Satellite remote sensor cross calibration method and device based on invariant pixels

Abstract

The present invention provides a satellite remote sensor cross-calibration method and device based on invariant pixels. The method comprises: determining an image pair sequence; obtaining the apparent reflectance of a single pixel in each image pair and inputting it into an invariant pixel detection model to obtain the invariant pixel in each image pair; obtaining the spectral matching factor of the remote sensor to be calibrated and the reference remote sensor for spectral matching to determine the spectrally corrected apparent reflectance of the reference remote sensor of the invariant pixel in each image pair; orthogonally regressing the spectrally corrected apparent reflectance of the reference remote sensor of the invariant pixel in each image pair with the corresponding apparent reflectance of the remote sensor to be calibrated to determine the cross-calibration coefficient. The present invention automatically detects the invariant pixel target in the scene, calculates the spectral matching factor to correct the spectral response difference between the two sensors, establishes a linear fitting relationship of the apparent reflectance of the invariant pixel sample, and obtains the cross-calibration result and its long-time series cross-calibration coefficient.

Description

Technical Field

[0001] The invention relates to the technical field of remote sensor calibration, and in particular to a satellite remote sensor cross calibration method and device based on invariant pixels. Background Art

[0002] With the rapid development of informatization and globalization and the continuous advancement of aerospace remote sensing technology, remote sensing applications have gradually penetrated into all areas of human activities. Accurate calibration of remote sensors is an important prerequisite for quantitative remote sensing applications.

[0003] There are many methods for radiation calibration of remote sensors. In absolute radiation calibration, it is necessary to obtain the absolute radiation value of the observed target in advance. The implementation process is difficult and difficult to apply to historical satellite data for re-calibration. The on-orbit calibration method relies on on-board calibration equipment. Many satellites lack on-board calibration equipment or are limited by the development level of on-board calibration equipment, resulting in low calibration accuracy or inability to perform on-orbit calibration.

[0004] Cross-calibration is an effective and feasible on-orbit alternative calibration method, which uses sensors with high absolute calibration accuracy as a reference and cross-calibrates the sensors to be calibrated by observing the same target at the same time. However, the existing cross-calibration methods have the disadvantages of difficulty in selecting invariant targets, single surface characteristics of the observed targets, small dynamic range of reflectivity, and low calibration frequency. Summary of the invention

[0005] The present invention provides a satellite remote sensor cross-calibration method and device based on invariant pixels, which is used to solve the defects of the prior art in cross-calibration of remote sensors, such as difficulty in selecting invariant targets, small dynamic range of target reflectivity, and low calibration frequency. The present invention does not need to manually select invariant targets, and can effectively increase the cross-calibration frequency between sensors. At the same time, a large number of invariant pixels effectively increase the coverage of reflectivity, and can be applied to large scene data.

[0006] The present invention provides a satellite remote sensor cross calibration method based on invariant pixels, comprising:

[0007] Determine a sequence of image pairs, wherein each of the image pairs includes multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day;

[0008] Obtaining the apparent reflectance of a single pixel in each image pair and inputting it into an invariant pixel detection model to obtain the invariant pixel in each image pair;

[0009] Acquire spectral matching factors of the remote sensor to be calibrated and the reference remote sensor, perform spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factors, and determine the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair;

[0010] The spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair is orthogonally regressed with the apparent reflectance of the corresponding remote sensor to be calibrated to determine the cross calibration coefficient.

[0011] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the step of determining an image pair sequence comprises:

[0012] The multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day when the satellites pass by at the same time;

[0013] The multispectral image collected by the remote sensor to be calibrated and the multispectral image collected by the reference remote sensor are preprocessed to obtain a target image pair.

[0014] The image pair sequence is determined based on the target image pairs collected over multiple days.

[0015] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the preprocessing includes resolution resampling, rasterization and elimination of invalid pixels.

[0016] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the elimination of invalid pixels comprises:

[0017] Eliminate cloud-contaminated pixels, water target pixels, and pixels with satellite observation zenith angles greater than or equal to 30°.

[0018] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the invariant pixel detection model is established based on an iterative weighted multivariate change detection IR-MAD method.

[0019] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the apparent reflectance of a single pixel in each image pair is obtained and input into an invariant pixel detection model, and the invariant pixel of each image pair is output, including:

[0020] Obtaining the detection result of a single pixel in each image pair and calculating the apparent reflectivity of the pixel;

[0021] Inputting the pixel apparent reflectance of a single pixel in each image pair into an invariant pixel detection model, constructing a MAD variable and constructing an observation value of a single pixel in each image pair based on the MAD variable;

[0022] Based on the observed value of a single pixel in each image pair and an invariant probability decision threshold, the invariant pixel of each image pair is determined.

[0023] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, the step of obtaining a spectral matching factor between a remote sensor to be calibrated and a reference remote sensor includes:

[0024] Acquire the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor;

[0025] Based on the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor, a spectral matching factor between the remote sensor to be calibrated and the reference remote sensor is determined.

[0026] According to a satellite remote sensor cross-calibration method based on invariant pixels provided by the present invention, orthogonal regression is performed on the spectrally corrected apparent reflectance of the reference remote sensor of the invariant pixel of each image pair and the apparent reflectance of the remote sensor to be calibrated of the invariant pixel of the corresponding image pair to determine the cross-calibration coefficient, including:

[0027] Establishing a linear fitting relationship between the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair and the apparent reflectance of the to-be-calibrated remote sensor of the unchanged pixel of the corresponding image pair to perform orthogonal regression;

[0028] Based on the slope and intercept of the orthogonal regression, the cross calibration coefficient was determined.

[0029] According to a satellite remote sensor cross calibration method based on invariant pixels provided by the present invention, the cross calibration coefficient is determined, and then the method further includes:

[0030] Performing long-time series cross-calibration on each channel of the long-time series data of the remote sensor to be calibrated;

[0031] Based on the result of the cross calibration of the long time series, a cross calibration coefficient of the long time series is determined.

[0032] The present invention also provides a satellite remote sensor cross calibration device based on invariant pixels, comprising:

[0033] An acquisition module is used to determine a sequence of image pairs, each of which includes multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day;

[0034] An invariant pixel detection module, used to obtain the apparent reflectance of a single pixel in each image pair and input it into an invariant pixel detection model to obtain the invariant pixel in each image pair;

[0035] A spectral matching module, used to obtain spectral matching factors of the remote sensor to be calibrated and the reference remote sensor, perform spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factors, and determine the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair;

[0036] The regression module is used to perform orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair and the apparent reflectance of the corresponding remote sensor to be calibrated to determine the cross calibration coefficient.

[0037] The satellite remote sensor cross-calibration method and device based on invariant pixels provided by the present invention are different from the traditional cross-calibration method, do not rely on the invariant calibration field artificially selected by the satellite transit, and are not limited by the harsh conditions of ground synchronous observation, and can effectively improve the cross-calibration frequency between sensors. The detection of invariant pixel targets is carried out automatically, which has a positive effect on the processing of long-term series data and recalibration of historical satellite data. The acquired invariant pixel targets are discontinuous in space. Although the spatial characteristics of the surface cannot be fully analyzed, a large number of invariant pixel samples effectively increase the coverage of reflectivity and can be applied to large scene data, improving the problems of small reflectivity range and single sample of the target site observed by the traditional cross-calibration method. The calibration results can be used for the research and application of subsequent quantitative inversion products of remote sensing data and long-term monitoring of the radiation response characteristics of remote sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use 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 creative work.

[0039] Figure 1 This is one of the flow charts of the satellite remote sensor cross calibration method based on invariant pixels provided by an embodiment of the present invention;

[0040] Figure 2 This is a second flow chart of a satellite remote sensor cross calibration method based on invariant pixels provided by an embodiment of the present invention;

[0041] Figure 3It is a schematic diagram of the FY-3B / VIRR and MERSI channel spectral response functions and SCIAMACHY hyperspectral samples provided by the embodiments of the present invention;

[0042] Figure 4 is a schematic diagram of orthogonal regression results of apparent reflectance of invariant pixels provided by an embodiment of the present invention;

[0043] Figure 5 It is a schematic diagram of a long time series of VIRR relative calibration slope provided by an embodiment of the present invention;

[0044] Figure 6 It is a structural schematic diagram of a satellite remote sensor cross-calibration device based on invariant pixels provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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 of 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.

[0046] The existing cross-calibration methods are mainly divided into two categories. One is the SNO cross-calibration method, and its main steps are:

[0047] 1) Obtain the geographical location and observation time of the intersection point of the two satellite orbits through orbit prediction;

[0048] 2) Based on the time and geographic location of the orbital forecast intersection, pixel matching of the observation data has been selected for the intersection calibration, including time, observation geometry, and spatial position matching;

[0049] 3) Spectral matching is performed based on the spectral response differences of similar channels of two remote sensors;

[0050] 4) Calculate the calibration coefficients by regression analysis.

[0051] The other is the site cross calibration method, the main steps of which are:

[0052] 1) Select a suitable unchanged site as the observation target;

[0053] 2) Obtain the invariant site observation data with matching observation conditions of the two remote sensors and project them to the same geographic grid;

[0054] 3) Using the radiation transfer model and spectral matching to obtain the simulated apparent reflectance or apparent radiance of the reference sensor;

[0055] 4) Calculate the calibration coefficients by regression analysis.

[0056] Based on the above two existing cross-calibration methods, the following three cross-calibration problems can be found:

[0057] The selection of invariant targets is difficult. SNO cross-calibration requires the use of orbit prediction models to select the orbit intersection points of the satellite. Only through strict pixel matching in the limited intersection points can the available cross-calibration data be obtained. Site cross-calibration also requires manual selection and demarcation of invariant sites, which consumes manpower and material resources.

[0058] The surface characteristics of the observed target are single, and the dynamic range of reflectivity is small. For example, when using the desert invariant field for site cross-calibration, the reflectivity of the visible light band of the desert site is approximately between 0.2-0.3, but the radiation performance of the remote sensor has a certain target dependence, that is, the calibration coefficient will change with the dynamics of the target radiation. Therefore, a large number of target samples are required to cover the dynamic range of the remote sensor as much as possible to obtain a high-precision calibration result.

[0059] Low calibration frequency. Whether it is SNO cross-calibration or field cross-calibration, the frequency of cross-calibration is low due to the difficulty of manually selecting unchanging targets, and it is impossible to carry out continuous radiation calibration to achieve the goal of long-term detection of remote sensor radiation performance.

[0060] Based on the above problems, an embodiment of the present invention provides a satellite remote sensor cross-calibration method based on invariant pixels to achieve cross-calibration of the remote sensor and solve the defects of difficulty in selecting invariant targets, small dynamic range of target reflectivity and low calibration frequency when cross-calibrating the remote sensor.

[0061] Combine the following Figure 1-Figure 5 The satellite remote sensor cross calibration method based on invariant pixels of the present invention is described as follows: Figure 1 As shown, the method comprises at least the following steps:

[0062] Step 101, determining a sequence of image pairs, wherein each image pair includes multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day;

[0063] Step 102: Obtain the apparent reflectance of a single pixel in each image pair and input it into an invariant pixel detection model to obtain the invariant pixel in each image pair;

[0064] Step 103, obtaining spectral matching factors of the remote sensor to be calibrated and the reference remote sensor, performing spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factors, and determining the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair;

[0065] Step 104: Perform orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair and the apparent reflectance of the corresponding remote sensor to be calibrated to determine a cross calibration coefficient.

[0066] Regarding step 101, it should be noted that the image pair sequence includes multiple image pairs, and each group of image pairs is a multispectral image of the same target scene collected by two remote sensors at the same time when the satellite passes by at the same time. The two image pairs may be two groups of multispectral images collected at the same time for the same observation scene on two consecutive days, or two groups of multispectral images collected at different times of the same day for the same observation scene, or two groups of multispectral images with different dates, collection times, and observation scenes. The image pair sequence needs to collect satellite data for multiple consecutive days, and the continuous collection time must be within the satellite orbit regression period.

[0067] In addition, each multispectral image contains satellite data of multiple channel bands. When selecting image pairs, it is also necessary to perform channel matching on two multispectral images of the same target scene that are taken at the same time, and to match channels of similar bands one by one.

[0068] Regarding step 102, it should be noted that the detection result of a single pixel in each image pair is generally a remote sensing sensor radiation measurement value, i.e., a DN count value. The apparent reflectance of each pixel can be calculated based on the detection result. The invariant pixel detection model is a pre-established model that can filter out invariant pixels from the input image pairs based on the apparent reflectance. For each single-day image pair, after passing through the invariant pixel detection model, the invariant pixels of the image pair can be output. After the image pair sequence is sequentially input into the invariant pixel detection model, a sequence of invariant pixels corresponding to different image pairs can be obtained.

[0069] Regarding step 103, it should be noted that since there is a difference in the spectral response of the corresponding channels of the remote sensor to be calibrated and the reference remote sensor, that is, for the same entrance pupil radiation, the two remote sensors will obtain different measurement values, it is necessary to perform spectral matching on the corresponding channels of the remote sensors to correct the calibration error caused by the difference in spectral response.

[0070] In this embodiment, the spectral matching factor of the remote sensor to be calibrated and the reference remote sensor needs to use the hyperspectral instrument observation data as the hyperspectral sample, which is convolved with the channel spectral response function of the remote sensor to be calibrated and the reference remote sensor respectively to obtain the entrance pupil radiance of the corresponding remote sensor, and then it can be obtained by establishing a relationship between the entrance pupil radiance of the matching channel of the remote sensor to be calibrated and the reference remote sensor.

[0071] Regarding step 104, it should be noted that the invariant pixels obtained by the invariant pixel detection model of any image pair can obtain the cross calibration coefficient after processing in steps 103 and 104. However, due to the vast scene used for statistical analysis, the single-day image pair results only contain partial area information in the scene that meets the observation zenith angle restriction conditions. In order to include more abundant invariant pixel samples in a single regression as much as possible, this embodiment collects the image pair sequence to merge the invariant pixels obtained by continuous detection for multiple days and performs fitting processing. Based on the short-term stability of the remote sensor radiation response, this merging process is feasible and will help expand the reflectivity value range of the pixel samples, effectively improving the regression effect and calibration accuracy.

[0072] The cross-calibration method for satellite remote sensors based on invariant pixels in the embodiment of the present invention is targeted at the problems of invariant target selection, dynamic range limitation of reflectivity of observed targets, and continuous calibration monitoring of remote sensors for long time series in the existing cross-calibration methods, and proposes an effective solution. This method is different from the traditional cross-calibration method. It does not rely on the invariant calibration field manually selected by the satellite transit, nor is it limited by the harsh conditions of ground synchronous observation, and can effectively improve the cross-calibration frequency between sensors. The detection of invariant pixel targets is automatic, which has a positive effect on processing long time series data and recalibration of historical satellite data. The acquired invariant pixel targets are discontinuous in space. Although the spatial characteristics of the surface cannot be fully analyzed, a large number of invariant pixel samples effectively increase the coverage of reflectivity, and can be applied to large scene data, improving the problems of small reflectivity range and single sample of the observed target site in the traditional cross-calibration method. The calibration results can be used for the research and application of subsequent quantitative inversion products of remote sensing data and long-term monitoring of the radiation response characteristics of remote sensors.

[0073] It can be understood that determining the image pair sequence includes:

[0074] The multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day when the satellites pass by at the same time;

[0075] The multispectral image collected by the remote sensor to be calibrated and the multispectral image collected by the reference remote sensor are preprocessed to obtain a target image pair.

[0076] Based on the target image pairs collected over multiple days, an image pair sequence is determined.

[0077] It should be noted that due to the different spatial resolutions between different remote sensors and between different channels, the different data dimensions make it impossible to carry out subsequent statistical analysis. Therefore, it is necessary to preprocess the multispectral images collected by the calibration remote sensor and the multispectral images collected by the reference remote sensor. The preprocessed image pairs are the target image pairs. The target image pairs are collected over multiple consecutive days to form an image pair sequence.

[0078] It can be understood that preprocessing includes resolution resampling, rasterization and elimination of invalid pixels.

[0079] It should be noted that during preprocessing, the same channel bands of the two multispectral images in the same image pair need to be matched one by one, and the basis for the subsequent step processing is the image pair under the same channel band, and the DN values ​​of the two multispectral images of each image pair under each channel band are obtained. In an embodiment of the present invention, the matching of channel bands can be achieved by resampling the resolution of the satellite data of each image pair. The satellite remote sensing data corresponding to the multispectral image is projected onto the same geographic grid, and the rasterized satellite remote sensing data is organized into a satellite data set. Among them. The geographic grid is a data form that divides the geographic space into regular grids, each grid is called a unit, and a corresponding attribute value is assigned to each unit to represent the satellite remote sensing data.

[0080] There are invalid data in the DN values ​​of different channels obtained by the two remote sensors. Since the invariant pixel detection model is based on statistical principles and requires the dimensions of the two sets of input data to be consistent, it is necessary to eliminate the invalid pixels in the two images of each image pair.

[0081] It is understandable that removing invalid pixels includes:

[0082] Eliminate cloud-contaminated pixels, water target pixels, and pixels with satellite observation zenith angles greater than or equal to 30°.

[0083] It should be noted that when the invariant pixel detection model is used to detect invariant pixels in a scene, the variable pixels will be gradually eliminated during the algorithm iteration process. However, considering the algorithm's operating efficiency, the significant unstable pixels in the scene, also known as cloud pollution pixels, such as clouds, dust and other targets, should be eliminated before the algorithm corresponding to the invariant pixel detection model is run. Although water bodies and oceans have relatively stable reflectivity characteristics, their reflectance spectral characteristics vary greatly, and the signals of some channel bands are weak. For water body target pixels, the land and ocean mask data in the data are used to eliminate them.

[0084] In addition, after removing invalid points, clouds, dust and ocean water pixels from the image pair sequence, there are still millions of pixel samples, among which there are a large number of pixel samples with too large satellite observation zenith angles. Too large satellite observation zenith angles will lead to a decrease in pixel spatial resolution and detection radiation accuracy. In order to improve the recognition accuracy of the invariant pixel detection model for invariant pixels, pixel samples with satellite observation zenith angles less than 30° are used for subsequent statistical analysis.

[0085] It can be understood that the invariant pixel detection model is established based on the iterative weighted multivariate change detection IR-MAD method.

[0086] It should be noted that with the advancement of computer information technology and the widespread application of statistical analysis methods in scientific research, remote sensing researchers have gradually applied some mathematical methods to the processing and analysis of satellite data. In the prior art, the difference image comparison method and the ratio method are suitable for the analysis of single-channel images, but not for multi-channel satellite remote sensor data. The principal component analysis (PCA) method can integrate the change information of each channel, but cannot eliminate the correlation between different remote sensor channels. A multivariate change detection technology (MAD) with linear scale invariance can eliminate the correlation within the sensor channel at the same time and the correlation between different channels of the sensor. Based on this, the embodiment of the present invention uses the IR-MAD method to construct an invariant pixel detection model for simultaneous observation scenes of different remote sensors, and applies it to the cross-calibration of remote sensors.

[0087] It can be understood that in the embodiment of the present invention, the invariant pixel detection model is constructed by the IR-MAD method to obtain the invariant pixels in each image pair. Among them, the invariant pixel detection model is used to analyze and process two n-channel multispectral images of the same scene collected by the remote sensor to be calibrated and the reference remote sensor at the same time, which at least includes the following steps:

[0088] Step 201, obtaining the detection result of a single pixel in each image pair and calculating the apparent reflectivity of the pixel;

[0089] Step 202, inputting the pixel apparent reflectance of a single pixel in each image pair into the invariant pixel detection model, constructing a MAD variable and constructing the observation value of a single pixel in each image pair based on the MAD variable;

[0090] Step 203: Determine the invariant pixels of each image pair based on the observed values ​​of the individual pixels in each image pair and the invariant probability decision threshold.

[0091] Regarding step 201, it should be noted that the apparent reflectance (TOA) refers to the reflectance of the top of the atmosphere, and its value is the sum of the contribution of the surface reflectance and the atmospheric transmittance. Using the fixed calibration coefficients in the early stage of satellite operation, the pixel apparent reflectance is calculated from the remote sensor channel DN value:

[0092]

[0093] The slope i 、bias i are the fixed calibration slope and intercept corresponding to the i-th channel of any remote sensor, DN i is the DN value obtained by any remote sensor, d 2 is the correction factor for the distance between the sun and the earth, θ s is the solar zenith angle.

[0094] Regarding step 202, it should be noted that after the pixel apparent reflectance of a single pixel in each image pair is input into the invariant pixel detection model, the MAD variable needs to be constructed first. The process of constructing the MAD variable includes:

[0095] Step 2021, respectively use vector F=(F1…F n ) T , G=(G1…G n ) T Represents the apparent reflectance after DN value conversion in two multispectral images in an image pair, and linearly combines all spectral bands to obtain typical variables U and V, as shown in Equation 1:

[0096]

[0097] Where n is the total number of matching channels of the current image pair, and m and n are constant vectors, which can be obtained by solving the coupled generalized eigenvalue equation of Equation 2:

[0098]

[0099] Among them, ρ is the correlation coefficient of typical variables U and V, ∑ ff ,∑ gg ,∑ fg ,∑ gf is the covariance matrix of image vectors F and G.

[0100] Step 2022: Obtain the MAD variable based on the difference of the typical variables U and V, as shown in Formula 3:

[0101]

[0102] Among them, MAD i Represents the MAD variable of the i-th channel of the remote sensor.

[0103] Step 2023: Based on the linear scale invariance of the MAD variable, the observed value Z is obtained, as shown in Formula 4:

[0104]

[0105] in, is the standard deviation of the MAD variable of the i-th channel of the remote sensor, Z is the sum of squares of the standardized MAD variables, which is approximately a chi-square distribution (χ 2 ), there are n degrees of freedom. The smaller the Z value, the higher the probability that the pixel remains unchanged.

[0106] Regarding step 203, it should be noted that for two n-channel multispectral images of the same scene collected by two remote sensors at the same time in any image pair, ideally, the difference between them is caused by the instrument response itself and random effects such as noise and atmospheric fluctuations. From the central limit theorem, the MAD variable conforms to the ideal normal distribution. However, the MAD variable associated with the change observation will deviate from this multivariate normal distribution to a greater or lesser extent. In the presence of changes, the sensitivity of the MAD transform will be improved by focusing on successive iterations to establish a better and better background without changes.

[0107] Since a single MAD transformation is difficult to achieve an ideal invariant pixel detection effect, when estimating the sample mean and covariance matrix, the embodiment of the present invention weights the sample data by the invariant probability determined by a single iteration, assigns a larger weight to pixels with a higher invariant probability, determines the MAD variable for the next iteration by canonical correlation analysis, and performs multiple iterations to obtain a better invariant pixel detection effect.

[0108] Specifically, step 203 includes the following sub-steps:

[0109] Step 2031, based on the observed value, calculate the unchanged probability weight value;

[0110] It should be noted that for each iteration, the constant probability weight value Pr can be determined by the chi-square distribution test result of the observation value Z, formula 5:

[0111]

[0112] For the IR-MAD algorithm, it is necessary to set three iteration stop thresholds: the typical variable correlation coefficient change threshold, the maximum number of iterations, and the minimum number of NCPs, so that the algorithm stops iterating on a certain image pair when the algorithm optimization converges and the small probability detection fails, so as to ensure the automatic operation of the algorithm. When the difference between the changes in the two typical variable correlation coefficients is less than the typical variable correlation coefficient change threshold, the algorithm is considered to have converged and the iteration is stopped; at the same time, in order to ensure that there are enough invariant pixel samples for subsequent cross-calibration regression analysis, the minimum NCPs number threshold needs to be set. When the number of NCPs is less than this value after a certain iteration, the iteration stops.

[0113] Step 2032: Set an invariant probability decision threshold, and determine the invariant pixels of each image pair based on the observed values ​​of individual pixels in each image pair and the invariant probability decision threshold.

[0114] It should be noted that, since this embodiment seeks to find an unchanged pixel with a greater probability, when the number of matching channel groups n is determined, multiple unchanged probability confidences are tested and analyzed, and the algorithm operation efficiency and unchanged pixel detection effect are comprehensively considered to select an appropriate unchanged probability decision threshold t. When the observed value Z < t, it is considered that the confidence of the pixel sample as an unchanged pixel is higher than Can be used for subsequent cross-calibration analysis.

[0115] Therefore, in this embodiment, the samples satisfying the condition of formula 6 are selected as the invariant pixels (NCPs):

[0116]

[0117] Among them, t is the upper quantile of the chi-square distribution with n degrees of freedom, that is, the probability decision threshold of the unchanged pixel, P is the probability that the chi-square test value is less than or equal to t.

[0118] In addition, the spatial positions of NCPs selected by the statistical characteristics of the data without prior knowledge of the surface correspond to the invariant features between image pairs, but the positions of NCPs will change with the differences in the radiation information of different image pairs.

[0119] It can be understood that obtaining the spectral matching factor between the remote sensor to be calibrated and the reference remote sensor includes:

[0120] Obtain the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor;

[0121] The spectral matching factor of the remote sensor to be calibrated and the reference remote sensor is determined based on the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor.

[0122] It should be noted that the embodiment of the present invention uses the hyperspectral instrument observation data as the hyperspectral sample, and convolves it with the channel spectral response function of the remote sensor to be calibrated and the reference remote sensor, respectively, to obtain the entrance pupil radiance of the corresponding remote sensor, as shown in Formula 7:

[0123]

[0124] Among them, R h is the hyperspectral sample radiance, f i_sensor is the spectral response function of channel i of the hyperspectral instrument, R i_sensor is the entrance pupil radiance of the remote sensor to be calibrated or the reference remote sensor after convolution.

[0125] Under the same surface, atmospheric and observation time geometric conditions, the ratio of the incoming radiation of the matching channel of the remote sensor to be calibrated and the reference remote sensor is called the spectral matching factor (SBAF). The relationship between the pupil brightness of the matching channel of the remote sensor to be calibrated and the reference remote sensor is established as shown in Equation 8, and the SBAF coefficients of the corresponding channels of the two remote sensors are obtained:

[0126] R i_CAL =A i，j ×R j_REF +B i，j Formula 8

[0127] Where R represents the radiance, A and B are the spectral matching factors, i and j are the matching channel numbers of the remote sensor to be calibrated (CAL) and the reference remote sensor (REF), respectively. i，j and B i，j It represents the spectral matching factor between the i channel of the remote sensor to be calibrated and the j channel of the reference remote sensor. The SBAF of each matching channel is calculated by least squares regression fitting.

[0128] It can be understood that the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair is orthogonally regressed with the apparent reflectance of the to-be-calibrated remote sensor of the unchanged pixel of the corresponding image pair to determine the cross calibration coefficient, including:

[0129] A linear fitting relationship is established between the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair and the apparent reflectance of the uncalibrated remote sensor of the corresponding unchanged pixel of the image pair to perform orthogonal regression;

[0130] Based on the slope and intercept of the orthogonal regression, the cross calibration coefficient was determined.

[0131] It should be noted that for the unchanged pixels obtained by IR-MAD detection, the apparent reflectance detection result ρ of the j channel of the reference remote sensor is used. j_REF The spectral matching factor is used for conversion to obtain the spectrally corrected apparent reflectance ρ under the spectral response of channel ii_REF , as shown in formula 9:

[0132] ρ i_REF =A i，j ×ρ j_REF +B i，j Formula 9

[0133] For any image pair, the unchanged pixel obtained by IR-MAD detection is converted into the spectrum-corrected apparent reflectance ρ of channel i obtained by spectral matching. i_REF The apparent reflectivity detection result of the remote sensor to be calibrated ρ i_CAL A linear fitting relationship is established for orthogonal regression as shown in Equation 10, thereby obtaining the cross calibration coefficient:

[0134] ρ i_REF =a×ρ i_CAL +b Formula 10

[0135] Where a is the calibration slope and b is the calibration intercept. i_CAL is the apparent reflectivity calculated based on the fixed calibration slope at the initial stage of emission. Therefore, the calibration coefficients a and b are relative calibration results compared to the fixed calibration coefficients at the initial stage of emission.

[0136] It can be understood that determining the cross calibration coefficient also includes:

[0137] Performing long-time series cross-calibration on each channel of the long-time series data of the remote sensor to be calibrated;

[0138] Based on the result of the cross calibration of the long time series, a cross calibration coefficient of the long time series is determined.

[0139] It should be noted that the long-term cross-calibration of each channel of the remote sensor to be calibrated is used to monitor the long-term radiation response of the remote sensor channel and determine the trend of the long-term cross-calibration coefficient.

[0140] Provide the specific embodiment of the method of the present invention below:

[0141] Fengyun-3B (FY-3B) is the second satellite of China's second generation of polar-orbiting meteorological satellites, carrying 11 remote sensing instruments, of which the Visible Infrared Scanning Radiometer (VIRR) and the Moderate Resolution Imaging Spectrometer (MERSI) are two of the main instruments. FY-3B / VIRR has a total of 10 channel bands with a spectral range of 0.43-12.5μm, including 7 visible near-infrared bands and 3 infrared emission bands, and a subsatellite spatial resolution of 1.1km. VIRR is mainly used to detect and identify cloud information, detect surface vegetation cover, obtain surface sea surface related information, monitor the total amount of atmospheric water vapor, etc. FY3B / MERSI has 19 reflected solar bands (0.41-2.13μm) and 1 thermal infrared band (11.25μm), and the subsatellite spatial resolution is 250m and 1000m. The settings of VIRR channels 1, 2, 6, 7, 8, and 9 are similar to those of MERSI channels 3, 4, 6, 1, 10_11, and 2 (VIRR channel 8 corresponds to MERSI channels 10 and 11). The specific indicators are shown in Table 1. Therefore, MERSI is used as the reference sensor and VIRR is used as the sensor to be calibrated. A total of 6 groups of corresponding channels of VIRR are cross-calibrated.

[0142] like Figure 2 As shown, this embodiment screens and preprocesses the image data pairs of the remote sensor to be calibrated VIRR and the reference remote sensor MERSI of the same phase, linearly combines the matching channel data of the two remote sensors to construct a MAD variable, and obtains the invariant pixel samples in the scene through IR-MAD invariant pixel detection model analysis. The obtained invariant pixels use the detection results of the reference remote sensor as the radiation benchmark, and after spectral matching correction, they are orthogonally regressed with the detection results of the remote sensor to be calibrated to obtain the cross calibration coefficient.

[0143] Table 1 FY3B / VIRR, MERSI spectral band indicators

[0144]

[0145]

[0146] It can be understood that for step 101, the observation scene of the image pair sequence selected in this embodiment is the northwest region of China, with the central coordinates of 91° east longitude and 39° north latitude. The terrain is mainly plateaus, basins, and mountains, with an altitude of more than 1 km. The landforms are mainly the Loess Plateau, Gobi Desert, desert steppe and Gobi Desert. This area has a continental climate and is affected by the southern Himalayas. It is cold and dry in winter, hot in summer, and has little precipitation. Drought is the main natural feature of the area. The seasonal variation of the surface reflectivity in the study area is small, which meets the ideal cross-calibration observation area standards of low aerosol, low water vapor content and high clear sky probability.

[0147] As shown in Table 1, there are differences in the resolution of each channel between VIRR and MERSI. There are three spatial resolutions: 250m, 1km and 1.1km. The detection data of each channel are resampled to 1km spatial resolution and projected to the same geographic grid. The size of a single scene data is 1400*3200 pixels, and the time range is from January 21, 2011 to November 14, 2018.

[0148] It can be understood that, for step 101, in the DN values ​​of the simultaneous multispectral images acquired by VIRR and MERSI, there are some invalid points in each channel, and it is necessary to ensure that the data of each channel of each pixel used for statistical analysis are valid values, so the pixels with invalid values ​​are eliminated.

[0149] Since the scene with significant changes in the observation scene area is mainly cloud, and the L1 data of VIRR and MERSI do not have cloud mask data, the channel data of VIRR is used, and the threshold discrimination method is used for preliminary cloud identification. First, the radiation ratio of the two bands of VIRR channel 1 (0.58-0.68μm) and channel 2 (0.84-0.89μm) can effectively distinguish between clouds and clear sky areas, and the radiation ratio of channel 2 (0.84-0.89μm) and channel 6 (1.55-1.64μm) can better distinguish between ground clouds and high reflectivity objects on the ground. For the target of ocean water pixel, the land and ocean mask data in VIRR and MERSI data are used to eliminate it. The field of view scanning range of VIRR and MERSI remote sensors is ±55.4°, and the pixels with satellite observation zenith angle less than 30° are selected for subsequent statistical analysis.

[0150] It can be understood that, for step 102, the fixed calibration coefficient of the remote sensor in the early stage of the operation of the FY-3B satellite is used to calculate the pixel apparent reflectance from the channel DN value to obtain the apparent reflectance image of the research scene at the same phase.

[0151] It can be understood that, for step 102, for the two multispectral images of Northwest China acquired by VIRR and MERSI at the same time, after data screening and preprocessing, 6 groups of matching channel apparent reflectance data of sample pixels for IR-MAD analysis are retained, that is, n=6.

[0152] In this embodiment, IR-MAD analysis is performed on multiple sets of image data pairs to test and select appropriate iteration stop threshold parameters. When the number of iterations is 0-10, the number of unchanged pixels decreases rapidly, and the downward trend gradually slows down after 10-20 times. The algorithm tends to converge at 20-25 times. At this time, the change in the correlation coefficient of the typical variables between two iterations is less than 0.001. Therefore, it can be set that the algorithm converges and stops iteration when the change difference of the correlation coefficient of the two typical variables is less than 0.001, and the maximum number of iterations of the algorithm is set to 30. At the same time, in order to ensure that there are enough unchanged pixel samples for subsequent cross-calibration regression analysis, the minimum number of NCPs is set to 400. When the number of NCPs is less than this value after a certain iteration, the iteration stops.

[0153] In this embodiment, it is necessary to obtain unchanged pixels with an unchanged probability of more than 90%. When the degree of freedom n=6 (6 groups of matching channels), the unchanged probability confidence levels of 90%, 92.5%, 95%, and 97.5% are tested and analyzed. Considering the algorithm operation efficiency and the unchanged pixel detection effect comprehensively, the selection As the invariant probability decision threshold, when the observed value Z < t = 1.635, the confidence level of the pixel sample being an invariant pixel is higher than 95%, which can be used for subsequent cross-calibration analysis.

[0154] It can be understood that, for step 203, a relationship between the VIRR and MERSI matching channel entrance pupil brightness is established to obtain the SBAF coefficients of the corresponding channels of the two remote sensors, specifically including:

[0155] R i_VIRR =A i，j ×R j_MERSI +B i，j Formula 11

[0156] R 8_VIRR =A 8，10 ×R 10_MERSI +A 8，11 ×R 11_MERSI +B i，j Formula 12

[0157] Where R represents radiance, A and B are spectral matching factors, and i and j are the matching channel numbers of VIRR and MERSI, respectively. For the one-to-one matching channel between VIRR and MERSI, Equation 11 is used, and the 8th channel of VIRR and the 10th and 11th channels of MERSI are used in Equation 12. The SBAF of each matching channel is calculated by least squares regression fitting, and the spectral matching factors of VIRR and MERSI are shown in Table 2.

[0158] Table 2 Spectral matching factor (SBAF) between FY-3B VIRR and MERSI

[0159]

[0160] It can be understood that, for step 203, for the unchanged pixels obtained by IR-MAD detection, the apparent reflectance detection result ρ of the MERSI j channel is j_MERSI The spectral matching factor is used for conversion to obtain the spectrally corrected apparent reflectance ρ under the spectral response of channel i i_MERSI .

[0161] It can be understood that, for step 204, for any image pair, the invariant pixel obtained by IR-MAD detection is corrected by the i-channel spectrum obtained by spectral matching. i_MERSI The apparent reflectivity detection result ρ of VIRR i_VIRR A linear fitting relationship is established as shown in Formula 12 to obtain the cross calibration coefficient.

[0162] ρ i_MERSI =a×ρ i_VIRR +b Formula 12

[0163] Where a is the calibration slope and b is the calibration intercept.

[0164] It should be noted that the nominal orbital regression period of the Fengyun-3 satellite is 5.5 days, and the unchanged pixels obtained from 5 consecutive days of detection are merged for orthogonal regression analysis. Figure 4 These are the orthogonal regression results of the matching channels of the unchanged pixels from April 8 to 12, 2011. The linear fitting effect of the unchanged pixels in each channel is good, and the TOA dynamic range is large.

[0165] Figure 5 It is the long-term series trend of the relative calibration slope of each channel of VIRR from January 21, 2011 to November 14, 2018, which is used to monitor the long-term radiation response of the remote sensor channel. The method of the present invention is based on the principle of mathematical statistics. When using a single-day image pair for analysis, the recognition result of the unchanged pixel will be affected by the cloud size, water vapor aerosol content, and extreme weather conditions (such as heavy rain, snow, dust, etc.), which will lead to poor TOA regression effect of the unchanged pixel. Therefore, the correlation coefficient, residual and other regression quality evaluation indicators of the regression results of each channel can be used to eliminate the data pairs that do not meet the expected effect in the long-term series. For some data pairs, when only some of the channels are disturbed, considering that the method eliminates the correlation within the channels of the same phase sensor and the correlation between different channels of different sensors, and the calibration and analysis of the channels are independent, the effective channel data results of such data pairs can be retained. The long-term series results show that the method of the present invention can be implemented with nearly uninterrupted automatic cross-calibration.

[0166] The following is a description of the satellite remote sensor cross-calibration device based on invariant pixels provided by the present invention. The satellite remote sensor cross-calibration device based on invariant pixels described below and the satellite remote sensor cross-calibration method based on invariant pixels described above can be referenced to each other.

[0167] like Figure 6 As shown, the embodiment of the present invention discloses a satellite remote sensor cross calibration device based on invariant pixels, comprising:

[0168] An acquisition module 601 is used to determine a sequence of image pairs, each image pair comprising multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day;

[0169] The invariant pixel detection module 602 is used to obtain the apparent reflectance of a single pixel in each image pair and input it into an invariant pixel detection model to obtain the invariant pixel in each image pair;

[0170] The spectral matching module 603 is used to obtain the spectral matching factor of the remote sensor to be calibrated and the reference remote sensor, and perform spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factor to determine the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair;

[0171] The regression module 604 is used to perform orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair and the apparent reflectance of the corresponding remote sensor to be calibrated to determine a cross calibration coefficient.

[0172] The invariant pixel-based satellite remote sensor cross-calibration device of the embodiment of the present invention aims at the calibration problem of optical remote sensing imagers and the need for recalibration of historical data, and proposes a remote sensor cross-calibration method based on intelligent selection of invariant pixels. The method uses simultaneous satellite scene images acquired by two remote sensors, and after data screening and pre-processing, automatically detects invariant pixel targets in the scene through an invariant pixel detection model, calculates spectral matching factors to correct the spectral response differences between the two sensors, establishes a linear fitting relationship for the apparent reflectance of the invariant pixel samples, and obtains the cross-calibration results and their long-term series calibration trends.

[0173] It can be understood that determining the image pair sequence in the acquisition module 601 includes:

[0174] The multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day when the satellites pass by at the same time;

[0175] The multispectral image collected by the remote sensor to be calibrated and the multispectral image collected by the reference remote sensor are preprocessed to obtain a target image pair.

[0176] Based on the target image pairs collected over multiple days, an image pair sequence is determined.

[0177] It can be understood that preprocessing includes resolution resampling, rasterization and elimination of invalid pixels.

[0178] It is understandable that removing invalid pixels includes:

[0179] Eliminate cloud-contaminated pixels, water target pixels, and pixels with satellite observation zenith angles greater than or equal to 30°.

[0180] It can be understood that the invariant pixel detection model is established based on the iterative weighted multivariate change detection IR-MAD method.

[0181] It can be understood that the unchanged pixel detection module 602 includes:

[0182] Obtain the detection results of a single pixel in each image pair and calculate the apparent reflectance of the pixel;

[0183] Input the pixel apparent reflectance of a single pixel in each image pair into the invariant pixel detection model, construct a MAD variable, and construct the observation value of a single pixel in each image pair based on the MAD variable;

[0184] Based on the observed values ​​of individual pixels in each image pair and the invariant probability decision threshold, the invariant pixels in each image pair are determined.

[0185] It can be understood that the spectrum matching module 603 obtains the spectrum matching factor of the remote sensor to be calibrated and the reference remote sensor, including:

[0186] Obtain the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor;

[0187] The spectral matching factor of the remote sensor to be calibrated and the reference remote sensor is determined based on the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor.

[0188] It is understandable that the regression module 604 includes:

[0189] A linear fitting relationship is established between the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair and the apparent reflectance of the uncalibrated remote sensor of the corresponding unchanged pixel of the image pair to perform orthogonal regression;

[0190] Based on the slope and intercept of the orthogonal regression, the cross calibration coefficient was determined.

[0191] It is understandable that the regression module 604 also includes:

[0192] Carry out long-time series cross-calibration on each channel of the long-time series data of the remote sensor to be calibrated;

[0193] Based on the result of the cross calibration of the long time series, the cross calibration coefficient of the long time series is determined.

[0194] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A satellite remote sensor cross calibration method based on invariant pixels, characterized in that: include: Determine a sequence of image pairs, wherein each of the image pairs includes multispectral images respectively observed by the remote sensor to be calibrated and the reference remote sensor on the same observation scene at the same time on the same day; Obtaining the apparent reflectance of a single pixel in each image pair and inputting it into an invariant pixel detection model to obtain the invariant pixel in each image pair; Acquire spectral matching factors of the remote sensor to be calibrated and the reference remote sensor, perform spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factors, and determine the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair; Performing orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair and the apparent reflectance of the corresponding remote sensor to be calibrated to determine a cross calibration coefficient; The invariant pixel detection model is established based on the iterative weighted multivariate change detection IR-MAD method; The step of obtaining the apparent reflectance of a single pixel in each image pair and inputting the reflected reflectance into an invariant pixel detection model, and outputting the invariant pixel in each image pair comprises: Obtaining the detection result of a single pixel in each image pair and calculating the apparent reflectivity of the pixel; Inputting the pixel apparent reflectance of a single pixel in each image pair into an invariant pixel detection model, constructing a MAD variable and constructing an observation value of a single pixel in each image pair based on the MAD variable; Determine the unchanged pixels in each image pair based on the observed values ​​of the single pixels in each image pair and the unchanged probability decision threshold; The process of constructing a MAD variable includes: Use vectors respectively , Represents the apparent reflectance after DN value conversion in two multispectral images in an image pair, and linearly combines all spectral bands to obtain typical variables U and V, as shown in Equation 1: Formula 1; in, Indicates the total number of matching channels for the current image pair, and is a constant vector, which can be obtained by solving the coupled generalized eigenvalue equation of Equation 2: Formula 2; in, is the correlation coefficient of the typical variables U and V, , , , is the image vector , The covariance matrix of The MAD variable is obtained based on the difference of the typical variables U and V, as shown in Formula 3: Formula 3; in, represents the MAD variable of the i-th channel of the remote sensor; Step 2023: Based on the linear scale invariance of the MAD variable, the observed value Z is obtained, as shown in Formula 4: Formula 4; in, is the standard deviation of the MAD variable of the i-th channel of the remote sensor, Z is the sum of squares of the standardized MAD variables, there are n degrees of freedom, the smaller the Z value, the higher the probability that the pixel remains unchanged.

2. The satellite remote sensor cross calibration method based on invariant pixels according to claim 1, characterized in that: The determining of the image pair sequence comprises: The multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day when the satellites pass by at the same time; Preprocessing the multispectral image collected by the remote sensor to be calibrated and the multispectral image collected by the reference remote sensor to obtain a target image pair; The image pair sequence is determined based on the target image pairs collected over multiple days.

3. The satellite remote sensor cross calibration method based on invariant pixels according to claim 2, characterized in that: The preprocessing includes resolution resampling, rasterization and elimination of invalid pixels.

4. The satellite remote sensor cross calibration method based on invariant pixels according to claim 3 is characterized in that: The removal of invalid pixels comprises: Eliminate cloud-contaminated pixels, water target pixels, and pixels with satellite observation zenith angles greater than or equal to 30°.

5. The satellite remote sensor cross calibration method based on invariant pixels according to claim 1, characterized in that: The step of obtaining the spectral matching factor of the remote sensor to be calibrated and the reference remote sensor includes: Acquire the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor; Based on the entrance pupil radiance of the remote sensor to be calibrated and the entrance pupil radiance of the reference remote sensor, a spectral matching factor between the remote sensor to be calibrated and the reference remote sensor is determined.

6. The satellite remote sensor cross calibration method based on invariant pixels according to claim 1, characterized in that: The step of performing orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair and the apparent reflectance of the to-be-calibrated remote sensor of the unchanged pixel of the corresponding image pair to determine the cross calibration coefficient comprises: Establishing a linear fitting relationship between the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel of each image pair and the apparent reflectance of the to-be-calibrated remote sensor of the unchanged pixel of the corresponding image pair to perform orthogonal regression; Based on the slope and intercept of the orthogonal regression, the cross calibration coefficient was determined.

7. The satellite remote sensor cross calibration method based on invariant pixels according to claim 1, characterized in that: The determining of the cross calibration coefficient further comprises: Performing long-time series cross-calibration on each channel of the long-time series data of the remote sensor to be calibrated; Based on the result of the cross calibration of the long time series, a cross calibration coefficient of the long time series is determined.

8. A satellite remote sensor cross-calibration device based on invariant pixels for executing the satellite remote sensor cross-calibration method based on invariant pixels as claimed in any one of claims 1 to 7, characterized in that: include: An acquisition module is used to determine a sequence of image pairs, each of which includes multispectral images obtained by the remote sensor to be calibrated and the reference remote sensor respectively observing the same observation scene at the same time on the same day; An invariant pixel detection module, used to obtain the apparent reflectance of a single pixel in each image pair and input it into an invariant pixel detection model to obtain the invariant pixel in each image pair; A spectral matching module, used to obtain spectral matching factors of the remote sensor to be calibrated and the reference remote sensor, perform spectral matching on the apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair based on the spectral matching factors, and determine the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair; The regression module is used to perform orthogonal regression on the spectrally corrected apparent reflectance of the reference remote sensor of the unchanged pixel in each image pair and the apparent reflectance of the corresponding remote sensor to be calibrated to determine the cross calibration coefficient.

Images