutually matching feature points in the two images are then retrieved to obtain their corresponding DN values, and a least-squares relative radiometric correction regression operation is performed on them.

[0036] For methods based on regional averages of ground features, to prevent the data from being too singular and random at a few matching points, several identical ground features are delineated in both low-altitude local UAV imagery and high-altitude global UAV imagery. The average pixel value of each ground feature's image region is calculated, and the average DN values ​​of corresponding regions in the two images are linearly regressed using the least squares method. The resulting relative radiometric correction parameters are then used for relative radiometric correction based on the regional averages of the ground features, thus ensuring the accuracy and reliability of the linear regression method for relative radiometric correction of the images.

[0037] Based on the high-altitude global imagery information of the UAV, relative radiometric correction is performed on the low-altitude local imagery information of the UAV, so that the low-altitude local imagery of the UAV has a tone that is more consistent with the high-altitude global imagery of the UAV.

[0038] S130 generates global low-altitude imagery information of the UAV based on relative radiation correction parameters and local low-altitude imagery information of the UAV.

[0039] Based on the relative radiometric correction parameters and the UAV low-altitude local image information, relative radiometric correction is performed on each pixel in the UAV low-altitude local image to obtain the UAV low-altitude local image after relative radiometric correction.

[0040] By stitching together radiometrically corrected low-altitude local images from a UAV, a seamless wide-field-of-view image can be obtained. Furthermore, Pix4d software can be used to stitch together radiometrically corrected low-altitude local images from a UAV to obtain a global low-altitude image from a UAV containing geographic coordinates and its associated information.

[0041] S140, based on the spectral information of ground objects and the low-altitude global image information of UAVs, performs absolute radiometric correction to obtain the true ground reflectance.

[0042] Among them, the ground object spectral information can include the spectral reflectance of the ground object, and the UAV low-altitude global image information can include the location information and pixel value of each pixel in the low-altitude global image.

[0043] Based on the image area of ​​each calibration blanket in the low-altitude global image, a linear regression function is obtained between the average pixel value of the calibration blanket and the spectral reflectance of ground features. Based on the linear regression function, absolute radiometric correction is performed on each pixel in the low-altitude global image to obtain the true ground reflectance.

[0044] As can be seen, the embodiments of this application acquire ground object spectral information, UAV high-altitude global image information, and UAV low-altitude local image information. Based on the UAV high-altitude global image information, relative radiometric correction is performed on the UAV low-altitude local image information to obtain relative radiometric correction parameters. Based on the relative radiometric correction parameters and the UAV low-altitude local image information, UAV low-altitude global image information is generated. Based on the ground object spectral information and the UAV low-altitude global image information, absolute radiometric correction is performed to obtain ground reflectance. Ground reflectance is used to correct the UAV low-altitude image, eliminating the influence of weather, illumination, UAV attitude, etc., improving radiometric correction accuracy, and realizing more precise quantitative applications.

[0045] As one feasible approach, relative radiometric correction is performed on the low-altitude local imagery of the UAV based on the UAV's high-altitude global imagery information, resulting in relative radiometric correction parameters, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, the UAV's high-altitude global imagery is cropped to obtain the high-altitude target area imagery. The high-altitude target area imagery is the same as the imagery containing the calibration blanket in the UAV's low-altitude local imagery. Using high-altitude target area images as reference images, histogram matching operations were performed on the low-altitude local image information of the UAV using remote sensing image processing software to obtain relative radiometric correction parameters.

[0046] Based on visual interpretation, the image data of each low-altitude UAV image was observed in five bands. Since the gray levels of vegetation and concrete are similar in the visible light band, the contrast between different ground features in the image is low and difficult to distinguish. Therefore, the near-infrared band of the image was selected for histogram matching.

[0047] First, the UAV high-altitude global imagery is cropped to obtain a high-altitude target area imagery that overlaps with the UAV low-altitude local imagery. The overlapping area is the region in both the UAV high-altitude global imagery and the low-altitude local imagery where the calibration blanket corresponds. Based on the information from both the UAV high-altitude global imagery and the UAV low-altitude local imagery, the position of the calibration blanket in the UAV low-altitude local imagery is determined, and the corresponding region in the UAV high-altitude global imagery is then determined based on this position.

[0048] Histogram matching was performed in the remote sensing image processing software ENVI, using high-altitude target area images as histogram reference images to correct low-altitude local images of UAVs.

[0049] Figure 2 A diagram illustrating the intermediate and final results of histogram matching, such as... Figure 2 As shown, Figure 2 (a) and Figure 2 (c) Reference histograms of high-altitude target area imagery and low-altitude local UAV imagery, respectively. Figure 2 (b) and Figure 2 (d) shows the histogram changes of the high-altitude target area image and the low-altitude local image from the UAV after histogram equalization. It can be seen that after histogram equalization, the histograms of both images cover almost the entire grayscale range, and except for a few grayscale values ​​that are relatively prominent, the overall grayscale distribution is almost uniform. Therefore, the images have a wider grayscale dynamic range, higher contrast, and richer details.

[0050] Figure 2 (e) The result of histogram matching on the UAV low-altitude local imagery. It can be seen that the histogram after matching is significantly closer to the histogram of the high-altitude target area imagery compared to the result using only equalization. The histogram-matched UAV low-altitude local imagery also shows that the overall color tone of the histogram-matched UAV low-altitude local imagery is more consistent with the color tone of the high-altitude target area imagery. Histogram matching achieves both consistent color tone across multiple low-altitude images and maintains a high level of image resolution.

[0051] As one feasible approach, relative radiometric correction is performed on the low-altitude local imagery of the UAV based on the UAV's high-altitude global imagery information, resulting in relative radiometric correction parameters, including: Based on the UAV's high-altitude global imagery and low-altitude local imagery, the initial matching feature points of the UAV's high-altitude global imagery and low-altitude local imagery are determined. Image matching points with pixel values ​​of maximum and minimum values ​​within a preset scale range are selected as candidate matching feature points; Determine the scale-invariant feature transformation feature similarity of candidate matching feature points, and select candidate matching feature points whose scale-invariant feature transformation feature similarity is greater than a preset threshold as matching feature points; The relative radiometric correction parameters are determined based on the matching feature points, the UAV's high-altitude global image information, and the UAV's low-altitude local image information.

[0052] For image feature point-based methods, the SIFT operator is used to extract and match feature points between the UAV's high-altitude global image and low-altitude local image. Then, based on the UAV's high-altitude global image and low-altitude local image information, the pixel values ​​corresponding to the matching feature points are found. Relative radiometric correction regression is then performed based on the pixel values ​​to obtain the relative radiometric correction parameters.

[0053] The specific steps for determining the matching feature points between UAV high-altitude global imagery and UAV low-altitude local imagery using the SIFT operator extraction and matching algorithm include: Based on the UAV's high-altitude global imagery and low-altitude local imagery, initial matching feature points are determined for the UAV's high-altitude global imagery and low-altitude local imagery. Specifically, based on the UAV's high-altitude global imagery and low-altitude local imagery, continuous Gaussian blurring is performed on the UAV's high-altitude global imagery and low-altitude local imagery using different Gaussian kernel standard deviations to generate a series of images of the same size; Based on a series of images of the same size, the initial matching feature points of the UAV high-altitude global image and the UAV low-altitude local image are determined.

[0054] Based on the UAV's high-altitude global imagery and low-altitude local imagery, the position of each pixel in the UAV's high-altitude global imagery and low-altitude local imagery is determined. Continuous Gaussian blurring is then applied to the UAV's high-altitude global imagery and low-altitude local imagery using different Gaussian kernel standard deviations to generate a series of images of the same size, referred to as an octave.

[0055] The most blurred image in the octave is downsampled by halving its size. This process is repeated on the resulting image to construct a Gaussian pyramid image composed of multiple octaves. The difference between two adjacent images in the same octave of the Gaussian pyramid is then calculated to obtain the Difference of Gaussian Pyramid (DoG). At this point, the pixels present in the DoG of the UAV high-altitude global image and the UAV low-altitude local image are used as the initial matching feature points.

[0056] The preset scale range can be the range of 8 neighboring points at the same scale and the range of 9 neighboring points at each of the upper and lower adjacent scales. In DoG, each pixel is compared with its 8 neighboring points at the same scale and the 9 neighboring points at each of the upper and lower adjacent scales. If the value of the pixel is an extreme value, that is, the maximum or minimum value, it is marked as a candidate matching feature point in that scale space.

[0057] To obtain stable and effective candidate matching feature points, unstable extreme points are eliminated. First, a ternary quadratic function (x, y, σ) is fitted to the candidate extreme points to obtain their sub-pixel positions. Simultaneously, contrast is calculated, and a threshold is set to eliminate extreme points with contrast below the threshold. Then, edge effects are removed by calculating the Hessian matrix (H) and its eigenvalues ​​at the extreme points. Edge points among the extreme points are eliminated using these eigenvalues, resulting in the final candidate matching feature points.

[0058] Within a 16×16 neighborhood of the extreme point, the gradient magnitude and direction of each pixel are calculated. The gradient direction of each pixel is discretized, with 10° increments, and each direction is accumulated using its magnitude as a weight to generate a direction histogram within the neighborhood. The maximum value in the histogram is taken as the principal direction of the extreme point, and directions with values ​​greater than 80% of the principal direction are taken as the secondary directions of the extreme point.

[0059] The coordinate axes of the neighborhood of the extreme point are rotated to the main direction of that point. A 16x16 pixel window is taken centered on the extreme point, and this window is divided into 4x4 sub-regions, for a total of 16 sub-regions. Within each sub-region, the gradient magnitude and direction of each pixel are calculated, generating a direction histogram with 45° increments. The direction histograms of the 16 sub-regions are concatenated to form a 128-dimensional feature vector, i.e., the SIFT feature. The distance between two 128-dimensional feature vectors is calculated; the smaller the distance, the higher the SIFT feature similarity; the larger the distance, the lower the SIFT feature similarity. The SIFT algorithm can not only determine differences between different scales but also select relatively stable feature points that do not change with lighting conditions, rotation, etc.

[0060] The preset threshold can be set according to the user's accuracy requirements. For example, if the preset threshold is 0.75, candidate matching feature points with SIFT feature similarity greater than 0.75 will be selected as matching feature points.

[0061] Based on the UAV's high-altitude global imagery and low-altitude local imagery, the corresponding positions of matching feature points in the UAV's high-altitude global imagery and low-altitude local imagery are determined. Based on the corresponding coordinates of the matching feature points in the two images, their corresponding pixel values ​​are found. The relative radiometric correction parameters can be obtained by performing a least-squares relative radiometric correction regression operation on them in MATLAB software.

[0062] As one feasible approach, relative radiometric correction is performed on the low-altitude local imagery of the UAV based on the UAV's high-altitude global imagery information, resulting in relative radiometric correction parameters, including: Based on the high-altitude global image information and the low-altitude local image information of the UAV, the first and second location areas of the preset ground features in the high-altitude global image and the low-altitude local image of the UAV are determined respectively. Calculate the average pixel value for the first location region and the second location region respectively; The average pixel values ​​of the first and second location regions of each preset feature are linearly regressed using the least squares method to obtain the relative radiometric correction parameters.

[0063] To prevent the data from several matching points from being too singular and random, a method based on the average value of the land cover area can be used to obtain the relative radiometric correction parameters. This involves delineating a large region of interest for the same land cover in both images and calculating the average pixel value of that region, then performing regression based on the region. This ensures the accuracy and reliability of the linear regression method for relative radiometric correction of the images.

[0064] Preset ground features are pre-selected ground features used for relative radiometric correction, which may include calibration blankets, concrete surfaces, car hoods, trees, and grasslands.

[0065] Based on UAV high-altitude global imagery and UAV low-altitude local imagery, the positions of preset ground features within these images are determined, and thus the corresponding regions of interest (ROIs) for each feature are identified. Specifically, the first location region is the ROI for the preset ground feature in the UAV high-altitude global imagery, and the second location region is the ROI for the preset ground feature in the UAV low-altitude local imagery. Multiple identical preset ground features are delineated in the UAV high-altitude global and low-altitude local images using ENVI software, and the first and second location regions corresponding to each preset ground feature are determined.

[0066] All pixel values ​​within the first and second location regions are acquired, and the average value of all pixel values ​​is calculated to obtain the average pixel value. Relative radiometric correction is then performed based on the average pixel value of the preset ground feature location regions, thereby ensuring the accuracy and reliability of the linear regression method for relative radiometric correction of the image.

[0067] The average pixel values ​​of the first and second location regions of each preset feature are linearly regressed using the least squares method to obtain the relative radiometric correction parameters.

[0068] As one feasible approach, based on relative radiometric correction parameters and UAV low-altitude local imagery information, low-altitude global imagery information of the UAV is generated, including: Based on the relative radiometric correction parameters and UAV low-altitude local image information, the pixel values ​​in each UAV low-altitude local image are subjected to relative radiometric correction to obtain the UAV low-altitude local corrected image. By stitching together locally corrected low-altitude images from the UAV, global low-altitude image information of the UAV can be obtained.

[0069] After obtaining the relative radiometric correction parameters, the pixel value of each pixel in the UAV low-altitude local image information is obtained. The relative radiometric correction parameters are used to perform relative radiometric correction on each pixel value in the UAV low-altitude local image to obtain the UAV low-altitude local corrected image.

[0070] Pix4D software can be used to directly input UAV low-altitude local corrected imagery with forward and lateral overlap to obtain UAV low-altitude global imagery information containing geographic coordinates. This UAV low-altitude global imagery information can be used for interpretation and display, and also facilitates subsequent absolute radiometric correction and agricultural applications.

[0071] As one feasible approach, ground feature spectral information includes ground feature spectral reflectance; based on the ground feature spectral information and UAV low-altitude global imagery, absolute radiometric correction is performed to obtain the ground reflectance, including: Based on the low-altitude global imagery information from the UAV, determine the calibration blanket image area in the low-altitude global imagery of the UAV. Calculate the average pixel value of the calibration blanket image area; Based on the average pixel value and spectral reflectance of the ground cover in the calibration blanket image area, the linear regression coefficients were determined by the least squares method. Absolute radiometric correction is performed on each pixel value in the UAV low-altitude global image based on the linear regression coefficients and preset formulas to obtain the true ground reflectance.

[0072] Based on the pixel location information and pixel values ​​in the low-altitude global imagery of the UAV, the image area of ​​each calibration blanket is delineated. The average pixel value is obtained by averaging the pixel values ​​of all pixels in each calibration blanket image area. Based on the average pixel value of the calibration blanket image area and the spectral reflectance of ground features, the linear regression coefficients are determined using the least squares method. The preset formula, consisting of linear regression coefficients, can be expressed as: ; in, This represents the true reflectivity of the ground. , is the linear regression coefficient, and DN is the average pixel value of the calibration blanket image area in the UAV low-altitude global image information.

[0073] After obtaining the linear regression coefficients, the linear regression coefficients and the average pixel value of the calibration blanket image area are substituted into the above preset formula to obtain the ground true reflectance of the UAV low-altitude global image, thus completing the joint radiometric correction of UAV remote sensing images.

[0074] This application uses relative radiometric correction to ensure consistent grayscale values ​​between UAV high-altitude global imagery and UAV low-altitude local imagery, eliminating the impact of weather and lighting changes during low-altitude image capture. Then, the relative radiometric correction result is compared with the ground-measured spectral reflectance of ground features using absolute radiometric correction to obtain the true ground reflectance. This combined method of relative and absolute radiometric correction eliminates the influence of weather, lighting, and UAV attitude, resulting in higher radiometric correction accuracy, enhanced comparability of UAV imagery, and greater convenience for standardized UAV image production and subsequent analysis and application.

[0075] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated in this application, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Furthermore, Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0076] Figure 3 This application provides a schematic diagram of a UAV remote sensing image joint radiometric correction device, used to perform, as described in this embodiment. Figure 1 The method flow is shown below. Figure 3 As shown, the device may include: an acquisition module 310, a relative radiation correction module 320, a generation module 330, and an absolute radiation correction module 340. The main functions of each component module are as follows: The acquisition module 310 is used to acquire ground feature spectral information, UAV high-altitude global image information, and UAV low-altitude local image information. The relative radiometric correction module 320 is used to perform relative radiometric correction on the low-altitude local image information of the UAV based on the high-altitude global image information of the UAV, and obtain the relative radiometric correction parameters. The generation module 330 is used to generate low-altitude global image information of the UAV based on the relative radiometric correction parameters and the low-altitude local image information of the UAV. The absolute radiometric correction module 340 is used to perform absolute radiometric correction based on ground object spectral information and UAV low-altitude global image information to obtain the true ground reflectance and complete the joint radiometric correction of UAV remote sensing images.

[0077] As an feasible approach, the relative radiometric correction module 320 is specifically used for: cropping the UAV's high-altitude global imagery based on the UAV's low-altitude local imagery to obtain a high-altitude target area imagery, wherein the high-altitude target area imagery is identical to the imagery containing the calibration blanket in the UAV's low-altitude local imagery; using the high-altitude target area imagery as a reference image, performing histogram matching on the UAV's low-altitude local imagery using remote sensing image processing software to obtain relative radiometric correction parameters.

[0078] As one feasible approach, the relative radiometric correction module 320 is specifically used for: determining initial matching feature points for the UAV's high-altitude global imagery and low-altitude local imagery based on the UAV's high-altitude global imagery and low-altitude local imagery; selecting image matching points with pixel values ​​of maximum and minimum within a preset scale range as candidate matching feature points; determining the scale-invariant feature transformation similarity of the candidate matching feature points, and selecting candidate matching feature points with a scale-invariant feature transformation similarity greater than a preset threshold as matching feature points; and determining relative radiometric correction parameters based on the matching feature points, the UAV's high-altitude global imagery, and the UAV's low-altitude local imagery.

[0079] As one feasible approach, the relative radiometric correction module 320 is specifically used to: perform continuous Gaussian blurring on the UAV's high-altitude global imagery and low-altitude local imagery using different Gaussian kernel standard deviations, based on the UAV's high-altitude global imagery and low-altitude local imagery, to generate a series of images of the same size; and determine the initial matching feature points of the UAV's high-altitude global imagery and low-altitude local imagery based on the series of images of the same size.

[0080] As one feasible approach, the relative radiometric correction module 320 is specifically used to: determine the first and second location regions of preset ground features in the UAV's high-altitude global imagery and low-altitude local imagery, respectively, based on the UAV's high-altitude local imagery and low-altitude local imagery; calculate the average pixel values ​​of the first and second location regions, respectively; and perform linear regression on the average pixel values ​​of the first and second location regions of each preset ground feature using the least squares method to obtain the relative radiometric correction parameters.

[0081] As one possible approach, the generation module 330 is specifically used to: perform relative radiometric correction on the pixel values ​​in each UAV low-altitude local image based on the relative radiometric correction parameters and the UAV low-altitude local image information to obtain a UAV low-altitude local corrected image; and stitch together the UAV low-altitude local corrected images to obtain UAV low-altitude global image information.

[0082] As an feasible approach, the absolute radiometric correction module 340 is specifically used for: determining the calibration blanket image area in the UAV's low-altitude global image based on the UAV's low-altitude global image information; calculating the average pixel value of the calibration blanket image area; determining the linear regression coefficient using the least squares method based on the average pixel value of the calibration blanket image area and the spectral reflectance of ground objects; and performing absolute radiometric correction on each pixel value in the UAV's low-altitude global image based on the linear regression coefficient and a preset formula to obtain the true ground reflectance.

[0083] The same or similar parts among the above embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.

[0084] It should be noted that the embodiments of this application may involve the use of user data. In practical applications, user-specific personal data may be used in the scheme described herein within the scope permitted by applicable laws and regulations, provided that it complies with the applicable laws and regulations of the country (e.g., explicit consent from the user, actual notification to the user, explicit authorization from the user, etc.).

[0085] According to embodiments of this application, this application also provides a computer device and a computer-readable storage medium.

[0086] like Figure 4 The diagram shown is a block diagram of a computer device according to an embodiment of this application. The term "computer device" is intended to represent various forms of digital computers or mobile devices. The digital computer may include a desktop computer, a portable computer, a workbench, a personal digital assistant, a server, a mainframe computer, and other suitable computers. The mobile device may include a tablet computer, a smartphone, a wearable device, etc.

[0087] like Figure 4 As shown, the computer device 400 includes a computing unit 401, a ROM 402, a RAM 403, a bus 404, and an input / output (I / O) interface 405. The computing unit 401, ROM 402, and RAM 403 are interconnected via the bus 404. The input / output (I / O) interface 405 is also connected to the bus 404.

[0088] The computing unit 401 can execute various processes in the method embodiments of this application according to computer instructions stored in the read-only memory (ROM) 402 or computer instructions loaded from the storage unit 408 into the random access memory (RAM) 403. The computing unit 401 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. The computing unit 401 can include, but is not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. In some embodiments, the methods provided in the embodiments of this application can be implemented as computer software programs, which are tangibly contained in a computer-readable storage medium, such as the storage unit 408.

[0089] RAM 403 can also store various programs and data required for the operation of computer device 400. Part or all of the computer program can be loaded and / or installed on computer device 400 via ROM 402 and / or communication unit 409.

[0090] The input unit 406, output unit 407, storage unit 408, and communication unit 409 in the computer device 400 can be connected to the I / O interface 405. The input unit 406 can be, for example, a keyboard, mouse, touchscreen, or microphone; the output unit 407 can be, for example, a monitor, speaker, or indicator light. The computer device 400 can exchange information and data with other devices through the communication unit 409.

[0091] It should be noted that the device may also include other components necessary for normal operation. It may also include only the components necessary for implementing the solution of this application, without necessarily including all the components shown in the figures.

[0092] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SOCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof.

[0093] The computer instructions used to implement the methods of this application may be written in any combination of one or more programming languages. These computer instructions may be provided to the computing unit 401 such that when executed by the computing unit 401, such as a processor, the computer instructions cause the execution of the steps involved in the embodiments of the methods of this application.

[0094] The computer-readable storage medium provided in this application can be a tangible medium that can contain or store computer instructions for performing the steps involved in the method embodiments of this application. The computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, and other forms of storage media.

[0095] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for joint radiometric correction of UAV remote sensing images, characterized in that, The method includes: Acquire spectral information of ground features, global high-altitude imagery from UAVs, and local low-altitude imagery from UAVs; Based on the high-altitude global image information of the UAV, relative radiometric correction is performed on the low-altitude local image information of the UAV to obtain the relative radiometric correction parameters. Based on the relative radiation correction parameters and the UAV low-altitude local image information, generate UAV low-altitude global image information; Absolute radiometric correction is performed based on the spectral information of the ground features and the low-altitude global image information of the UAV to obtain the true ground reflectance, thus completing the joint radiometric correction of the UAV remote sensing image.

2. The method according to claim 1, characterized in that, The process involves performing relative radiometric correction on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information, resulting in relative radiometric correction parameters, including: Based on the UAV's high-altitude global image information and the UAV's low-altitude local image information, the UAV's high-altitude global image is cropped to obtain a high-altitude target area image. The high-altitude target area image is the same as the image containing the calibration blanket in the UAV's low-altitude local image. Using the high-altitude target area image as a reference image, histogram matching is performed on the low-altitude local image information of the UAV using remote sensing image processing software to obtain relative radiometric correction parameters.

3. The method according to claim 1, characterized in that, The process involves performing relative radiometric correction on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information, resulting in relative radiometric correction parameters, including: Based on the UAV's high-altitude global image information and the UAV's low-altitude local image information, the initial matching feature points of the UAV's high-altitude global image and the UAV's low-altitude local image are determined. Image matching points with pixel values ​​of maximum and minimum values ​​within a preset scale range are selected as candidate matching feature points; Determine the scale-invariant feature transformation feature similarity of the candidate matching feature points, and select candidate matching feature points whose scale-invariant feature transformation feature similarity is greater than a preset threshold as matching feature points; The relative radiometric correction parameters are determined based on the matching feature points, the UAV's high-altitude global image information, and the UAV's low-altitude local image information.

4. The method according to claim 1, characterized in that, The step of determining the initial matching feature points of the UAV's high-altitude global imagery and low-altitude local imagery based on the UAV's high-altitude global imagery and low-altitude local imagery includes: Based on the UAV's high-altitude global image information and the UAV's low-altitude local image information, the UAV's high-altitude global image and the UAV's low-altitude local image are continuously Gaussian blurred using different Gaussian kernel standard deviations to generate a series of images of the same size. Based on the series of images of the same size, the initial matching feature points of the UAV high-altitude global image and the UAV low-altitude local image are determined.

5. The method according to claim 1, characterized in that, The process involves performing relative radiometric correction on the low-altitude local image information of the UAV based on the UAV's high-altitude global image information, resulting in relative radiometric correction parameters, including: Based on the UAV's high-altitude global image information and the UAV's low-altitude local image information, the first and second location regions of the preset ground features in the UAV's high-altitude global image and the UAV's low-altitude local image are determined respectively. Calculate the average pixel value for the first location region and the second location region respectively; The average pixel values ​​of the first and second location regions of each preset feature are linearly regressed using the least squares method to obtain the relative radiometric correction parameters.

6. The method according to claim 1, characterized in that, The step of generating low-altitude global image information of the UAV based on the relative radiometric correction parameters and the UAV low-altitude local image information includes: Based on the relative radiometric correction parameters and the UAV low-altitude local image information, the pixel values ​​in each UAV low-altitude local image are subjected to relative radiometric correction to obtain the UAV low-altitude local corrected image. By stitching together the low-altitude local corrected images of the UAV, the low-altitude global image information of the UAV is obtained.

7. The method according to claim 1, characterized in that, The ground feature spectral information includes ground feature spectral reflectance; the process of performing absolute radiometric correction based on the ground feature spectral information and the UAV low-altitude global imagery to obtain ground reflectance includes: Based on the UAV low-altitude global image information, determine the calibration blanket image area in the UAV low-altitude global image. Calculate the average pixel value of the calibration blanket image area; Based on the average pixel value of the calibration blanket image area and the spectral reflectance of the ground features, the linear regression coefficients are determined by the least squares method. Based on the linear regression coefficients and the preset formula, absolute radiometric correction is performed on each pixel value in the low-altitude global image of the UAV to obtain the true ground reflectance.

8. A device for joint radiometric correction of UAV remote sensing images, characterized in that, The device includes: The acquisition module is used to acquire spectral information of ground features, global high-altitude imagery information of UAVs, and local low-altitude imagery information of UAVs. The relative radiometric correction module is used to perform relative radiometric correction on the low-altitude local image information of the UAV based on the high-altitude global image information of the UAV, and obtain the relative radiometric correction parameters. The generation module is used to generate low-altitude global image information of the UAV based on the relative radiometric correction parameters and the low-altitude local image information of the UAV. The absolute radiometric correction module is used to perform absolute radiometric correction based on the spectral information of the ground features and the low-altitude global image information of the UAV, to obtain the true ground reflectance and complete the joint radiometric correction of the UAV remote sensing image.

9. A computer device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores computer instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 7.

Images