A method and system for dodging and burning of remote sensing imagery

Abstract

The application discloses a method and system for remote sensing image dodging and color uniformization, and relates to the technical field of remote sensing image processing. The method comprises the following steps: a template image containing various typical ground objects in a survey area is made, and statistical parameters of the template image in a Lab color space are calculated, the statistical parameters including mean values target_mean[] and standard deviations target_std[] of each channel; a digital orthographic image to be dodged and color-uniformized is grouped according to color consistency, and images with consistent colors and in a continuous piece are arranged in the same processing unit; for aerial images, the images are organized in units of flights and taken as processing units; for each processing unit, overall statistical parameters of the processing unit in the Lab color space are calculated based on sample images randomly selected, the overall statistical parameters including mean values source_mean[] and standard deviations source_std[] of each channel of all sample images in the processing unit; and dodging and color-uniformization conversion is performed on each image in the processing unit according to a preset formula.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and in particular to a method and system for uniform illumination and color matching of remote sensing images. Background Technology

[0002] Aerial imagery and digital orthophotos (DOM) are commonly used surveying and mapping geographic information products, widely applied in natural resource management, environmental monitoring, urban planning, and many other fields. In aerial remote sensing image processing, due to factors such as the type of sensor used in the original image data source, the timing of the event, weather conditions, or the quality of the image itself, serious quality problems arise in the acquired aerial imagery and the produced digital orthophotos, including poor color quality, significant color differences between different regions and batches of images, and inconsistent colors.

[0003] Because aerial images and digital orthophotos are numerous, have large data volumes, and exhibit significant color variations, traditional operating methods cannot meet the requirements for image color quality and production efficiency.

[0004] Solving the problems of color quality and color consistency in large-area aerial imagery and digital orthophotos has become an important and critical issue that the surveying and mapping geographic information industry urgently needs to address. Summary of the Invention

[0005] This application provides a method and system for uniform illumination and color matching of remote sensing images, which addresses the problems of color difference and large color aberration in existing remote sensing images, thereby overcoming the shortcomings of traditional operations in terms of quality and efficiency.

[0006] On the one hand, this application provides a method for uniform illumination and color matching of remote sensing images. The method includes the following steps: Step S1: Create a template image containing various typical land features in the survey area and calculate its statistical parameters in the Lab color space. The statistical parameters include the mean target_mean[] and standard deviation target_std[] of each channel; Step S2: Group the digital orthophoto images to be uniformly illuminated and colored according to their color consistency, and organize the images with consistent and contiguous colors into the same processing unit; for aerial images, they are organized by flight as a unit, which is used as a processing unit; Step S3: For each processing unit, calculate its overall statistical parameters in the Lab color space based on randomly selected sample images. The overall statistical parameters include the mean source_mean[] and standard deviation source_std[] of each channel of all sample images in the processing unit; Step S4: Perform uniform illumination and color matching transformation on each image in the processing unit according to a preset formula.

[0007] In one implementation of this application, step S1, the creation of the template image includes: selecting high-quality original aerial images with uniform tone and moderate brightness contrast, stitching them together into a rectangular image, and manually adjusting the colors to cover all typical features of the survey area.

[0008] In one implementation of this application, step S3, which calculates the overall statistical parameters of a randomly selected sample image in the Lab color space, specifically includes: step S31: traversing all images within the processing unit and randomly selecting a preset number of images as sample images; step S32: for each sample image, determining whether the number of pixels with a pixel value of (0,0,0) is less than a preset threshold; if it is less, then determining it as a valid sample image; step S33: based on all valid sample images, calculating the overall mean source_mean[] and overall standard deviation source_std[] of each channel of the processing unit in the Lab color space.

[0009] In one implementation of this application, in step S33, the method for calculating the overall mean value source_mean[c] of each channel of the processing unit in the Lab color space is: source_mean[c] = (the sum of pixel values ​​in channel c of all valid sample images) / (the total number of pixels in all valid sample images); that is, following the formula:

[0010] Where g is the overall mean of channel c, source_mean[c], M and N are the total number of rows and columns after the images of all valid samples are superimposed, respectively, and f(i,j) is the pixel value of the i-th row and j-th column in channel c.

[0011] In one implementation of this application, in step S3, the method for calculating the overall standard deviation source_std[c] of each channel of the processing unit in the Lab color space is as follows: calculate the sum of squares of the differences between each pixel value in channel c of all valid sample images and the overall mean source_mean[c], and obtain the sum of squares ss; then calculate ss / (the total number of pixels in all valid sample images), and perform a square root operation on the result to obtain source_std[c]; that is, following the formula:

[0012] Where K is the overall standard deviation of channel c, source_std[c], g is the overall mean of channel c, source_mean[c], M and N are the total number of rows and columns after the images of all valid samples are superimposed, and f(i,j) is the pixel value in the i-th row and j-th column of channel c.

[0013] In one implementation of this application, step S4 further includes processing the data-free area after color transformation: before performing color transformation, the pixel index of the source image with a pixel value of (0,0,0) is recorded; after the color transformation is completed, the pixel value of the corresponding position in the image after light and color equalization is reset to (0,0,0) according to the pixel index.

[0014] In one implementation of this application, step S4 is executed using a multi-threaded parallel processing method, including: creating a thread pool containing multiple worker threads; submitting the light and color equalization task for each image in the processing unit to the thread pool; and having the thread pool schedule the worker threads to execute multiple light and color equalization tasks in parallel.

[0015] In one implementation of this application, the method further includes step S5, template color adjustment: after step S4, the effect of the image after uniform lighting and color is evaluated. If the preset requirements are not met, the process returns to step S1, and at least one parameter of the template image, such as hue, brightness, saturation, color balance, shadow / highlight, exposure, or contrast, is adjusted, and subsequent steps S1-S4 are re-executed.

[0016] In one implementation of this application, in step S4, when the object being processed is a digital orthophoto, the corresponding coordinate positioning file is read or copied simultaneously when reading the image file.

[0017] On the other hand, this application also provides a system for uniform illumination and color matching of remote sensing images. The system includes: a template processing module for creating a template image and calculating its Lab color space statistical parameters; an image processing module for processing the image to be processed into different processing units; a parameter calculation module for calculating the overall Lab color space statistical parameters of each processing unit; and a uniform illumination and color matching execution module for performing color transformation on the image within the processing unit based on the template image statistical parameters and the overall statistical parameters of the processing unit.

[0018] The method and system for homogenizing and color-matching remote sensing images provided in this application have the following beneficial effects: 1. This scheme produces template images covering typical features in the survey area and calculates Lab color space statistical parameters. It also groups the images according to color consistency, ensuring the pertinence of uniform lighting and color from both the perspective of benchmark establishment and image classification. It can effectively solve the problems of large color difference and inconsistent color in remote sensing images of different regions and batches, and make the image color match the color characteristics of the template image.

[0019] 2. By randomly selecting sample images and filtering out valid samples, the mean and standard deviation of the Lab color space of the processing unit are calculated using a standardized formula. This eliminates the interference of invalid samples and makes the statistical parameters more consistent with the true color characteristics of the processing unit, providing an accurate and reliable data foundation for subsequent uniform light and color transformation.

[0020] 3. Balancing processing quality and efficiency, it restores pixel values ​​in data-free areas after light and color equalization to ensure image data integrity, while also employing multi-threaded parallel processing to improve work efficiency. It also supports iterative adjustment of template colors, allowing for parameter optimization and reprocessing based on the desired effect, ensuring that the final image meets the preset color quality requirements. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a method for homogenizing light and color in remote sensing images provided in this application embodiment; Figure 2 An effect diagram provided in this application embodiment for use in the uniform illumination and color matching of remote sensing images; Figure 3 An image showing the effect of homogenizing and color-matching remote sensing images, provided as an embodiment of this application; Figure 4 This is a system composition diagram for uniform illumination and color matching of remote sensing images, provided as an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] This application provides a method and system for uniform illumination and color matching of remote sensing images. The technical solution proposed in this application will be described in detail below with reference to the accompanying drawings. The main technical process for uniform illumination and color matching of images includes: template image creation, image processing, template image parameter calculation, image parameter calculation, and uniform illumination and color matching of the image. A flowchart is attached. Figure 1 As shown, the specific process is as follows: 1. Template image creation.

[0024] This method uses template imagery as the target template for image homogenization and color matching. Selecting a suitable template image is crucial for successful homogenization and color matching. Unprocessed raw aerial imagery possesses rich color information and is more suitable for template image creation than processed digital orthophotos. High-quality aerial imagery should be selected as the data source for template image creation. The selected imagery should have uniform and soft tones, moderate brightness and contrast, and good performance in hue, lightness, saturation, and other aspects. The terrain features within the template should cover as much of the survey area as possible, including buildings representing local architectural colors in urban and rural residential areas, as well as industrial and mining enterprises, colorful amusement parks, water systems, forests, farmland, etc. The proportions of various colors in the template should be close to the color proportions in the images of the survey area.

[0025] After selecting the aerial images, use Photoshop to stitch them together to form a rectangular template image. The template image must not contain any areas without data. After stitching, experienced professionals need to adjust the color of the template data to make the various indicators of the template image more scientifically accurate.

[0026] 2. Image organization.

[0027] Aerial imagery from different sorties uses the same sensors and has consistent timeliness, so it can be organized by sortie and does not require further processing. However, digital orthophotos exhibit significant color variations within the mission area due to different sensors and time phases. Before homogenization, the digital orthophotos should be processed. First, all orthophotos should be mosaicked to generate low-resolution orthophotos and overlaid with an orthophoto result combination table. Then, based on the color characteristics of the low-resolution orthophotos and according to the orthophoto result combination table, digital orthophotos with consistent and contiguous colors should be organized into the same folder.

[0028] 3. Template image parameter calculation.

[0029] Template image data reading. The Pillow (PIL) library is an image processing library supported by Python. Compared with other libraries, its advantage is that it can read image results in all TIFF formats. The Pillow library is used to read the template image, and the cvtColor method of the OpenCV library is used to convert the read template image into Lab mode. In this mode, the L channel represents the brightness or lightness value, ranging from black to white; the a channel represents the range from green (negative value) to red (positive value); and the b channel represents the range from blue (negative value) to yellow (positive value).

[0030] The mean_mean[] and standard deviation_target_std[] of the three channels L, a, and b of the template image are calculated using the meanStdDev method of the OpenCV library. Both target_mean[] and target_std[] are one-dimensional arrays that can store the mean and standard deviation of the three bands of the image.

[0031] Mean: This is the average value of all pixel values ​​in each channel, representing the average brightness or intensity of the image in the color channels. Its formula is shown in Formula 1:

[0032] in, Let M be the average pixel value of a certain channel, and M and N be the number of rows and columns of the image. Let be the pixel value in the i-th row and j-th column of a certain image channel.

[0033] Standard deviation: The standard deviation is the square root of the mean of the sum of the squares of the differences between all pixel values ​​in each channel of an image and the mean of the corresponding channel. It reflects the dispersion of pixel values ​​from the mean in an image. The larger the standard deviation, the more dispersed the pixel values ​​are, and the richer the image's tonal and color information. The formula for image standard deviation is shown in Formula 2:

[0034] Where k is the standard deviation of pixel values ​​in a certain color channel of the image, and other variables are shown in Formula 1.

[0035] 4. Image parameter calculation.

[0036] The mean and standard deviation of an image are crucial parameters for image homogenization and color matching. Different images have different means and standard deviations. If homogenization and color matching are performed separately for each image sheet, color differences will occur between digital orthophoto sheets, resulting in inconsistent colors. Furthermore, for aerial images of large areas of water damage, the excessive deviation in the mean and standard deviation of individual images can lead to color distortion after homogenization and color matching. To avoid these problems, the image folders prepared in step 2 should be used as homogenization and color matching units. The mean and standard deviation of each image folder should be calculated separately, and then these values ​​should be used as uniform homogenization and color matching parameters for the entire folder.

[0037] The mean and standard deviation parameters are calculated using a folder as a whole. Considering that there may be tens of thousands of images in a folder, it would be very time-consuming to include all of them in the calculation. Therefore, an appropriate number of images in the folder can be randomly selected as sample images for the calculation of mean and standard deviation. The sample size should be moderate. Too many images will affect the calculation speed, while too few images will make it impossible to accurately calculate the mean and standard deviation parameters of the images in the folder. When there are few images in the folder, all images can be selected as sample images for the calculation.

[0038] The main steps are as follows: (1) Get all image file names with the extension .tif in the specified folder. Randomly select images according to the specified number of samples using the random.choice method provided by the numpy library, and put the selected images into the specified sample image file list file_list. If no sample number is specified, all images in the folder are put into the sample image file list file_list as sample images.

[0039] (2) Use the Pillow library to traverse and read all images in the sample image file list file_list, and then calculate the sample image parameters by folder.

[0040] (3) Considering that some incomplete images contain a large number of black pixels with a value of (0, 0, 0) (no data area), too many of these pixels will affect the statistical accuracy of the uniform lighting and color matching parameters, resulting in color cast in the image after uniform lighting and color matching. Therefore, when calculating the uniform lighting and color matching parameters of the sample image, the index value of the (0, 0, 0) value pixels in the image should be calculated first, and then the sum method of the NumPy library should be used to calculate the number of (0, 0, 0) value pixels based on the index. If the number of (0, 0, 0) value pixels in the image is less than the threshold (1000), it can be used as the final sample image to participate in the calculation of uniform lighting and color matching parameters; otherwise, it will not be used as a sample image to participate in the calculation of uniform lighting and color matching parameters.

[0041] (4) Convert the read single sample image into Lab mode and calculate the number of pixels in the sample image. Then calculate the mean of each channel of the sample image using the meanStdDev method of the OpenCV library.

[0042] (5) Calculation of the mean value of the sample image. Multiply the mean value of each channel of a single sample image obtained in (4) by the number of pixels in the image to obtain the total value of each channel of the sample image. Repeat steps (3)-(5) to sum the total value of each channel and the number of pixels of all sample images respectively. Divide the total value of each channel of all sample images by the number of pixels of all sample images to obtain the mean value of each channel of all sample images in the folder. source_mean[] is a one-dimensional array that can store the mean values ​​of the three bands of the image.

[0043] (6) Calculation of standard deviation of sample images. After obtaining the mean value source_mean[] of each channel of all sample images in the folder through step (5), iterate through all images in the sample image file list file_list again, and determine the sample images that meet the conditions according to the requirements of step (3). Calculate the sum of squares of the differences between each pixel value f(i,j) of each channel of a single sample image and the mean value source_mean[] of each channel of all sample images in the folder. Accumulate the sum of squares of all pixels in all sample images to obtain ss, which is the sum of squares of the differences between each pixel value of each channel of all sample images in the folder and the mean value source_mean[] of each channel of the sample images in the folder. Divide ss by the sum of the number of pixels of all sample images in the folder, and then perform a square root operation to obtain the standard deviation source_std[] of each channel of the sample images in the folder.

[0044] 5. Image uniform lighting and color.

[0045] Convert the source image to Lab mode. Taking the L channel as an example, its light and color homogenization is as shown in Formula 3: result_lab[:,:,0]=(source_lab[:,:,0]-source_mean[0])*(target_std[0] / source_std[0])+target_mean[0].

[0046] Where, result_lab[:,:,0] is the value of the L channel of the target image (after light and color homogenization); source_lab[:,:,0] is the value of the L channel of the source image data (before light and color homogenization); source_mean[0] is the mean of the L channels of all sample images in the folder; target_std[0] is the standard deviation of the L channels of the template image; source_std[0] is the standard deviation of the L channels of the sample images in the folder; target_mean[0] is the mean of the L channels of the template image. The light and color homogenization of channels a and b is the same as above, only 0 needs to be changed to 1 and 2, 1 is channel a and 2 is channel b.

[0047] After all channels are uniformly illuminated and colored, the cvtColor method of the OpenCV library is used to convert the Lab mode of the image into an RGB channel mode that meets the specifications. In actual production operations, some incomplete images have no-data areas with pixel values ​​of (0, 0, 0) black pixels. After uniform illumination and color matching, these pixel values ​​will change and will no longer be (0, 0, 0), which does not meet the specifications. Therefore, before uniform illumination and color matching, the index of the (0, 0, 0) black pixel value in the image should be calculated. After uniform illumination and color matching, the corresponding pixel value is replaced with (0, 0, 0) according to the index to ensure that the image conforms to the relevant specifications after uniform illumination and color matching. Through the above uniform illumination and color matching method, the brightness and color of the source image are made to be consistent with the template image, thereby achieving uniform illumination and color matching of the image. Since the digital orthophoto image format is generally TIF+TFW format, where the image is a TIF format file and the TFW is a coordinate positioning file, both are indispensable. When traversing and reading all digital orthophoto images in the folder, the corresponding TFW file with the same name should be copied to the target folder.

[0048] Finally, the `save` method provided by the Pillow (PIL) library is used to output the homogenized and color-matched images as required. This method allows setting the compression method, output quality, and data format of the output image. To ensure consistent data size, digital orthophotos should be output in an uncompressed format. During image output, the image organization should be consistent with the source image to reduce the workload of manual copying.

[0049] 6. Improved efficiency in image homogenization and color matching.

[0050] Due to the large volume and quantity of image data, the efficiency of image homogenization is a bottleneck problem. Since image homogenization involves multiple tasks that do not interfere with each other, multi-threading technology can be used to fully utilize computer CPU resources and improve the efficiency of image homogenization.

[0051] The implementation is as follows: First, import the concurrent.futures parallel management module; second, create a thread pool executor (ThreadPoolExecutor) and set the maximum number of threads in the thread pool, i.e., the number of tasks the computer can execute in parallel; third, traverse all images in the specified folder (including subfolders), create a light and color balancing task for each image, and submit all light and color balancing tasks to the thread pool executor; finally, the thread pool executor processes the light and color balancing tasks in batches according to the submission order and the set maximum number of threads. The maximum number of threads in the thread pool should not exceed the number of CPU cores. It can be reasonably set according to CPU parameters, computer memory size, and image data size. Setting the maximum number of threads too high may exhaust computer memory resources and cause a system crash. Using multi-threading technology can improve the efficiency of light and color balancing by more than ten times, effectively solving the bottleneck problem of image light and color balancing efficiency.

[0052] 7. Template color adjustment.

[0053] During the image homogenization and color matching process, a small number of images should be selected first to conduct homogenization and color matching experiments using template images. If the homogenization and color matching do not achieve the desired effect, manual color adjustments need to be made to the template image. For example, if a certain color in the output image does not meet the requirements, Photoshop can be used to adjust parameters such as hue, brightness, saturation, color balance, and vibrance of that color in the template image. If the output image has shadows or highlights that make the image texture unclear, shadow / highlight adjustments need to be made to the template image. If the output image has overexposure or excessive contrast, exposure and contrast adjustments need to be made to the template image. After adjusting the template colors, the experiment should be repeated until the resulting image meets the requirements before large-scale image homogenization and color matching can be performed.

[0054] Taking digital orthophoto homogenization as an example, the image quality and color consistency before and after homogenization are shown in the attached figure. Figure 2 Appendix Figure 3 As shown.

[0055] The above describes a method for homogenizing and color-matching remote sensing images according to embodiments of this application. Based on the same inventive concept, embodiments of this application also provide a system for homogenizing and color-matching remote sensing images. Figure 4 A system composition diagram for uniform illumination and color matching of remote sensing images is provided in the embodiments of this application, as follows: Figure 4As shown, the system mainly includes: a template processing module 401, used to create a template image and calculate its Lab color space statistical parameters; an image processing module 402, used to process the image to be processed into different processing units; a parameter calculation module 403, used to calculate the overall Lab color space statistical parameters of each processing unit; and a uniform light and color execution module 404, used to perform color transformation on the image within the processing unit based on the template image statistical parameters and the overall statistical parameters of the processing unit.

[0056] This application provides a method and system for uniform illumination and color matching of remote sensing images. The method and system, based on Lab statistical information matching technology, solves the problem of uniform illumination and color matching for large-area, multi-temporal, and multi-source aerial and digital orthorectified images. This method and system can be applied to uniform illumination and color matching of raw aerial images and digital orthorectified images produced by different work units, in different batches, and using different sensors. It can solve the color consistency problem of aerial and digital orthorectified images with different colors, greatly improving the color quality of the image results. This method and system have high speed and efficiency in uniform illumination and color matching; the efficiency of uniform illumination and color matching for large aerial images by a single computer node reaches 800 images / hour, which is four times that of mainstream equipment in China; the efficiency of uniform illumination and color matching for 1:2000 standard sheet digital orthorectified images reaches up to 3000 images / hour. The system also features one-click operation, full automation, and no manual intervention required.

[0057] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. 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; relevant parts can be referred to the descriptions of the method embodiments.

[0058] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0059] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A method for uniform illumination and color matching in remote sensing images, characterized in that, The method includes the following steps: Step S1: Create a template image containing various typical land features in the survey area and calculate its statistical parameters in the Lab color space. The statistical parameters include the mean target_mean[] and standard deviation target_std[] of each channel. Step S2: The digital orthophotos to be uniformly illuminated and colored are grouped according to their color consistency, and images with consistent and contiguous colors are organized into the same processing unit; for aerial images, they are organized by sortie as processing units. Step S3: For each processing unit, based on randomly selected sample images, calculate its overall statistical parameters in the Lab color space. The overall statistical parameters include the mean source_mean[] and standard deviation source_std[] of each channel of all sample images in the processing unit. Step S4: For each image in the processing unit, perform uniform light and color transformation according to a preset formula.

2. The method for uniform illumination and color matching of remote sensing images according to claim 1, characterized in that, In step S1, creating the template image includes: selecting high-quality original aerial images with uniform tone and moderate brightness contrast, stitching them together into a rectangular image, and manually adjusting the colors to cover all typical features in the survey area.

3. The method for uniform illumination and color matching of remote sensing images according to claim 1, characterized in that, Step S3, which involves calculating the overall statistical parameters of randomly selected sample images in the Lab color space, specifically includes: Step S31: Traverse all images within the processing unit and randomly select a preset number of images as sample images; Step S32: For each sample image, determine whether the number of pixels with a pixel value of (0,0,0) is less than a preset threshold. If it is less, then determine it as a valid sample image. Step S33: Based on all valid sample images, calculate the overall mean source_mean[] and overall standard deviation source_std[] of each channel of the processing unit in the Lab color space.

4. The method for uniform illumination and color matching of remote sensing images according to claim 3, characterized in that, In step S33, the method for calculating the overall mean value source_mean[c] of each channel of the processing unit in the Lab color space follows the formula: Where g is the overall mean of channel c, source_mean[c], M and N are the total number of rows and columns after the images of all valid samples are superimposed, respectively, and f(i,j) is the pixel value of the i-th row and j-th column in channel c.

5. A method for homogenizing light and color in remote sensing images according to claim 3, characterized in that, In step S3, the method for calculating the overall standard deviation source_std[c] of each channel of the processing unit in the Lab color space is as follows: Calculate the sum of squares of the differences between each pixel value in channel c of all valid sample images and the overall mean source_mean[c], and obtain the total sum of squares ss; then calculate ss divided by the total number of pixels in all valid sample images, and perform a square root operation on the result to obtain source_std[c]; That is, follow the formula: Where K is the overall standard deviation of channel c, source_std[c], g is the overall mean of channel c, source_mean[c], M and N are the total number of rows and columns after the images of all valid samples are superimposed, and f(i,j) is the pixel value in the i-th row and j-th column of channel c.

6. The method for uniform illumination and color matching of remote sensing images according to claim 1, characterized in that, Step S4 also includes processing the data-free area after color transformation: before performing color transformation, the pixel index of the source image with a pixel value of (0,0,0) is recorded; after the color transformation is completed, the pixel value of the corresponding position in the image after light and color equalization is reset to (0,0,0) according to the pixel index.

7. The method for uniform illumination and color matching of remote sensing images according to claim 1, characterized in that, Step S4 is executed using a multi-threaded parallel processing method, including: Create a thread pool containing multiple worker threads; The task of uniform lighting and color matching for each image in the processing unit is submitted to the thread pool. The thread pool schedules the worker threads to execute multiple light and color equalization tasks in parallel.

8. A method for homogenizing light and color in remote sensing images according to claim 1, characterized in that, The method further includes step S5, template color adjustment: after step S4, the effect of the image after uniform lighting and color is evaluated. If the preset requirements are not met, the method returns to step S1 and adjusts at least one parameter of the template image, such as hue, brightness, saturation, color balance, shadow / highlight, exposure, or contrast, and then re-executes subsequent steps S1-S4.

9. A method for homogenizing light and color in remote sensing images according to claim 1, characterized in that, In step S4, the light and color homogenization follows the following formula: result_lab[:,:,c]=(source_lab[:,:,c]-source_mean[c])*(target_std[c] / source_std[c])+target_mean[c] Where c is 0, 1 or 2, representing the L, a, and b channels in the Lab color space respectively; source_lab[:,:,c] is the pixel value matrix of the source image in channel c, and result_lab[:,:,c] is the pixel value matrix of the image after light and color homogenization in channel c.

10. A system for uniform illumination and color matching of remote sensing images, characterized in that, The system includes: The template processing module is used to create template images and calculate their Lab color space statistical parameters; The image processing module is used to organize the images to be processed into different processing units; The parameter calculation module is used to calculate the overall statistical parameters of the Lab color space for each of the processing units; The uniform light and color execution module is used to perform color transformation on the image within the processing unit based on the statistical parameters of the template image and the overall statistical parameters of the processing unit.

Images