An unmanned aerial vehicle remote sensing mapping image enhancement method and system

Abstract

The application provides a UAV remote sensing mapping image enhancement method and system, a global mapping relationship of radiation brightness between two time phase images is constructed through homonymic feature points in a reference time phase image and a time phase image to be corrected, and then the time phase image to be corrected is corrected in radiation through the global mapping relationship, so that a preliminary correction image is obtained; multi-dimensional similar features in each super pixel region of the preliminary correction image and the reference time phase image are determined; a stable region mask in which no real object changes occur is extracted, the global mapping relationship is fitted in radiation bias through the stable region mask, and a bias field of remote sensing mapping radiation brightness is obtained; and then a remote sensing mapping enhanced image consistent with the radiation characteristics of the reference time phase image is generated. The scheme of the application can reduce radiation distortion of the enhanced image caused by regional light difference.

Description

Technical Field

[0001] This application relates to the field of image enhancement technology, and more specifically, to a method and system for enhancing remote sensing and mapping images from unmanned aerial vehicles (UAVs). Background Technology

[0002] Image enhancement refers to the process of transforming blurry, low-quality, or information-limited two-dimensional image data into an optimized version with clearer visual representation and more prominent features through a series of digital image processing operations such as contrast adjustment, noise suppression, edge sharpening, and color correction. This enhances the image content from its original perception to its enhanced expression, ultimately providing a more reliable and efficient visual information foundation for subsequent analysis, recognition, and visualization.

[0003] UAV remote sensing image enhancement refers to the technology of improving the quality and optimizing the features of raw remote sensing images acquired by UAV platforms through a series of aerospace image processing algorithms, such as radiometric and geometric correction and multispectral information fusion. This transforms the initial image data, which is limited by atmospheric conditions, sensors, and environmental interference, into enhanced image products with higher geometric fidelity, better radiometric consistency, and richer ground feature resolution. In the traditional UAV remote sensing image enhancement process, radiometric normalization is mostly based on direct brightness mapping based on a global linear model. This makes it difficult to effectively distinguish between spatially varying radiometric differences caused by factors such as terrain undulations, bidirectional reflection effects, and regional thin clouds and real ground feature changes. As a result, enhanced images still have local radiometric distortion problems in complex scenes. Therefore, how to reduce radiometric distortion caused by regional illumination differences in enhanced images has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for enhancing remote sensing and mapping images from unmanned aerial vehicles (UAVs), which can reduce radiation distortion in enhanced images caused by regional differences in illumination.

[0005] In a first aspect, this application provides a method for enhancing remote sensing mapping images from unmanned aerial vehicles (UAVs), comprising the following steps: The baseline and phase images to be corrected of the target geographic area are collected using a drone platform. Based on the corresponding feature points in the reference time-phase image and the time-phase image to be corrected, a global mapping relationship of radiance between the two time-phase images is constructed. Then, the time-phase image to be corrected is radiometrically corrected through the global mapping relationship to obtain a preliminary corrected image. Superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to determine the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region; Based on the multidimensional similarity features, a stable region mask that has not undergone real ground object changes is extracted from all superpixel regions. The global mapping relationship is then fitted with the stable region mask to obtain the bias field of the radiance during remote sensing mapping. The radiometric transformation of the time-phase image to be corrected is performed by the bias field of the radiometric brightness, thereby generating a remote sensing mapping enhancement image with radiometric characteristics consistent with the reference time-phase image.

[0006] In some embodiments, constructing a global mapping relationship of radiance between the two temporal images based on corresponding feature points in the reference temporal image and the temporal image to be corrected specifically includes: Multiple feature points with the same name are extracted from the reference time-phase image and the time-phase image to be corrected; A radiative conversion model between the reference temporal image and the temporal image to be corrected is fitted by the radiance of all corresponding feature points; A global mapping relationship of radiance between two temporal images is constructed based on the aforementioned radiative conversion model.

[0007] In some embodiments, performing radiometric correction on the phase image to be corrected using the global mapping relationship to obtain a preliminary corrected image specifically includes: The gain coefficient and bias coefficient of radiance during remote sensing mapping are determined through the global mapping relationship. The radiance of each pixel in the phase image to be corrected is linearly transformed based on the gain coefficient and the bias coefficient. A preliminary corrected image is generated based on the linear transformation results.

[0008] In some embodiments, performing superpixel segmentation on the preliminary corrected image and the reference temporal image, and then determining the multidimensional similarity features between the preliminary corrected image and the reference temporal image in each superpixel region, specifically includes: Superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to obtain a set of reference time-phase superpixel regions and a set of preliminary corrected superpixel regions. A spatial correspondence between superpixel regions is established based on the overlap of geographic coordinates between the regions in the reference temporal superpixel region set and the preliminary correction superpixel region set. The spatial correspondence is used to calculate the multidimensional similarity features between the preliminary corrected image and the reference temporal image in each corresponding superpixel region.

[0009] In some embodiments, extracting a stable region mask that has not undergone real-world feature changes from all superpixel regions based on the multidimensional similarity features specifically includes: Based on the multidimensional similarity features, a comprehensive similarity score between the preliminary corrected image and the reference temporal image is determined in each superpixel region; Based on a preset similarity threshold and the comprehensive similarity score, the unchanged regions are extracted from all superpixel regions; A stable region mask that does not exhibit any actual changes in ground features is generated based on the unchanged region.

[0010] In some embodiments, the radiative bias fitting of the global mapping relationship using the stable region mask to obtain the bias field of radiance during remote sensing mapping specifically includes: Radiance sample data were collected within the masked area of ​​the stable region. Spatial radiance optimization is performed on the bias parameters in the global mapping relationship using the radiance sample data; The bias field of radiance during remote sensing mapping is generated based on the optimization results.

[0011] In some embodiments, performing a radiometric transformation on the time-phase image to be corrected using the radiance bias field to generate a remote sensing mapping enhancement image with radiometric characteristics consistent with the reference time-phase image specifically includes: The radiance bias field is combined with the global mapping relationship to construct a set of radiance correction parameters for image enhancement; The radiometric transformation is performed on each pixel in the phase image to be corrected using the radiometric correction parameter set. Based on the transformation results, a remote sensing mapping enhancement image consistent with the radiometric characteristics of the reference temporal image is generated.

[0012] Secondly, this application provides a UAV remote sensing mapping image enhancement system, comprising: The acquisition module is used to acquire baseline and phase images of the target geographic area via a drone platform. The processing module is used to construct a global mapping relationship of radiometric brightness between the reference time-phase image and the time-phase image to be corrected based on the same feature points in the reference time-phase image and the time-phase image to be corrected, and then perform radiometric correction on the time-phase image to be corrected through the global mapping relationship to obtain a preliminary corrected image; The processing module is further configured to perform superpixel segmentation on the preliminary corrected image and the reference time-phase image, thereby determining the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region; The processing module is further configured to extract a stable region mask that has not undergone real ground object changes from all superpixel regions based on the multidimensional similarity features, and to perform radiative bias fitting on the global mapping relationship using the stable region mask to obtain the bias field of radiance during remote sensing mapping. The execution module is used to perform a radiometric transformation on the time-phase image to be corrected through the bias field of the radiometric brightness, thereby generating a remote sensing mapping enhancement image that is consistent with the radiometric characteristics of the reference time-phase image.

[0013] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described UAV remote sensing mapping image enhancement method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described UAV remote sensing mapping image enhancement method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The UAV remote sensing mapping image enhancement method and system provided in this application firstly constructs a global mapping relationship of radiance between the two time-phase images based on the corresponding feature points in the reference time-phase image and the time-phase image to be corrected. Then, the time-phase image to be corrected is radiometrically corrected using this global mapping relationship to obtain a preliminary corrected image. This process establishes a foundation for overall radiometric consistency between the two time-phase images, corrects systematic radiometric deviations caused by environmental factors through the global mapping relationship, and provides image data with preliminary uniformity of radiometric levels for subsequent processing. Subsequently, superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to determine the multidimensional similarity features in each superpixel region. Based on these multidimensional similarity features, stable region masks that have not undergone real ground feature changes are extracted from all superpixel regions. The global mapping relationship is then fitted with these stable region masks to obtain a radiometric bias field for remote sensing mapping. This process accurately identifies geographic units with radiometric stability between time phases by quantifying the radiometric stability of local areas. The similarity in multidimensional features such as vegetation index and texture effectively distinguishes radiation anomalies caused by regional illumination fluctuations that only occur in a single time phase from the radiation characteristics of real ground objects. This process provides local spatial and spectral basis for radiation distortion identification. Furthermore, in the process of generating remote sensing mapping enhanced images, a radiation transformation is performed on the image to be corrected using a bias field of radiance. This process can apply high-confidence bias parameters verified in stable regions to the entire image in the form of a spatial field, achieving pixel-level fine radiometric correction. This process dynamically quantifies the amount of correction required at each spatial location through the bias field, providing adjustment parameters that have been rigorously verified by local radiometric consistency for the final image enhancement. This ensures that the enhanced image is no longer just the result of global coarse correction, but a realistic scene reproduction in which the radiation characteristics are highly consistent with the reference time phase in local areas, thereby effectively suppressing color distortion caused by regional illumination unevenness. In summary, this scheme can reduce radiation distortion caused by regional illumination differences in enhanced images. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for enhancing remote sensing and mapping images from an unmanned aerial vehicle (UAV) according to some embodiments of this application; Figure 2 This is a schematic diagram illustrating the process of constructing a global mapping relationship according to some embodiments of this application; Figure 3 This is a schematic diagram of the process for collecting radiance sample data according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a UAV remote sensing mapping image enhancement system according to some embodiments of this application; Figure 5This is an internal structural diagram of a computer device for implementing a UAV remote sensing mapping image enhancement method according to some embodiments of this application. Detailed Implementation

[0017] To better understand the technical solutions in this embodiment, the technical solutions in this embodiment will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0018] refer to Figure 1 The figure is a flowchart illustrating a method for enhancing remote sensing images of an unmanned aerial vehicle (UAV) according to some embodiments of this application. The method mainly includes the following steps: In step 101, a reference time-phase image and a time-phase image to be corrected of the target geographic area are acquired through an unmanned aerial vehicle (UAV) platform.

[0019] In practice, an unmanned aerial vehicle (UAV) platform equipped with a multispectral imager and a high-precision positioning and attitude determination system can conduct aerial surveys of the target geographic area twice, under similar weather conditions and solar altitude angles. During the first flight to acquire the baseline time-phase image, flight altitude, flight path overlap rate, and sensor parameters are recorded. Centimeter-level absolute geographic coordinates are obtained using a Global Navigation Satellite System (GNSS) receiver, and three-axis attitude angles are recorded using an Inertial Measurement Unit (IMU). Subsequent flights maintain the same platform configuration and aerial survey plan. While acquiring the time-phase image to be corrected, positioning and attitude determination data are simultaneously collected and timestamped. This yields both the baseline time-phase image and the time-phase image to be corrected for the target geographic area. It should be noted that the multispectral imager refers to a professional remote sensing sensor capable of simultaneously acquiring data in multiple spectral bands such as blue, green, red, and near-infrared, and its spectral response characteristics must remain stable between two acquisitions; the high-precision positioning and attitude determination system refers to a combined measurement system integrating a GNSS receiver and an IMU, where the GNSS provides centimeter-level absolute positioning accuracy and the IMU provides attitude measurement accuracy better than 0.01 degrees; the reference time phase image refers to a high-quality remote sensing image acquired during the stable period of ground features, serving as a radiometric reference; the time phase image to be corrected refers to a subsequently acquired image that needs to undergo radiometric consistency processing with the reference time phase; the target geographical area includes radiometrically stable ground features such as bare soil and water bodies.

[0020] In step 102, a global mapping relationship of radiance between the two time phase images is constructed based on the corresponding feature points in the reference time phase image and the time phase image to be corrected. Then, the time phase image to be corrected is radiometrically corrected through the global mapping relationship to obtain a preliminary corrected image.

[0021] In some embodiments, reference Figure 2 As shown in the figure, this is a schematic flowchart illustrating the construction of a global mapping relationship in some embodiments of this application. The global mapping relationship of radiance between two time-phase images based on the corresponding feature points in the reference time-phase image and the time-phase image to be corrected can be achieved by the following steps: First, in step 1021, multiple feature points with the same name are extracted from the reference time phase image and the time phase image to be corrected; Then, in step 1022, the radiative conversion model between the reference phase image and the phase image to be corrected is fitted by the radiance of all corresponding feature points; Finally, in step 1023, a global mapping relationship of radiance between the two temporal images is constructed based on the radiative conversion model.

[0022] In specific implementation, extracting multiple corresponding feature points from the reference time-phase image and the time-phase image to be corrected can be achieved in the following way: First, the reference time-phase image and the time-phase image to be corrected are preprocessed, including radiometric calibration and atmospheric correction, converting pixel values ​​into surface reflectance; then, a scale-invariant feature transform feature detection algorithm is used to extract key points from the reference time-phase image and the time-phase image to be corrected, and their 128-dimensional feature descriptors are calculated; next, a fast nearest neighbor search library matcher is used to perform similarity matching of feature descriptors, retaining feature point pairs that are successfully matched bidirectionally; then, a random sampling consensus algorithm is used to remove mismatched point pairs, and an image pyramid strategy is used to ensure the uniformity of feature point distribution in different scale spaces; finally, feature points distributed on radiometrically invariant ground features (such as bare soil, built-up areas, and stable water bodies) are selected as corresponding feature points, thus obtaining multiple corresponding feature points.

[0023] It should be noted that the aforementioned corresponding feature points refer to pairs of corresponding feature points representing the same geographical location in remote sensing images of different time phases. These are used to establish pixel-level spatial correspondences between two temporal images, providing a geometric registration basis for radiometric consistency conversion.

[0024] In specific implementation, fitting the radiative conversion model between the reference time-phase image and the time-phase image to be corrected using the radiance of all corresponding feature points can be achieved in the following way: First, obtain the radiance values ​​of all corresponding feature points at corresponding positions in the reference time-phase image and the time-phase image to be corrected, forming a sample dataset; then, use the least squares linear regression method, with the radiance value of the time-phase image to be corrected as the independent variable and the radiance value of the reference time-phase image as the dependent variable, to establish a linear regression model; minimize the sum of squared prediction residuals of all sample points through iterative calculation, and use the obtained regression model as the radiative conversion model between the reference time-phase image and the time-phase image to be corrected. Preferably, the slope parameter of the regression model can be used as the gain coefficient and the intercept parameter as the bias coefficient; as a preferred embodiment, the fitting results can be subjected to a significance test, and abnormal sample points that deviate from three times the standard deviation of the mean can be removed before refitting to ensure the statistical significance of the radiative conversion model.

[0025] It should be noted that the aforementioned radiation conversion model refers to a mathematical model that describes the conversion relationship of radiance values ​​between two temporal images. It is used to quantify the systematic differences in radiation characteristics between the temporal phase to be corrected and the reference temporal phase, and to provide a mathematical conversion basis for radiation normalization.

[0026] In specific implementation, the global mapping relationship of radiance between two temporal images based on the radiation conversion model can be achieved in the following way: using the gain coefficient and bias coefficient of the fitted regression model as core parameters, a radiance mapping function based on linear transformation is constructed; the radiance mapping function is encapsulated into a transformation operator that can be applied to the entire image, wherein the gain coefficient is responsible for adjusting the relative radiation difference between images, and the bias coefficient is responsible for correcting the absolute radiation deviation; at the same time, metadata records of the mapping relationship are established, including fitting accuracy indicators, the number of sample points involved in fitting, and spatial distribution information, forming a global mapping relationship of radiance between two temporal images, providing a parameterized model that can be directly called for subsequent radiometric correction.

[0027] It should be noted that the global mapping relationship refers to the radiometric conversion rule applicable to the entire image, used to achieve overall correction of the radiometric characteristics of the image to be corrected to the reference image, ensuring radiometric consistency between images.

[0028] In some embodiments, performing radiometric correction on the phase image to be corrected using the global mapping relationship to obtain a preliminary corrected image can be achieved through the following steps: The gain coefficient and bias coefficient of radiance during remote sensing mapping are determined through the global mapping relationship. The radiance of each pixel in the phase image to be corrected is linearly transformed based on the gain coefficient and the bias coefficient. A preliminary corrected image is generated based on the linear transformation results.

[0029] In specific implementation, the gain coefficient and bias coefficient of radiance during remote sensing mapping can be determined through the global mapping relationship in the following way: extract the parameters of the mathematical expression (i.e., the linear regression equation) of the radiative conversion model corresponding to the global mapping relationship, and use the slope parameter of the linear regression equation as the gain coefficient and the intercept parameter as the bias coefficient; preferably, the gain coefficient can be confirmed to be in the effective range of 0.8-1.2 and the bias coefficient to be in the range of -0.1 to 0.1 reflectance units through the parameter verification process; when the parameters exceed the reasonable range, return to the feature point screening and model reconstruction process.

[0030] In specific implementation, the linear transformation of the radiance of each pixel in the time-phase image to be corrected based on the gain coefficient and the bias coefficient can be achieved in the following way: taking the radiance value of the time-phase image to be corrected as input, the radiance value of each pixel is first multiplied by the gain coefficient, and then added to the bias coefficient to obtain the transformed radiance value; the same linear transformation operation is performed independently on each band of the multi-band image, thereby realizing the linear transformation of the radiance of each pixel in the time-phase image to be corrected; preferably, a block segmentation parallel processing strategy can be adopted to divide the time-phase image to be corrected into several processing blocks, and multi-threading technology is used to process the pixel data in each block synchronously.

[0031] In specific implementation, the preliminary corrected image can be generated based on the linear transformation result in the following way: the corrected radiance values ​​of all pixels obtained by the linear transformation (i.e., the linear transformation result) are used as the image data source, and reorganized into a two-dimensional digital matrix according to the spatial size of the phase image to be corrected; the digital matrix is ​​bound with the georeferenced information, projection parameters and metadata of the phase image to be corrected to generate the preliminary corrected image; at the same time, the gain coefficient, bias coefficient and correction timestamp information used are recorded in the metadata segment of the image file.

[0032] It should be noted that the preliminary corrected image refers to the intermediate result image obtained after global radiometric correction, which is used to provide input data for subsequent fine processing while maintaining basic radiometric characteristics.

[0033] In step 103, superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to determine the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region.

[0034] In some embodiments, performing superpixel segmentation on the preliminary corrected image and the reference temporal image to determine the multidimensional similarity features between the preliminary corrected image and the reference temporal image in each superpixel region can be achieved through the following steps: Superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to obtain a set of reference time-phase superpixel regions and a set of preliminary corrected superpixel regions. A spatial correspondence between superpixel regions is established based on the overlap of geographic coordinates between the regions in the reference temporal superpixel region set and the preliminary correction superpixel region set. The spatial correspondence is used to calculate the multidimensional similarity features between the preliminary corrected image and the reference temporal image in each corresponding superpixel region.

[0035] In specific implementation, superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to obtain the reference time-phase superpixel region set and the preliminary corrected superpixel region set. This can be achieved in the following way: a linear iterative clustering algorithm is used to perform superpixel segmentation on the reference time-phase image and the preliminary corrected image, respectively. Specifically, the following steps are included: First, the image is converted from the RGB color space to the CIELAB color space. Preferably, the number of superpixel targets can be set to 500, and the compactness factor can be set to 20. Then, cluster centers are initialized and similar pixels are searched in a 5×5 neighborhood. Pixel similarity is evaluated by calculating the weighted sum of color distance and spatial distance, where the color distance weight can be set to 2 and the spatial distance weight to 1. Next, iterative optimization is performed, updating the cluster centers and reassigning pixel labels in each iteration. Finally, post-processing operations are performed, using connected component analysis to merge fragmented regions with an area smaller than a limited pixel threshold and smoothing the superpixel boundaries, ultimately obtaining the reference time-phase superpixel region set and the preliminary corrected superpixel region set with uniform spectral characteristics.

[0036] It should be noted that the superpixel region set refers to a set of pixels with uniform spectral characteristics obtained through image segmentation, which is used to improve image analysis from the pixel level to the region level, thereby enhancing the stability and efficiency of feature calculation.

[0037] In specific implementation, the spatial correspondence between superpixel regions can be established based on the overlap of geographical coordinates between the reference time-phase superpixel region set and the preliminary correction superpixel region set. This can be achieved in the following way: First, extract the coordinates of the minimum bounding rectangle of each superpixel region and calculate the intersection area between the reference time-phase superpixel region and the preliminary correction superpixel region; then calculate the overlap index, which is the ratio of the intersection area to the smaller area of ​​the two regions; mark regions with an overlap exceeding a preset threshold as candidate corresponding regions; then perform bidirectional verification on the candidate corresponding regions to ensure that each reference region establishes a correspondence with at most one correction region; finally, for regions without a correspondence, search for the nearest neighbor region in its spatial neighborhood and supplement it with spatial interpolation to establish a complete region correspondence table. The region number mapping relationship and overlap value in the region correspondence table are used as the spatial correspondence between superpixel regions.

[0038] It should be noted that the spatial correspondence refers to the geographic location mapping relationship between the superpixel regions of two temporal images, which is used to ensure that subsequent feature comparisons are performed within the same geographic unit and to avoid errors caused by spatial misalignment.

[0039] In specific implementation, the calculation of multidimensional similarity features between the preliminary corrected image and the reference temporal image in each corresponding superpixel region based on the spatial correspondence can be achieved in the following way: For each pair of superpixel regions with established spatial correspondence, in the radiation dimension, the reflectance histogram of the superpixel region pair is extracted, and then the Bach distance algorithm is used to calculate the histogram similarity. Specifically, the similarity index is obtained by calculating the sum of the square roots of the products of the two histograms in each group; in the vegetation feature dimension, the normalized vegetation index distribution of the two regions is calculated respectively, and the distribution similarity is evaluated by the Pearson correlation coefficient algorithm. The correlation coefficient is calculated by the ratio of covariance to standard deviation; in the texture dimension, contrast and homogeneity features are extracted based on the gray-level co-occurrence matrix, and the difference in texture features between the two regions is calculated by the Euclidean distance algorithm. The final distance is obtained by taking the square root of the sum of the squares of the differences of each feature; the three-dimensional feature vector composed of the similarity measurements of these three dimensions is used as the multidimensional similarity feature between the preliminary corrected image and the reference temporal image in each corresponding superpixel region.

[0040] It should be noted that the aforementioned multidimensional similarity features refer to feature vectors that quantify the degree of similarity between corresponding regions at two different times from multiple dimensions, and are used to comprehensively evaluate the similarity between regions in terms of land cover attributes and spatial features.

[0041] In step 104, a stable region mask that has not undergone real ground object changes is extracted from all superpixel regions based on the multidimensional similarity features. The global mapping relationship is then fitted with the stable region mask to obtain the bias field of the radiance during remote sensing mapping.

[0042] In some embodiments, extracting a stable region mask that has not undergone real-world feature changes from all superpixel regions based on the multidimensional similarity features can be achieved using the following steps: Based on the multidimensional similarity features, a comprehensive similarity score between the preliminary corrected image and the reference temporal image is determined in each superpixel region; Based on a preset similarity threshold and the comprehensive similarity score, the unchanged regions are extracted from all superpixel regions; A stable region mask that does not exhibit any actual changes in ground features is generated based on the unchanged region.

[0043] In specific implementation, the comprehensive similarity score between the preliminary corrected image and the reference temporal image in each superpixel region based on the multidimensional similarity features can be achieved in the following way: For the multidimensional similarity features of each superpixel region, the measured values ​​of its three dimensions of radiation similarity, vegetation feature similarity, and texture similarity are extracted respectively; a weighted average algorithm is used to calculate the comprehensive similarity score, wherein the weight coefficients of each dimension can be allocated based on the category probabilities output by the trained land cover classifier, or preset according to the prior knowledge of land cover. For example, for the vegetation category, the weight of the vegetation feature dimension is set to 0.5, the weight of the radiation dimension is 0.3, and the weight of the texture dimension is 0.2. The weight of the vegetation feature dimension can be appropriately increased in the vegetation-covered area, and the weight of the texture feature dimension can be appropriately increased in the building area. This application does not limit the specific weight allocation; after standardizing the measured values ​​of each dimension, the weighted sum is calculated according to the allocated weight coefficients, and the weighted sum is used as the comprehensive similarity score.

[0044] In specific implementation, the extraction of unchanged regions from all superpixel regions based on a preset similarity threshold and the comprehensive similarity score can be achieved in the following way: a preset comprehensive similarity score threshold is set, wherein the threshold can be adjusted according to image quality requirements and sensitivity to changes in ground features. A higher threshold can be set in scenarios with strict requirements, and a relatively lower threshold can be set in scenarios with higher tolerance, for example, it can be set between 0.7 and 0.9; superpixel regions with a comprehensive similarity score greater than or equal to the threshold are marked as candidate unchanged regions; spatial continuity analysis is performed on the candidate regions, and isolated regions with too small an area are eliminated, while spatially clustered continuous regions are retained, finally obtaining the unchanged regions.

[0045] In specific implementation, the generation of a stable region mask based on the unchanged region without any real changes in ground features can be achieved in the following way: according to the spatial distribution information of the unchanged region, a binary matrix with the same size as the reference time-phase image is created; the pixel positions corresponding to the unchanged region are set to a first value, and the other regions are set to a second value; the generated binary matrix is ​​morphologically processed to fill the gaps inside the region and smooth the irregular boundaries, and the processed binary matrix is ​​used as the stable region mask.

[0046] It should be noted that the stable region mask refers to a binary image that identifies areas where no changes in ground features have occurred. This image is used to limit the extraction range of reliable samples in subsequent processing, thereby ensuring the accuracy of parameter estimation.

[0047] In some embodiments, the radiative bias fitting of the global mapping relationship using the stable region mask to obtain the bias field of radiance during remote sensing mapping can be achieved by the following steps: Radiance sample data were collected within the masked area of ​​the stable region. Spatial radiance optimization is performed on the bias parameters in the global mapping relationship using the radiance sample data; The bias field of radiance during remote sensing mapping is generated based on the optimization results.

[0048] For specific implementation, refer to Figure 3 As shown in the figure, this is a schematic diagram of the process for collecting radiance sample data in some embodiments of this application. The collection of radiance sample data within the stable region mask can be achieved in the following manner: First, traverse all pixel positions marked as unchanged in the stable region mask, and extract radiance values ​​at the corresponding positions in the preliminary correction image and the reference time-phase image respectively; then, pair the extracted radiance values ​​according to pixel positions to form a sample dataset containing radiance values ​​of the two time phases; at the same time, record the spatial coordinate information of each sample point to provide basic data for subsequent spatial analysis.

[0049] In specific implementation, the spatial radiometric optimization of the bias parameters in the global mapping relationship using the radiance sample data can be achieved in the following way: Based on the collected radiance sample data, a weighted regression analysis method is used to re-estimate the bias parameters in the global mapping relationship. Specifically, this includes: firstly, calculating the weight of each sample point, with the weight value determined according to the spatial distribution density and radiometric consistency of the sample points, assigning higher weights to sample clustering areas and areas with high radiometric consistency; then, establishing a weighted least squares optimization model, using the baseline temporal radiance value as the target value and the preliminary corrected image radiance value as the observation value, and minimizing the prediction residual through an iterative optimization algorithm to obtain the optimized bias parameter estimate.

[0050] In specific implementation, the bias field of radiance during remote sensing mapping based on the optimization results can be achieved in the following way: using the optimized bias parameters as the base value, combined with the spatial distribution characteristics of the sample points, a spatial interpolation method is used to generate a bias field covering the entire map. Specifically, this includes: first, establishing a spatial index based on the spatial location of the sample points; then, using an inverse distance weighted interpolation algorithm to calculate the bias value at each location in dense sample areas; and using a kriging interpolation algorithm based on spatial autocorrelation characteristics to estimate the bias value in sparse sample areas. The division between dense and sparse samples can be based on the local density of the spatial distribution of the sample points. For example, the K-nearest neighbor algorithm can be used to calculate the density of sample points around each location, and when the density is higher than a preset density threshold, it is determined to be a dense area. The power parameter of the inverse distance weighted interpolation can be set to 2, and a spherical model can be used as the variogram model for kriging interpolation. Finally, the interpolation result is quantized into a numerical matrix with fixed precision as the bias field of radiance during remote sensing mapping.

[0051] It should be noted that the bias field of radiance refers to the distribution field of radiation correction bias parameters that reflects spatial variations, used to achieve local adaptive fine radiation correction and eliminate radiation differences due to spatial heterogeneity.

[0052] In step 105, the phase image to be corrected is subjected to radiometric transformation by the bias field of the radiance, thereby generating a remote sensing mapping enhancement image with radiometric characteristics consistent with the reference phase image.

[0053] In some embodiments, the following steps can be used to generate a remote sensing mapping enhancement image with the same radiometric characteristics as the reference time-phase image by performing a radiometric transformation on the time-phase image to be corrected using the radiance bias field: The radiance bias field is combined with the global mapping relationship to construct a set of radiance correction parameters for image enhancement; The radiometric transformation is performed on each pixel in the phase image to be corrected using the radiometric correction parameter set. Based on the transformation results, a remote sensing mapping enhancement image consistent with the radiometric characteristics of the reference temporal image is generated.

[0054] In specific implementation, the radiometric correction parameter set for image enhancement can be constructed by combining the radiance bias field with the global mapping relationship in the following manner: First, the gain coefficients are extracted from the global mapping relationship, and the radiance bias field data is obtained simultaneously; then, a three-dimensional data structure with the same spatial dimension as the bias field is created, which contains two data layers: the first layer is the gain coefficient layer, which stores the gain coefficient values ​​extracted from the global mapping relationship; the second layer is the bias field layer, which stores the bias field values ​​at the corresponding spatial locations; finally, the metadata description of the parameter set is established, including parameter source identifiers, spatial reference system information, and data accuracy description, forming a complete radiometric correction parameter set.

[0055] It should be noted that the radiation correction parameter set refers to a complete set of correction parameters that includes gain and bias parameters, providing all the necessary correction parameters for the final radiation conversion and ensuring the integrity of the correction process.

[0056] In specific implementation, the radiometric transformation of each pixel in the phase image to be corrected using the radiometric correction parameter set can be achieved in the following way: All pixel positions in the phase image to be corrected are traversed in a raster scan order. For each pixel coordinate, the gain value at the corresponding position is read from the gain coefficient layer of the radiometric correction parameter set, and the bias value at the corresponding position is read from the bias field layer. Then, the radiometric transformation calculation is performed: the original radiance value of the current pixel in the image to be corrected is multiplied by the gain value, and the product result is added to the bias value to obtain the transformed radiance value. The above calculation process is performed independently for each multi-band image, thereby realizing the radiometric transformation of each pixel in the phase image to be corrected.

[0057] In specific implementation, the remote sensing mapping enhancement image that is consistent with the radiometric characteristics of the reference time-phase image can be generated based on the transformation result in the following manner: the radiometric values ​​of all pixels that have undergone radiometric transformation are reorganized into a two-dimensional array structure according to the spatial arrangement order of the original image; all metadata information, including geographic coordinate system parameters, projection transformation parameters, and acquisition time identifiers, of the time-phase image to be corrected is fully inherited; at the same time, pixel samples in the stable region are selected, and the radiometric consistency verification index between the transformed image and the reference time-phase image is calculated, including the root mean square error and correlation coefficient; finally, the verified image data and the complete metadata are encapsulated into a standard format remote sensing image file to generate the remote sensing mapping enhancement image that is consistent with the radiometric characteristics of the reference time-phase image.

[0058] It should be noted that the aforementioned enhanced remote sensing mapping image with consistent radiometric characteristics refers to a processed image with comparable radiometric characteristics to the reference time phase, used to support subsequent quantitative remote sensing analysis and change detection applications.

[0059] In another aspect, in some embodiments, this application provides a UAV remote sensing mapping image enhancement system, with reference to... Figure 4 The figure is a schematic diagram of the structure of a UAV remote sensing mapping image enhancement system according to some embodiments of this application. The UAV remote sensing mapping image enhancement system 200 includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to acquire the reference time phase image and the time phase image to be corrected of the target geographical area through the UAV platform; Processing module 202, in this application, is mainly used to construct a global mapping relationship of radiometric brightness between the reference time-phase image and the time-phase image to be corrected based on the same feature points in the reference time-phase image and the time-phase image to be corrected, and then perform radiometric correction on the time-phase image to be corrected through the global mapping relationship to obtain a preliminary corrected image; In addition, the processing module 202 in this application is also used to perform superpixel segmentation on the preliminary corrected image and the reference time-phase image, thereby determining the multidimensional similarity features of the preliminary corrected image and the reference time-phase image in each superpixel region; In addition, the processing module 202 in this application is also used to extract a stable region mask that has not undergone real ground object changes from all superpixel regions based on the multidimensional similarity features, and to perform radiation bias fitting on the global mapping relationship using the stable region mask to obtain the bias field of radiation brightness during remote sensing mapping. The execution module 203 in this application is mainly used to perform radiometric transformation on the phase image to be corrected through the bias field of the radiance, thereby generating a remote sensing mapping enhancement image that is consistent with the radiometric characteristics of the reference phase image.

[0060] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described UAV remote sensing mapping image enhancement method.

[0061] In some embodiments, reference Figure 5 This figure is an internal structural diagram of a computer device implementing a UAV remote sensing mapping image enhancement method according to some embodiments of this application. The UAV remote sensing mapping image enhancement method in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 300 includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0062] The processor 301 may be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the UAV remote sensing mapping image enhancement method in this application.

[0063] The communication bus 302 is used to transmit information between the aforementioned components.

[0064] Memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 303 may exist independently and be connected to processor 301 via communication bus 302. Memory 303 may also be integrated with processor 301.

[0065] The memory 303 stores program code for executing the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the UAV remote sensing mapping image enhancement method can be implemented by the processor 301 and one or more software modules in the program code in the memory 303.

[0066] Communication interface 304 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0067] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core processor or a multi-core processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0068] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device may be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0069] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described UAV remote sensing mapping image enhancement method.

[0070] In summary, the UAV remote sensing mapping image enhancement method and system disclosed in this application involves: acquiring a reference temporal image and a temporal image to be corrected for a target geographical area via a UAV platform; constructing a global mapping relationship of radiance between the two temporal images based on corresponding feature points in the reference and temporal images; performing radiometric correction on the temporal image to be corrected using the global mapping relationship to obtain a preliminary corrected image; performing superpixel segmentation on the preliminary corrected image and the reference temporal image to determine multidimensional similarity features between the preliminary corrected image and the reference temporal image in each superpixel region; extracting stable region masks from all superpixel regions based on the multidimensional similarity features, and using the stable region masks to perform radiometric bias fitting on the global mapping relationship to obtain a bias field of radiance during remote sensing mapping; performing radiometric transformation on the temporal image to be corrected using the radiance bias field to generate a remote sensing mapping enhancement image consistent with the radiometric characteristics of the reference temporal image; and reducing radiometric distortion caused by regional illumination differences in the enhancement image.

[0071] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0072] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for enhancing remote sensing images from unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: The baseline and phase images to be corrected of the target geographic area are collected using a drone platform. Based on the corresponding feature points in the reference time-phase image and the time-phase image to be corrected, a global mapping relationship of radiance between the two time-phase images is constructed. Then, the time-phase image to be corrected is radiometrically corrected through the global mapping relationship to obtain a preliminary corrected image. Superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to determine the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region; Based on the multidimensional similarity features, a stable region mask that has not undergone real ground object changes is extracted from all superpixel regions. The global mapping relationship is then fitted with the stable region mask to obtain the bias field of the radiance during remote sensing mapping. The radiometric transformation of the time-phase image to be corrected is performed by the bias field of the radiometric brightness, thereby generating a remote sensing mapping enhancement image with radiometric characteristics consistent with the reference time-phase image; Specifically, the method of performing radiative bias fitting on the global mapping relationship using the stable region mask to obtain the bias field of radiance during remote sensing mapping includes: Radiance sample data were collected within the masked area of ​​the stable region. Spatial radiance optimization is performed on the bias parameters in the global mapping relationship using the radiance sample data; The bias field of radiance during remote sensing mapping is generated based on the optimization results.

2. The method as described in claim 1, characterized in that, Constructing a global mapping relationship of radiance between the two temporal images based on corresponding feature points in the reference temporal image and the temporal image to be corrected specifically includes: Multiple feature points with the same name are extracted from the reference time-phase image and the time-phase image to be corrected; A radiative conversion model between the reference temporal image and the temporal image to be corrected is fitted by the radiance of all corresponding feature points; A global mapping relationship of radiance between two temporal images is constructed based on the aforementioned radiative conversion model.

3. The method as described in claim 1, characterized in that, The radiometric correction of the phase image to be corrected using the global mapping relationship to obtain the preliminary corrected image specifically includes: The gain coefficient and bias coefficient of radiance during remote sensing mapping are determined through the global mapping relationship. The radiance of each pixel in the phase image to be corrected is linearly transformed based on the gain coefficient and the bias coefficient. A preliminary corrected image is generated based on the linear transformation results.

4. The method as described in claim 1, characterized in that, Performing superpixel segmentation on the preliminary corrected image and the reference time-phase image, and then determining the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region, specifically includes: Superpixel segmentation is performed on the preliminary corrected image and the reference time-phase image to obtain a set of reference time-phase superpixel regions and a set of preliminary corrected superpixel regions. A spatial correspondence between superpixel regions is established based on the overlap of geographic coordinates between the regions in the reference temporal superpixel region set and the preliminary correction superpixel region set. The spatial correspondence is used to calculate the multidimensional similarity features between the preliminary corrected image and the reference temporal image in each corresponding superpixel region.

5. The method as described in claim 1, characterized in that, The stable region mask that does not undergo changes to real ground features, extracted from all superpixel regions based on the multidimensional similarity features, specifically includes: Based on the multidimensional similarity features, a comprehensive similarity score between the preliminary corrected image and the reference temporal image is determined in each superpixel region; Based on a preset similarity threshold and the comprehensive similarity score, the unchanged regions are extracted from all superpixel regions; A stable region mask that does not exhibit any actual changes in ground features is generated based on the unchanged region.

6. The method as described in claim 1, characterized in that, The process of performing a radiometric transformation on the time-phase image to be corrected using the radiance bias field to generate a remote sensing mapping enhancement image with radiometric characteristics consistent with the reference time-phase image specifically includes: The radiance bias field is combined with the global mapping relationship to construct a set of radiance correction parameters for image enhancement; The radiometric transformation is performed on each pixel in the phase image to be corrected using the radiometric correction parameter set. Based on the transformation results, a remote sensing mapping enhancement image consistent with the radiometric characteristics of the reference temporal image is generated.

7. A UAV remote sensing image enhancement system, which uses the method described in any one of claims 1 to 6 to enhance UAV remote sensing images, characterized in that, The system includes: The acquisition module is used to acquire baseline and phase images of the target geographic area via a drone platform. The processing module is used to construct a global mapping relationship of radiometric brightness between the reference time-phase image and the time-phase image to be corrected based on the same feature points in the reference time-phase image and the time-phase image to be corrected, and then perform radiometric correction on the time-phase image to be corrected through the global mapping relationship to obtain a preliminary corrected image; The processing module is further configured to perform superpixel segmentation on the preliminary corrected image and the reference time-phase image, thereby determining the multidimensional similarity features between the preliminary corrected image and the reference time-phase image in each superpixel region; The processing module is further configured to extract a stable region mask that has not undergone real ground object changes from all superpixel regions based on the multidimensional similarity features, and to perform radiative bias fitting on the global mapping relationship using the stable region mask to obtain the bias field of radiance during remote sensing mapping. The execution module is used to perform a radiometric transformation on the time-phase image to be corrected through the bias field of the radiometric brightness, thereby generating a remote sensing mapping enhancement image that is consistent with the radiometric characteristics of the reference time-phase image.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the UAV remote sensing mapping image enhancement method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the UAV remote sensing mapping image enhancement method as described in any one of claims 1 to 6.

Images