Quality-graded remote sensing image composite cloud removal method, system and equipment

Abstract

The invention discloses a quality-graded remote sensing image composite cloud removal method, system and device. The method comprises the following steps: preprocessing an acquired remote sensing image; based on the preprocessed data, thin cloud radiation correction is carried out; based on the data after thin cloud radiation correction, thick cloud detection is carried out; based on the data after thick cloud detection, thick cloud missing reconstruction is carried out; performing atmospheric correction based on the data after the thick cloud missing reconstruction; and performing data quality grading based on the data after atmospheric correction. According to the method, a series of standardized processing is realized on data acquired under different space and time conditions, the influence of atmospheric radiation attenuation is eliminated, and the data are formatted into surface reflection data with consistent observation characteristics, so that the utilization rate of remote sensing images is improved, and effective comparison and analysis of subsequent application research are promoted.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing technology, and specifically relates to a quality-graded remote sensing image composite cloud removal process, which can provide a high-quality data basis for remote sensing application research. Background Art

[0002] Remote sensing is an important means of detecting information on the earth's surface. It is widely used in resource and environmental surveys, space exploration, communications, navigation, meteorology, surveying and mapping, military reconnaissance and other fields. Its importance to the effective utilization of earth resources and sustainable development is growing.

[0003] As the main medium for radiation transmission, the state of the atmosphere has a very important and non-negligible impact on the remote sensing imaging process. When the atmosphere is clean, the surface radiation can be effectively recorded by the sensor to generate high-quality image data; however, when the atmosphere is turbid, suspended particles of different sizes scatter radiation, hindering the radiation transmission process, resulting in different degrees of cloud obscuration in the acquired data. According to the degree of scattering of radiation by atmospheric suspended particles, the cloud obscuration problem in remote sensing images can be divided into thin clouds and thick clouds. For the former, the surface reflected radiation is partially scattered and the radiation transmission is partially weakened. At this time, the sensor includes additional cloud and fog information in addition to recording the surface information, which greatly reduces the accuracy of the observation; while for the latter, the suspended particles in the atmosphere completely scatter the surface reflected radiation, interrupting the radiation transmission path, resulting in the complete loss of surface information. The above adverse effects make it difficult for cloud and fog optical data to meet the subsequent application requirements such as feature extraction, target detection, object recognition, and quantitative analysis, which seriously reduces the accuracy of remote sensing interpretation and limits the potential for data application.

[0004] Existing cloud removal methods generally only focus on a certain type of degradation problem, but lack research on the overall process of composite cloud removal. Existing studies usually distinguish between thin clouds and thick clouds and develop multiple types of correction methods. Removal methods can be classified into multiple categories according to their technical means, data requirements, and scope of application.

[0005] Thin cloud radiation correction methods can be roughly divided into three categories. 1) Methods based on radiation transfer models. They all rely on physical prior knowledge and integrate factors such as the atmosphere, surface, and sensors to establish models, hoping to accurately simulate the radiation transfer process, estimate solar radiation and atmospheric parameters, and achieve absolute atmospheric correction. 2) Methods based on statistical information. Aims to enhance surface information or eliminate cloud information by mining the statistical features contained in the image, and have been widely used in thin cloud removal of remote sensing images. 3) Methods that couple radiation transfer models with data statistics. Combining the radiation transfer model method to accurately compensate for atmospheric effects and the statistical information method does not require any prior information, achieving efficient and accurate correction.

[0006] The reconstruction method of thick cloud loss mainly uses complementary information to reconstruct the cloud coverage area, which can be roughly divided into two categories: single-image reconstruction and multi-source data assisted reconstruction. 1) Single-image reconstruction method. It belongs to the extraction and utilization of spatial information, and uses clean pixels in the cloud coverage image to reconstruct the cloud coverage area. This method is suitable for small areas of cloud coverage, but not for large areas of cloud coverage. 2) Multi-source data assisted reconstruction method. The most popular method is to use multi-phase data and microwave data that is not interfered by clouds to reconstruct information on the target phase cloud coverage image. Although a large number of cloud removal processing methods have been developed for sensor images of different types and band configurations, most of the existing methods are only applicable to specific scenarios, resulting in incomplete correction problems such as spectral distortion in thin cloud areas and obvious boundaries in thick cloud areas.

[0007] Therefore, the present invention designs a remote sensing image composite cloud removal process to achieve integrated intelligent processing of thin cloud correction and thick cloud reconstruction, and gives quality grading standards for cloud removal results according to the degradation of the original data and the processing method. Summary of the invention

[0008] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a quality-graded remote sensing image composite cloud removal method, system and equipment, which implement a series of standardized processing on data acquired under different spatial and temporal conditions, eliminate the influence of atmospheric radiation attenuation and format it into surface reflection data with consistent observation characteristics, thereby improving the utilization rate of remote sensing images and promoting effective comparison and analysis of subsequent application research.

[0009] According to one aspect of the present invention, a method for removing composite clouds from remote sensing images with quality classification is provided, comprising:

[0010] Preprocess the acquired remote sensing images;

[0011] Based on the preprocessed data, thin cloud radiation correction is performed, wherein the thin cloud radiation correction includes clear sky sample selection and cirrus cloud correction based on scattering law;

[0012] Based on the data after the thin cloud radiation correction, thick cloud detection is performed, wherein the thick cloud detection includes using the Fmask3.3 algorithm to identify clouds and cloud shadows and create a mask image;

[0013] Based on the data after thick cloud detection, the thick cloud missing reconstruction is performed, wherein the thick cloud missing reconstruction includes reconstructing the thick cloud missing area by using a fast radiation adjustment and residual correction method or a multi-year weighted average method after matching the reference data;

[0014] Based on the data reconstructed after thick cloud loss, atmospheric correction is performed based on the 6S atmospheric correction model;

[0015] Data quality is graded based on the atmospherically corrected data.

[0016] As a further technical solution, the acquired remote sensing images are preprocessed, including:

[0017] Data screening is performed based on cloud coverage in remote sensing image metadata;

[0018] Based on the images after data screening, radiation correction and band stacking are performed in sequence.

[0019] As a further technical solution, thin cloud radiation correction is performed, including:

[0020] Using the cirrus band in the preprocessed image data, sample points in the cloud-free area are selected as clear sky samples;

[0021] Based on the selected clear sky samples, cirrus correction based on the scattering law is performed.

[0022] As a further technical solution, thick cloud detection is performed, including:

[0023] Based on the multispectral reflectance data after thin cloud radiation correction and the thermal infrared band data of the original remote sensing image, the Fmask3.3 algorithm is used to identify clouds and cloud shadows.

[0024] Based on the identified clouds and cloud shadows, morphological processing is performed to generate a mask image.

[0025] As a further technical solution, the method further includes: matching reference data based on thick cloud detection results.

[0026] As a further technical solution, the method also includes: after matching the reference data, according to different data missing situations, for thick cloud data that can provide sufficient clear sky reference data for adjacent phase images, a fast radiation adjustment and residual correction method is used to achieve reconstruction of the thick cloud missing area; for severely missing thick cloud degradation data, a multi-year weighted average method is used to achieve reconstruction of the thick cloud missing area.

[0027] As a further technical solution, fast radiation adjustment and residual correction methods are used to achieve reconstruction of thick cloud missing areas, including: using a fast radiation adjustment method based on a local linear histogram matching method to fill in the thick cloud coverage area, and performing fast residual correction through global optimization.

[0028] As a further technical solution, a multi-year weighted average method is used to achieve reconstruction of thick cloud missing areas, including: synthesizing reference images year by year and performing moment matching normalization processing to make their mean and standard deviation tend to the target data.

[0029] According to one aspect of the present invention, a system for removing composite clouds from remote sensing images with quality classification is provided, comprising:

[0030] Data preprocessing module, used to preprocess the acquired remote sensing images;

[0031] A thin cloud radiation correction module, used for performing thin cloud radiation correction based on the preprocessed data, wherein the thin cloud radiation correction includes clear sky sample selection and cirrus cloud correction based on the scattering law;

[0032] A thick cloud detection module, used for performing thick cloud detection based on the data after the thin cloud radiation correction, wherein the thick cloud detection includes using the Fmask3.3 algorithm to identify clouds and cloud shadows and create a mask image;

[0033] A thick cloud reconstruction module is used to reconstruct the missing thick clouds based on the data after thick cloud detection. The missing thick clouds reconstruction includes reconstructing the missing thick cloud areas by using a fast radiation adjustment and residual correction method or a multi-year weighted average method after matching the reference data.

[0034] Atmospheric correction module, used to perform atmospheric correction based on the 6S atmospheric correction model based on the data reconstructed after thick cloud loss;

[0035] The quality grading module is used to grade data quality based on the atmospheric corrected data.

[0036] According to one aspect of the present invention, there is provided a device for removing composite clouds from remote sensing images with quality graded, comprising a memory and a processor, wherein the memory stores program instructions to be executed by the processor, and the processor calls the program instructions to execute the steps of a method for removing composite clouds from remote sensing images with quality graded.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The present invention implements quality grading and adopts a series of standardized processing based on the acquired remote sensing images, and finally obtains the surface reflectance data that has eliminated the influence of atmospheric radiation attenuation and restored the surface observation under the cloud. The composite cloud removal process of the present invention mainly includes five steps: data preprocessing, thin cloud radiation correction, thick cloud detection, thick cloud missing reconstruction, and atmospheric correction. By implementing a series of standardized processing on the data acquired under different spatial and temporal conditions, the influence of atmospheric radiation attenuation is eliminated and formatted into surface reflectance data with consistent observation characteristics, thereby improving the utilization rate of remote sensing images and promoting effective comparison and analysis of subsequent application research.

[0039] 2. The present invention realizes batch processing and engineering application of algorithms at each stage, and realizes quality grading of processing results according to data processing conditions, which can ensure the radiation accuracy of images after cloud removal and provide high-quality data for subsequent remote sensing mission research. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction is given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 A schematic flow chart of a method for removing composite clouds from remote sensing images with quality grading provided in an embodiment of the present invention.

[0042] Figure 2 A schematic structural diagram of a quality-graded remote sensing image composite cloud removal system provided in an embodiment of the present invention.

[0043] Figure 3 A schematic structural diagram of a quality-graded remote sensing image composite cloud removal device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or the structural composition mode, but must be based on the ability of ordinary technicians in this field to achieve. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0045] The quality graded remote sensing image composite cloud removal method of the present invention is as follows: Figure 1 As shown, it mainly includes step 1 data preprocessing, step 2 thin cloud radiation correction, step 3 thick cloud detection, step 4 thick cloud missing reconstruction, step 5 atmospheric correction and step 6 data quality classification. Finally, the input Landsat8 / 9 image can be processed into cloud-free space seamless surface reflectance data.

[0046] The present invention can be implemented through four parts during specific implementation, namely: operating environment preparation, target and reference image preparation, composite cloud removal and data sorting.

[0047] (1) Prepare the operating environment

[0048] The algorithm of the present invention is implemented based on the Python programming language and requires high-performance computing resources to process large-scale image data. Specifically, the operating environment should include but is not limited to the following configurations: high-performance processors to ensure fast processing capabilities; sufficient memory capacity to support the loading and processing of large data sets; high-speed storage devices to increase data reading and writing speeds; appropriate software environments, including Python and its scientific computing libraries (such as NumPy, SciPy, Pandas) and remote sensing image processing libraries (such as GDAL, Libtiff, OpenCV), etc.

[0049] (2) Prepare the target and reference images

[0050] The present invention is designed for Landsat 8 / 9 satellite images and aims to perform efficient composite cloud removal tasks. To this end, the target and reference image data should meet the following requirements: the data set should include complete multispectral image files, covering visible light, near infrared and short-wave infrared bands; each image file should be accompanied by a corresponding metadata file, providing key parameters such as image acquisition time, geographic location, satellite orbit information, etc.; image data should be organized in the form of folders, and classified and stored according to the image row and column numbers (such as WRS-2 path / row number) and acquisition time (such as YYYYMMDD format) for fast retrieval and processing; in order to improve processing efficiency and accuracy, it is recommended to use standardized data formats, such as GeoTIFF or IMG format, and ensure that all image files have been calibrated to a unified geographic reference system.

[0051] (3) Composite cloud removal

[0052] By using the prepared target and reference images as input to the composite cloud removal process, the entire process of automated radiation correction, thin cloud removal, cloud detection, thick cloud removal, and atmospheric correction can be realized, and quality-graded cloud-free surface reflectance data can be output eventually.

[0053] (4) Data collation

[0054] After the composite cloud removal process is completed, the generated cloud-free surface reflectance data usually has a large file size. Therefore, a lossless compression algorithm is used to effectively compress the data to optimize storage and transmission performance. This not only helps to reduce storage costs, but also facilitates data sharing and distribution, providing support for subsequent remote sensing applications.

[0055] The embodiment of the present invention, in the part (3) of removing composite clouds, specifically includes the following steps:

[0056] Step 1: Remote sensing image preprocessing

[0057] The main technical steps of remote sensing image preprocessing include data screening, radiation correction, and band stacking. Due to the high cloud coverage, the surface information it provides is very limited. In order to reduce meaningless workload, data screening should be performed first. According to the cloud coverage in the remote sensing image metadata obtained by Landsat8 / 9 satellites, images with cloud coverage less than 50% are screened out and enter the composite cloud removal process. Secondly, radiation correction should be performed to eliminate the systematic and random radiation distortion or distortion generated by the data acquisition and transmission system. The main operation steps are to extract the three necessary parameters of gain, offset, and solar altitude angle from the image metadata file, and then convert the original unsigned 16-bit integer data into the physical atmospheric top reflectance data through formula (1).

[0058]

[0059] where ρ TOA Represents the reflectivity of the top atmosphere, DN represents unsigned 16-bit integer data, Gain represents gain, Offset represents offset, θ SE Represents the sun's altitude angle.

[0060] Finally, multiple single-band remote sensing images after radiation correction are merged into a multi-band image for comprehensive analysis and visualization.

[0061] Step 2: Thin cloud radiation correction

[0062] The main technical steps of thin cloud radiation correction include clear sky sample selection and thin cloud correction. Using the cirrus band in Landsat8 / 9, sample points in the cloud-free area are selected to provide clear sky reference for the thin cloud correction step. Because the coverage of a single scene image can be represented by approximately 50 million pixels, the 5 million points with the smallest value in the cirrus band, i.e. 1 / 10 of the data volume, are selected as clear sky samples for thin cloud correction.

[0063] In the thin cloud correction stage of this process, the cirrus correction method based on scattering law (CCM_SL) is selected. The specific steps are as follows: Fit the linear expression of the coastline band reflectance I1 and the blue band reflectance I2 of the clear sky sample area of ​​Landsat8 / 9 image I1=a·I2+b, where a and b are fitting coefficients. According to the radiation transfer model, the top atmosphere reflectance ρ obtained by the sensor is TOA It is the sum of the top-level reflectivity I provided by the ground objects in the clear sky and the cloud reflectivity C under the current atmospheric conditions. Subscript 1 represents the coastline band, and subscript 2 represents the blue band. Therefore, the top-level reflectivity of the atmosphere obtained by the sensor in the coastline band can be expressed as ρ TOA1=I1+C1, the blue band can be expressed as ρ TOA2 =I2+C2, other visible light and near-infrared bands can be expressed as ρ TOAi =I i +C i , subscript i represents the specified band. According to the law of atmospheric optical scattering, cloud radiation has band correlation. Subscript 9 represents the cirrus line band, so the cloud radiation correlation between the cirrus band and the coastline band can be expressed by the scattering law as The correlation with the blue band can be expressed as The correlation with other visible light and near-infrared bands can be expressed as Where λ represents the center wavelength of the band. By combining the above equations, the atmospheric environment parameter γ can be obtained, and according to the center wavelength of the band set by the sensor, the reflectivity of the top layer of the atmosphere provided by the ground object can be obtained. The above steps can achieve high-precision correction for the radiation difference area caused by thin cloud cover.

[0064] Step 3: Thick Cloud Detection

[0065] Thin clouds and thick clouds in images often appear at the same time in the form of composite clouds. After the thin cloud correction is achieved, the image should also process the thick cloud coverage area to reconstruct the missing information. Thick cloud detection aims to comprehensively use the characteristics of multispectral image data to identify features such as clouds and cloud shadows. In this process, the Fmask3.3 algorithm is selected to achieve thick cloud detection. The specific steps are as follows: First, the multispectral reflectance data after thin cloud correction is retained, and the thermal infrared band data is converted into brightness temperature. Secondly, the reflectance is greater than 0.03, the brightness temperature is less than 27℃, the normalized snow cover index and the normalized vegetation index should be less than 0.8, the whiteness index is less than 0.7, the haze optimization transformation index is greater than 0, and the radiation ratio of the red band to the near-infrared band is greater than 0.75. The spectral thresholds are used to separate potential cloud pixels and absolutely cloud-free pixels. Then, according to the unique physical characteristics of the cloud layer, the normalized temperature probability, spectral variation probability and brightness probability are combined to calculate the cloud probability value of each pixel for land and water bodies respectively. Subsequently, the potential shadow layer is generated by the flood filling algorithm and the cloud shadow position is predicted by combining the sensor's viewing angle, illumination angle, and cloud probability. Finally, the image is segmented based on the empirical threshold to obtain the cloud and cloud shadow mask image, with pixels with a value of 0 representing invalid data, pixels with a value of 1 representing clear sky data, pixels with a value of 128 representing cloud shadows, and pixels with a value of 255 representing clouds. In addition, morphological techniques are used to enhance the mask to ensure that edge pixels are correctly included in the cloud and cloud shadow areas, and the final cloud detection result, namely the cloud mask image M, is obtained.

[0066] Step 4: Thick cloud missing reconstruction

[0067] Thick cloud missing reconstruction is divided into two cases according to different data missing situations: when adjacent phase images can provide sufficient clear sky reference data, the Fast Radiometric Adjustment and Residual Correction method (FRARC) can significantly improve the reconstruction quality because the surface changes are small and the seasonal characteristics are closer. The multi-year weighted average method (MWAM) is used to achieve seamless spatial reconstruction of the remaining severely missing thick cloud degraded data. The specific reconstruction steps are as follows:

[0068] 1.1 Reference data matching

[0069] Compare the cloud mask image M of the target reconstructed image T and the cloud mask image M of its adjacent phase images R , when the clear sky area (pixel with cloud detection result of 1) of the adjacent phase image can completely cover the cloud area (pixel with cloud detection result of 255) and cloud shadow area (pixel with cloud detection result of 128) of the target reconstructed image. The image of the adjacent phase is the reference data of the target reconstructed image, and the thick cloud removal is completed using the fast radiometric adjustment and residual correction method. When the above conditions are not met, all multi-phase images of the corresponding seasons collected by Landsat8 / 9 will be used as the reference data of the target image, and the thick cloud removal is completed using the multi-year weighted average method.

[0070] 1.2 Rapid radiometric adjustment and residual correction methods

[0071] The input of the FRARC method is a target image covered by thick clouds ρ TOA_T (T indicates the target image), reference image ρ TOA_R , and the cloud mask image M of the target image T and the cloud mask image M of the reference image R The FRARC method consists of two main steps. First, a fast radiation adjustment method based on a local linear histogram matching method is used to fill the thick cloud cover area, as shown in Equation (2).

[0072]

[0073] Among them, (i,j) represents the pixel in the thick cloud coverage area, ρ TOA_T′ represents the result after rapid radiation adjustment, σ represents the standard deviation, and μ represents the mean.

[0074] Then, a fast residual correction is performed through global optimization, as shown in Equation (3), which further eliminates the inconsistency of radiometric measurements between the reconstructed area and the clear sky area.

[0075]

[0076] Among them, ρ TOA_T″ represents the result of thick cloud reconstruction, represents the residual term after fast radiation adjustment, Ω is the correction region, ▽ is the gradient operator, and ω is the weight used to balance the gradient and intensity fidelity.

[0077] 1.3 Multi-year weighted average method

[0078] The multi-year weighted average method is based on the K-period reference image ρ TOA_R (R indicates the original reference image) Reconstructed target image ρ TOA_T First, synthesize the reference image ρ year by year TOA_R′ (R′ denotes the annual synthetic reference image), and the synthesis rule is to calculate the median reflectance of multiple reference images in each season for each pixel, as shown in formula (4).

[0079] ρ TOA_R′(i,j) =median({ρ TOA_R(i,j) [k]|k=1,2,…,K}) (4)

[0080] Then, the mean and standard deviation of the overlapping part of the effective area of ​​the target data are used as the standard to synthesize all the reference images ρ TOA_R′ Perform moment matching normalization processing to make its mean and standard deviation approach the target data, as shown in formula (5).

[0081]

[0082] Where n represents the data of the nth year, and A_R′ and A_T represent the pixels corresponding to the overlapping areas of R′ and T in their respective images.

[0083] After normalization, the distribution level of the pixel reflectance of the reference data is similar to that of the pixel reflectance of the target data. Then the correlation coefficient CC between the normalized data and the overlapping part of the valid area of ​​the target data is calculated, and the weight W is obtained as follows:

[0084]

[0085] Finally, the missing areas are filled by multi-year weighted averaging to obtain the reconstruction result.

[0086] ρ TOA_T″ =ρ TOA_T +W R″ ρ TOA_R″ (8)

[0087] Step 5: Atmospheric Correction

[0088] Based on the 6S atmospheric correction model, atmospheric correction is achieved according to the satellite geometric conditions, image acquisition time and regional location, corresponding atmospheric model, aerosol model, target altitude, sensor altitude, spectral information and other parameters recorded in the metadata, and finally the multi-band cloud-free surface reflectance data is obtained, as shown in the following formula.

[0089] ρ SR_T =6S(ρ TOA_T″ ,Meta_data) (9)

[0090] Among them, ρ SR_T It represents the surface reflectance of the target image, and Meta_data represents metadata and correction related parameters.

[0091] Step 6: Data quality grading

[0092] According to the cloud coverage of the image, the reference data, the processing flow and the reconstruction accuracy, the data quality classification standard is proposed as follows: (1) Level L1, representing the highest quality, that is, the cloud-free surface reflectance data obtained after the original clear sky data is calibrated by radiation and atmospheric correction; (2) Level L2, representing the second highest quality, that is, the cloud-free surface reflectance data obtained after the image contains only thin clouds and is corrected by thin clouds and atmosphere; (3) Level L3, representing relatively high quality, that is, the cloud-free surface reflectance data reconstructed with thick clouds using images of adjacent time phases; (4) Level L4, representing medium quality, that is, the cloud-free surface reflectance data reconstructed with thick clouds using images of multiple years and multiple periods.

[0093] The implementation basis of each embodiment of the present invention is to implement programmed processing through a device with a processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, on the basis of the above embodiments, an embodiment of the present invention provides a quality-graded remote sensing image composite cloud removal system, which is used to execute a quality-graded remote sensing image composite cloud removal method in the above method embodiment.

[0094] See also Figure 2The system includes: a data preprocessing module for preprocessing the acquired remote sensing images; a thin cloud radiation correction module for performing thin cloud radiation correction based on the preprocessed data, wherein the thin cloud radiation correction includes clear sky sample selection and cirrus correction based on the scattering law; a thick cloud detection module for performing thick cloud detection based on the data after the thin cloud radiation correction, wherein the thick cloud detection includes using the Fmask3.3 algorithm to identify clouds and cloud shadows and create a mask image; a thick cloud reconstruction module for performing thick cloud missing reconstruction based on the data after the thick cloud detection, wherein the thick cloud missing reconstruction includes, after matching the reference data, using a fast radiation adjustment and residual correction method or a multi-year weighted average method to achieve reconstruction of the thick cloud missing area; an atmospheric correction module for performing atmospheric correction based on the data after the thick cloud missing reconstruction based on the 6S atmospheric correction model; and a quality grading module for performing data quality grading based on the atmospheric corrected data.

[0095] The embodiment of the present invention provides a quality graded remote sensing image composite cloud removal system. Most existing methods are only applicable to specific scenes, resulting in incomplete correction problems such as spectral distortion in thin cloud areas and obvious boundaries in thick cloud areas. Figure 2 Several modules in the system are used to realize the integrated intelligent processing of thin cloud correction and thick cloud reconstruction, and the quality grading standards are given to the cloud removal results according to the degradation of the original data and the processing method.

[0096] It should be noted that the system embodiment provided by the present invention is used to implement the methods in the above method embodiment as well as the methods in other method embodiments provided by the present invention. The only difference is that the corresponding functional modules are set. The principle is basically the same as the principle of the above system embodiment provided by the present invention. As long as the technical personnel in this field refer to the specific technical solutions in other method embodiments on the basis of the above system embodiment, obtain the corresponding technical means and the technical solutions composed of these technical means by combining the technical features, and on the premise of ensuring the practicality of the technical solutions, improve the modules in the above system embodiment to obtain the corresponding system class embodiments, which are used to implement the methods in other method class embodiments. For example:

[0097] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the data preprocessing module is further used to execute the following instructions:

[0098] Data screening is performed based on cloud coverage in remote sensing image metadata;

[0099] Based on the images after data screening, radiation correction and band stacking are performed in sequence.

[0100] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the thin cloud radiation correction module is further used to execute the following instructions:

[0101] Using the cirrus band in the preprocessed image data, sample points in the cloud-free area are selected as clear sky samples;

[0102] Based on the selected clear sky samples, cirrus correction based on the scattering law is performed.

[0103] Based on the content of the above system embodiment, as a preferred embodiment, in a quality-graded remote sensing image composite cloud removal system provided in an embodiment of the present invention, the thick cloud detection module is further used to execute the following instructions:

[0104] Based on the multispectral reflectance data after thin cloud radiation correction and the thermal infrared band data of the original remote sensing image, the Fmask3.3 algorithm is used to identify clouds and cloud shadows.

[0105] Based on the identified clouds and cloud shadows, morphological processing is performed to generate a mask image.

[0106] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the thick cloud reconstruction module is further used to execute the following instructions:

[0107] Matching of reference data is performed based on thick cloud detection results.

[0108] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the thick cloud reconstruction module is further used to execute the following instructions:

[0109] After matching the reference data, according to different data missing situations, for thick cloud data whose adjacent phase images can provide sufficient clear sky reference data, the rapid radiation adjustment and residual correction methods are used to reconstruct the thick cloud missing area; for the thick cloud degradation data with severe missing data, the multi-year weighted average method is used to reconstruct the thick cloud missing area.

[0110] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the thick cloud reconstruction module is further used to execute the following instructions:

[0111] Fast radiation adjustment and residual correction methods are used to achieve reconstruction of thick cloud missing areas, including: using a fast radiation adjustment method based on a local linear histogram matching method to fill in thick cloud coverage areas, and performing fast residual correction through global optimization.

[0112] Based on the content of the above system embodiment, as a preferred embodiment, a quality-graded remote sensing image composite cloud removal system is provided in the embodiment of the present invention, wherein the thick cloud reconstruction module is further used to execute the following instructions:

[0113] The multi-year weighted average method is used to reconstruct the thick cloud missing area, including synthesizing reference images year by year and performing moment matching normalization processing to make its mean and standard deviation tend to the target data.

[0114] Based on the same inventive concept as the above-mentioned embodiment, an embodiment of the present invention also provides a quality-graded remote sensing image composite cloud removal device, including a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the quality-graded remote sensing image composite cloud removal method.

[0115] In an embodiment of the present invention, the memory may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory (volatile memory), such as a random-access memory (RAM). The memory is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in an embodiment of the present invention may also be a circuit or any other device that can implement a storage function, for storing program instructions and / or data.

[0116] In the embodiments of the present invention, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0117] In summary, the present invention provides a quality-graded remote sensing image composite cloud removal process, which is based on the multispectral images acquired by the Landsat8 / 9 satellites, realizes quality grading and adopts a series of standardized processing, and finally obtains the surface reflectance data that has eliminated the influence of atmospheric radiation attenuation and restored the surface observation under the cloud. The composite cloud removal process mainly includes five steps: data preprocessing, thin cloud radiation correction, thick cloud detection, thick cloud missing reconstruction, and atmospheric correction. The present invention realizes batch processing and engineering application of algorithms at each stage, and realizes quality grading of processing results according to the data processing situation, which can ensure the radiation accuracy of the image after the cloud layer is removed, and provide high-quality data for subsequent remote sensing mission research.

[0118] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A quality-graded remote sensing image composite cloud removal method, characterized in that: include: Preprocess the acquired remote sensing images; Based on the preprocessed data, thin cloud radiation correction is performed, wherein the thin cloud radiation correction includes clear sky sample selection and cirrus cloud correction based on scattering law; Based on the data after the thin cloud radiation correction, thick cloud detection is performed, wherein the thick cloud detection includes using the Fmask3.3 algorithm to identify clouds and cloud shadows and create a mask image; Based on the data after thick cloud detection, the thick cloud missing reconstruction is performed, wherein the thick cloud missing reconstruction includes reconstructing the thick cloud missing area by using a fast radiation adjustment and residual correction method or a multi-year weighted average method after matching the reference data; Based on the data reconstructed after thick cloud loss, atmospheric correction is performed based on the 6S atmospheric correction model; Data quality is graded based on the atmospherically corrected data.

2. A method for removing composite clouds from remote sensing images by quality classification according to claim 1, characterized in that: Preprocess the acquired remote sensing images, including: Data screening is performed based on cloud coverage in remote sensing image metadata; Based on the images after data screening, radiation correction and band stacking are performed in sequence.

3. The method for removing composite clouds from remote sensing images by quality classification according to claim 1, characterized in that: Perform thin cloud radiation correction, including: Using the cirrus band in the preprocessed image data, sample points in the cloud-free area are selected as clear sky samples; Based on the selected clear sky samples, cirrus correction based on the scattering law is performed.

4. The method for removing composite clouds from remote sensing images by quality classification according to claim 1, characterized in that: Perform thick cloud detection, including: Based on the multispectral reflectance data after thin cloud radiation correction and the thermal infrared band data of the original remote sensing image, the Fmask3.3 algorithm is used to identify clouds and cloud shadows. Based on the identified clouds and cloud shadows, morphological processing is performed to generate a mask image.

5. A method for removing composite clouds from remote sensing images by quality classification according to claim 4, characterized in that: The method further includes: matching reference data based on the thick cloud detection result.

6. The method for removing composite clouds from remote sensing images by quality classification according to claim 1, characterized in that: The method further comprises: after matching the reference data, according to different data missing situations, for thick cloud data that can provide sufficient clear sky reference data for adjacent phase images, a fast radiation adjustment and residual correction method is used to achieve reconstruction of the thick cloud missing area; for the severely missing thick cloud degradation data, a multi-year weighted average method is used to achieve reconstruction of the thick cloud missing area.

7. A method for removing composite clouds from remote sensing images by quality classification according to claim 6, characterized in that: Fast radiation adjustment and residual correction methods are used to achieve reconstruction of thick cloud missing areas, including: using a fast radiation adjustment method based on a local linear histogram matching method to fill in thick cloud coverage areas, and performing fast residual correction through global optimization.

8. The method for removing composite clouds from remote sensing images by quality classification according to claim 6, characterized in that: The multi-year weighted average method is used to reconstruct the thick cloud missing area, including synthesizing reference images year by year and performing moment matching normalization processing to make its mean and standard deviation tend to the target data.

9. A quality-graded remote sensing image composite cloud removal system, characterized in that: include: Data preprocessing module, used to preprocess the acquired remote sensing images; A thin cloud radiation correction module, used for performing thin cloud radiation correction based on the preprocessed data, wherein the thin cloud radiation correction includes clear sky sample selection and cirrus cloud correction based on the scattering law; A thick cloud detection module, used for performing thick cloud detection based on the data after the thin cloud radiation correction, wherein the thick cloud detection includes using the Fmask3.3 algorithm to identify clouds and cloud shadows and create a mask image; A thick cloud reconstruction module is used to reconstruct the missing thick clouds based on the data after thick cloud detection. The missing thick clouds reconstruction includes reconstructing the missing thick cloud areas by using a fast radiation adjustment and residual correction method or a multi-year weighted average method after matching the reference data. Atmospheric correction module, used to perform atmospheric correction based on the 6S atmospheric correction model based on the data reconstructed after thick cloud loss; The quality grading module is used to grade data quality based on the atmospheric corrected data.

10. A quality-graded remote sensing image composite cloud removal device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of a method for removing composite clouds from remote sensing images with quality grading as described in any one of claims 1 to 8.

Images