CN-121982118-A - Multi-scale remote sensing image color homogenizing method based on geographic position matching

CN121982118ACN 121982118 ACN121982118 ACN 121982118ACN-121982118-A

Abstract

The invention relates to the technical field of remote sensing image color homogenizing methods, in particular to a multi-scale remote sensing image color homogenizing method based on geographic position matching, which comprises the steps of acquiring satellite images covering an image area to be color homogenized, and processing the satellite images to form a reference image; the method comprises the steps of carrying out normalization processing on an image to be leveled, carrying out geographic position matching on the image to be leveled and a reference image to obtain an image in a corresponding coordinate range of the image to be leveled on the reference image, marking the image as a reference POI image, converting a time domain of the reference POI image and the image to be leveled into a frequency domain, carrying out three-level multi-scale decomposition on the reference POI image and the image to be leveled, calculating color feature vectors on the reference POI image, constructing a color mapping relation, carrying out color mapping on the image to be leveled, and restoring the image after being leveled by adopting a multi-scale restoration strategy.

Inventors

Yong Baohu
LAN YUANGE
YANG XIAFANG
WANG YUAN
CHEN SHAOLI
ZHOU XIN

Assignees

洞察时空(成都)科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260205

Claims (10)

1. A multi-scale remote sensing image color homogenizing method based on geographic position matching is characterized by comprising the following steps: Acquiring satellite images covering an image area to be leveled, carrying out dodging treatment and orthographic mosaic splicing treatment on each satellite image, and carrying out smoothing treatment on chromatic aberration at adjacent positions of a scene to form a reference image with consistent tone and full color; carrying out normalization processing including gray level normalization and spectrum normalization on the image to be uniformly colored; constructing a coordinate conversion relation, and realizing the geographic position matching of the image to be leveled and the reference image to obtain an image of the image to be leveled in a corresponding coordinate range on the reference image, wherein the image is marked as a reference POI image; Converting the reference POI image and the image to be leveled from a time domain to a frequency domain; performing three-level multi-scale decomposition on the reference POI image and the image to be uniformly colored in a frequency domain, wherein the first-level decomposition is performed based on original pictures, and the second-level and third-level classification is performed on low-frequency components of the previous-level decomposition; calculating a color feature vector for the reference POI image; Based on the color feature vector of the reference POI image, constructing a color mapping relation between the image to be leveled and the reference base image, and performing color mapping on the image to be leveled; based on the color mapping result and the signal decomposition result, adopting a multi-scale restoration strategy to reconstruct step by step from bottom to top, and restoring the image with consistent color tone and rich details after color homogenization.
2. The method for homogenizing a multi-scale remote sensing image based on geographic position matching as claimed in claim 1, wherein the specific steps of obtaining satellite images covering an image area to be homogenized, homogenizing each satellite image, performing orthomosaic stitching, smoothing chromatic aberration at a scene adjacent part, and forming a reference image with consistent hue and full color comprise the following steps: Acquiring a primary remote sensing image of which the multiple views cover an image area to be uniformly colored are subjected to orthorectification processing, and performing format conversion, digit conversion, unification of coordinate references, adjustment of spatial resolution and unification of dimension; carrying out uniform light treatment on each scene image to ensure that the internal tone of a single image is consistent and the brightness is moderate; Mosaic and splicing are carried out on the images subjected to the uniform light treatment according to geographic positions; and setting the pixel width of a transition region of adjacent scene splicing, and eliminating the splicing chromatic aberration of the adjacent scene by using an eclosion mode to form a reference base image with consistent hue, moderate brightness and geographic reference information.
3. The method for homogenizing a multi-scale remote sensing image based on geographic location matching according to claim 1, wherein in the step of performing normalization processing including gray scale normalization and spectrum normalization on the image to be homogenized, the specific step of gray scale normalization includes: Selecting the gray level of a reference image as a reference gray level, and converting the gray level of the image to be leveled into the range of the gray level of the reference image, wherein the gray level normalization formula is as follows: In the formula, 、、、 The method comprises the steps of respectively obtaining the maximum and minimum gray level of an image to be normalized and the maximum and minimum gray level of a reference image, wherein SRC is a DN value before normalization, and DST is a DN value after normalization.
4. The method for homogenizing a multi-scale remote sensing image based on geographic location matching according to claim 1, wherein in the step of performing normalization processing including gray scale normalization and spectrum normalization on the image to be homogenized, the specific step of spectrum normalization includes: By calculating the spectrum matching factor, the radiation difference between images of the same radiation quantity on different sensors caused by the spectrum response difference of different remote sensors is eliminated, and the calculation formula of the spectrum matching factor is as follows: wherein: 、 For the normalized spectral response functions of the image to be leveled and the reference image, 、、、 Respectively is 、 Upper and lower bounds of the spectral effective range; the formula for image correction by using the spectrum matching factor is as follows: where DN is the DN value of the image before correction, The DN value of the corrected image is obtained.
5. The method for homogenizing a multi-scale remote sensing image based on geographic position matching according to claim 1, wherein the specific steps of constructing a coordinate transformation relationship to realize geographic position matching of the image to be homogenized and the reference image and obtain an image of the image to be homogenized in a corresponding coordinate range on the reference image, and marking the image as a reference POI image include: analyzing geographic reference information of the reference image and the image to be leveled, and respectively obtaining ellipsoid references and affine six parameters of the reference image and the image to be leveled; selecting four corner points on the image to be leveled as image space control points; Calculating object space coordinates of four corner points under an ellipsoid reference of the image to be leveled based on four corner point image space coordinates of the image to be leveled and affine six parameters of the image to be leveled; Converting the object space coordinates of four corners under the ellipsoid reference of the image to be uniformly colored into the object space coordinates under the ellipsoid reference corresponding to the reference image based on the Boolean seven parameters; Based on affine six parameters of the reference image, converting the object side coordinates under the ellipsoidal reference corresponding to the converted reference image into image side coordinates of the reference image, wherein the image side coordinates are the positions of the images to be uniformly colored on the reference image, and acquiring row and column information through coordinate matching and then acquiring POI images through position cutting.
6. The method for homogenizing a multi-scale remote sensing image based on geographic location matching as set forth in claim 5, wherein in the step of calculating the object space coordinates of the four corners under the ellipsoidal reference of the image to be homogenized based on the four corners of the image to be homogenized and affine six parameters thereof, the calculation formula is: in the formula, GT [0], GT [1], GT [2], GT [3], GT [4], GT [5] are affine transformation six parameters, col, row are image space coordinates on the image, xgeo, ygeo are object space coordinates corresponding to the image space coordinates.
7. The method for homogenizing a multi-scale remote sensing image based on geographic location matching as set forth in claim 5, wherein in the step of converting the object coordinates of four corners under the ellipsoid reference of the image to be homogenized to the object coordinates under the ellipsoid reference corresponding to the reference image based on the boolean seven parameters, the formula of the coordinate conversion is as follows: In the formula, [ X1Y 1Z 1], [ X2Y 2Z 2] are the coordinates of the same point under two coordinate systems respectively, and [ X0Y 0Z 0] is a translation parameter between the two coordinate systems, Is a rotation parameter between the two coordinate systems and m is a scaling parameter.
8. The method for homogenizing a multi-scale remote sensing image based on geographic location matching as claimed in claim 1, wherein the specific step of converting the reference POI image and the image to be homogenized from the time domain to the frequency domain comprises: And converting the reference POI image and the image to be leveled from a time domain to a frequency domain by using two-dimensional Fourier transformation, and carrying out frequency spectrum normalization and dynamic range compression.
9. The method for homogenizing a multi-scale remote sensing image based on geographic location matching as claimed in claim 1, wherein the reference POI image and the image to be homogenized are subjected to three-level multi-scale decomposition in a frequency domain, the first-level decomposition is performed based on an original image, the second-level and third-level classification is performed on low-frequency components of the previous-level decomposition, The downsampling technique is adopted for the decomposition of each stage of the image to be uniformly colored, namely the size of the decomposition result of the next stage is only 1/2 of that of the previous stage, namely the size of the three-stage decomposition result is only 1/8 of that of the original image.
10. The method for homogenizing a multi-scale remote sensing image based on geographic position matching as claimed in claim 1, wherein the specific steps of adopting a multi-scale restoration strategy to reconstruct step by step from bottom to top based on the color mapping result and the signal decomposition result to restore an image with consistent hue and rich details after homogenizing the color comprise the following steps: performing three-level signal decomposition on the reference image, performing SVD decomposition on a low-frequency component Ls3 obtained by the third-level decomposition to obtain a feature vector UV and a feature value E, and further calculating a color migration matrix T; Performing color mapping on the third-stage low-frequency component L3 of the image to be uniformly colored based on the color migration matrix T to obtain a third-stage low-frequency correction result ; High-frequency component H3 obtained by third-stage decomposition of image to be uniformly-colored and low-frequency correction result obtained after color mapping Combining to obtain Then comparing the low-frequency component L2 with the low-frequency component L2 of the second layer of the image to be leveled to obtain a detail reduction factor of the third layer; Obtaining a restoration result of the third layer based on the color mapping result, the detail restoration factor of the third layer and the high-frequency component obtained by the decomposition of the third layer ; Upsampling the restored image I3 of the third layer to obtain a restored result of the low-frequency component of the second layer ; Combining the high-frequency component H2 obtained by decomposing the second layer with the low-frequency correction result obtained by recovering the second layer Combining to obtain Then comparing with the low-frequency component L1 of the first layer to obtain a detail reduction factor of the second layer; Recovery effort based on low frequency components The detail reduction factor of the second layer and the high frequency component obtained by the decomposition of the second layer to obtain the restoration result of the second layer ; Results of the second layer recovery Upsampling to obtain the recovery result of the first layer low frequency component ; Combining the high frequency component H1 obtained by the first layer decomposition with the low frequency correction result obtained by the second layer restoration Combining to obtain Comparing the first layer of detail reduction factors with the image I to be leveled to obtain the detail reduction factors of the first layer; Obtaining a restoration result of the first layer based on the restoration result of the low frequency component and the high frequency component obtained by the first layer decomposition ; Results of the first layer recovery Up-sampling to obtain low-frequency component result under original image size ; Calculating a high-frequency component H0 of the image to be uniformly colored under the original size based on the low-frequency component restored by the first layer under the original size; high frequency component based on original image size And restored low frequency component And combining to obtain the uniform-color result.

Description

Multi-scale remote sensing image color homogenizing method based on geographic position matching Technical Field The invention relates to the technical field of remote sensing image color homogenizing methods, in particular to a multi-scale remote sensing image color homogenizing method based on geographic position matching. Background The remote sensing image obtained by the satellite sensor is often influenced by external factors such as the characteristics of the sensor, human operation errors, weather conditions and the like, and cannot be directly applied to an actual scene. The image is required to be subjected to systematic correction processing at the ground end, and the key links comprise radiation correction, geometric correction, splicing and embedding, color homogenizing processing and the like. The color homogenizing treatment is used as a core difficulty of image post-treatment, and is faced with multiple challenges such as color deviation caused by remote sensor difference, brightness change caused by weather mutation, feature fluctuation caused by seasonal change, feature transition caused by human factors and inter-image tone difference in different time phases. In addition, the uncertainty introduced by subjective factors further exacerbates the processing difficulty. The interleaving of the objective factors and the subjective factors makes the color homogenizing treatment a bottleneck problem which restricts the large-scale application of the remote sensing image. The color homogenizing technology (also called color correction) is a key link in remote sensing image processing, and is mainly used for solving the problems of image color distortion and uneven brightness. The hue, the darkness and the contrast of the image are optimally adjusted through a professional algorithm, so that the processed remote sensing image presents a visual effect of coordinated overall hue, uniform brightness distribution and moderate contrast, the visual expressive force of the image is remarkably improved, a 'one-picture' result with consistent hue and seamless connection is formed, and the high-precision and high-quality image data support is provided for large-range application. The traditional color homogenizing method comprises a man-machine interaction processing method, a template processing method, a dodging method and a histogram method. The template processing method is characterized in that an image with better color is selected as a color reference standard to carry out color homogenization, the template processing method can obtain better effect on a similar region of a landform, but the image with large landform difference is not ideal in color homogenization effect, the problem that the ground objects are not matched exists, namely the template processing method cannot be applied to wide-range color homogenization processing, meanwhile, the color homogenization effect difference obtained by selecting different reference images is also large, the problem that unified judgment standards are lacking exists, and the light homogenization method is characterized in that the MASK method is adopted, the color consistency inside a single image can be solved, and color difference residues still exist at a multi-view splicing position. The histogram method is to count the mean value, variance, histogram and other information of the reference image and the image to be leveled, and to realize the leveling by means of mean value and variance mapping, histogram mapping and the like, and the histogram method can obtain better leveling effect under most conditions, but also has the problem of detail blurring and distortion. Therefore, how to realize the color uniformity process with effective color adjustment and abundant details is a current urgent problem to be solved. Disclosure of Invention The invention aims to provide a multi-scale remote sensing image color homogenizing method based on geographic position matching, which can realize effective color adjustment and color homogenizing correction processing with rich details on a remote sensing image. In order to achieve the above purpose, the invention provides a multi-scale remote sensing image color homogenizing method based on geographic position matching, which comprises the following steps: Acquiring satellite images covering an image area to be leveled, carrying out dodging treatment and orthographic mosaic splicing treatment on each satellite image, and carrying out smoothing treatment on chromatic aberration at adjacent positions of a scene to form a reference image with consistent tone and full color; carrying out normalization processing including gray level normalization and spectrum normalization on the image to be uniformly colored; constructing a coordinate conversion relation, and realizing the geographic position matching of the image to be leveled and the reference image to obtain an image of the image to be leveled in a corresponding coor