A comprehensive evaluation method for signal-to-noise ratio of remote sensing image data taking into account the influence of background noise

Abstract

The present invention discloses a comprehensive evaluation method for the signal-to-noise ratio of remote sensing image data taking into account the influence of background noise, and relates to the field of remote sensing image technology. The method comprises the following steps: constructing a grouped data set; constructing a surface reflectance grouping; constructing a typical sample grouping; designing and selecting a pseudo-invariant calibration field and optimizing the pseudo-invariant calibration field; collecting data, constructing an image data set and optimizing the ground target of the pseudo-invariant calibration field; extracting the image area for calculating the SNR, constructing and optimizing the image data sub-block set; calculating the SNR, and analyzing and verifying the SNR of the image data of the same sensor and different sensors. The present invention makes full use of the characteristics of the pseudo-invariant calibration field, improves the accuracy and reliability of the SNR evaluation, and provides an important reference for the processing and application of remote sensing image data.

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing images, and in particular relates to a comprehensive evaluation method for signal-to-noise ratio of remote sensing image data taking into account the influence of background noise. Background Art

[0002] The signal-to-noise ratio (SNR) of remote sensing images is a commonly used parameter for evaluating image quality. It reflects the radiation performance of the remote sensing imaging system and the anti-interference ability of the imaging device. It is an important indicator for measuring the imaging quality and radiation performance of the sensor, and is also an important factor in determining whether the target can be detected. The larger the signal-to-noise ratio, the better the image quality. The noise in the image data is mainly additive noise. The same level of noise has different effects on images with different signal strengths, and generally has a stronger impact on weak signal images than strong signal images. Accurately estimating the noise level of remote sensing images can provide prior parameters for subsequent data processing algorithms.

[0003] The essence of signal-to-noise ratio evaluation is to extract the signal mean and background standard deviation from the image data. The commonly used remote sensing image signal-to-noise ratio evaluation algorithm is the variance method, which selects an image area, calculates the mean and standard deviation of the pixel DN value, and gives the signal-to-noise ratio. This method is closely related to the selected image data area, and it is difficult to ensure the uniformity and stability of the data, and the SNR result obtained is highly uncertain. Summary of the invention

[0004] The purpose of the present invention is to provide a comprehensive evaluation method for the signal-to-noise ratio of remote sensing image data taking into account the influence of background noise. The method introduces pseudo-invariant calibration field data of typical ground samples as a restriction condition, and uses multiple criteria to compare and analyze the signal-to-noise ratio. The signal-to-noise ratio evaluation method has high reliability and comprehensiveness.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a method for comprehensively evaluating the signal-to-noise ratio of remote sensing image data taking into account the influence of background noise, comprising the following steps:

[0007] S1: Segment and group the remote sensing image data according to the grayscale value (DN) to construct a remote sensing image DN value grouping dataset {GroupDN};

[0008] S2: Referring to the remote sensing image DN value grouping dataset {GroupDN}, according to the sensor design indicators and radiation transfer model, combined with the prior knowledge of radiation calibration, the surface reflectance grouping dataset {GroupReflectance} is constructed;

[0009] S3: Based on the surface reflectance grouping data, typical sample classification design and selection are performed to construct a typical sample grouping data set {GroupSamples};

[0010] S4: Select the ground pseudo-invariant calibration field according to the surface reflectance grouping data set and the typical sample grouping data set, and construct the preliminary pseudo-invariant calibration field data set {AllPICS0};

[0011] S5: Use high calibration accuracy image data to evaluate the uniformity and stability of the pseudo-invariant calibration field data, and obtain the optimized pseudo-invariant calibration field data set {AllPICS1};

[0012] S6: For the ground targets in the pseudo-invariant calibration field, a combination of existing data screening and programmatic data acquisition is used to collect data and construct a remote sensing image dataset {AllImages0};

[0013] S7: Based on the remote sensing image dataset {AllImages0}, atmospheric parameters, solar altitude angle parameters and observation geometry parameters are introduced, and the data are screened again according to the set threshold to obtain the optimized remote sensing image dataset {AllImages1};

[0014] S8: Based on the remote sensing image dataset {AllImages1}, according to the geographic scope of the pseudo-invariant calibration field dataset {AllPICS1}, the image area used to calculate the SNR is extracted according to the spatial geometric relationship of the image data, and the remote sensing image data sub-tile set {AllImageTiles0} is constructed;

[0015] S9: further evaluating the spatial uniformity of the remote sensing image data based on the remote sensing image data sub-block set {AllImageTiles0}, optimizing the image data sub-block set and its range, and obtaining the optimized remote sensing image data sub-block set {AllImageTiles1};

[0016] S10: Calculate the SNR of the remote sensing image data sub-block set {AllImageTiles1} to obtain the remote sensing image sub-block signal-to-noise ratio data set {AllSNRs};

[0017] S11: Analyze and verify the SNR of the image data of the same sensor, take the grayscale mean of the image data sub-block as the independent variable, and analyze the variation of SNR with grayscale value;

[0018] S12: Analyze the SNR of the time series image data of the same sensor, detect the change trend of sensor performance, and verify the reliability of SNR;

[0019] S13: Analyze and verify the SNR of image data from different sensors, and comprehensively evaluate the SNR performance of different sensors.

[0020] As a preferred technical solution of the present invention, when the remote sensing image data is segmented and grouped according to the grayscale value (DN) in S1, the segmented grouping design covers the sensor SNR design indicator range and includes the low and high ends of the DN value; for image data containing multiple bands, the bands are processed separately.

[0021] As a preferred technical solution of the present invention, when constructing the surface reflectivity grouping data set in S2, typical reflectivity groups can be set to 5-10%, 20%, 30%, 40%, 50%, 60%.

[0022] As a preferred technical solution of the present invention, when designing and selecting the typical sample grouping data set in S3, artificial targets and stable lakes, deserts and grasslands with good uniformity are selected to serve as targets and as typical samples.

[0023] As a preferred technical solution of the present invention, the pseudo-invariant calibration field data set {AllPICS1} in S5 is stored in the following manner:

[0024] PICS1ID, center position, calibration field range (*.shp), focus target range (*.shp), reflectivity, ...

[0025] PICS1ID, center position, calibration field range (*.shp), focus target range (*.shp), reflectivity, ...

[0026] ….

[0027] As a preferred technical solution of the present invention, in S6, the existing data is screened by querying with the given pseudo-invariant calibration field range, cloud cover parameters and seasonal characteristics as constraint conditions to collect the existing calibration field data; for programmed data acquisition, it is necessary to arrange the sensor data acquisition task according to the sensor transit time forecast parameters, the geographic coordinates of the ground pseudo-invariant field target, and the solar altitude angle, observation angle and meteorological factors.

[0028] As a preferred technical solution of the present invention, the grayscale mean of the image data sub-block in S10 is mean, the standard deviation of the image data sub-block is stddev, and the calculation formula of SNR is:

[0029]

[0030] As a preferred technical solution of the present invention, in S11, when the main imaging state parameters of the same sensor do not change, a statistical time period within one year is selected, and the grayscale mean of the image data sub-block is used as the independent variable to analyze the change pattern of SNR with grayscale value.

[0031] As a preferred technical solution of the present invention, in S12, when the main imaging state parameters of the same sensor have not changed, the SNR data obtained in each year are compared to detect the change trend of the sensor performance.

[0032] As a preferred technical solution of the present invention, when performing SNR analysis and verification on image data from different sensors in S13, the radiation transmission model is used to convert the image data to the atmospheric apparent reflectivity to analyze the SNR variation law and conduct a horizontal comparison analysis of the SNR.

[0033] The present invention has the following beneficial effects:

[0034] (1) A pseudo-invariant calibration field database of typical ground samples was constructed to ensure the availability of remote sensing image data used for evaluation and improve the reliability of signal-to-noise ratio evaluation;

[0035] (2) Various data optimization strategies, such as atmospheric parameter screening and image data spatial uniformity assessment, were used to further improve the image dataset used for signal-to-noise ratio evaluation;

[0036] (3) Through time series analysis and cross-comparison analysis of the SNR data set, the SNRs were cross-verified again to ensure the reliability of the signal-to-noise ratio evaluation.

[0037] At the same time, compared with the prior art, the SNR evaluation method proposed in the present invention can have the following uses:

[0038] (1) Analyze the SNR differences of image data acquired under different imaging state parameters;

[0039] (2) Analyze the advantages and disadvantages of SNR between different relative radiation correction products;

[0040] (3) Analyze the SNR of time series image data and compare longitudinally whether the sensor performance is degraded;

[0041] (4) Horizontally compare the SNR performance differences between different sensors.

[0042] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0044] Figure 1 It is a specific flow chart of the patent of this invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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] Embodiment 1

[0047] like Figure 1 As shown: The present invention discloses a comprehensive evaluation method for the signal-to-noise ratio of remote sensing image data taking into account the influence of background noise, comprising the following steps:

[0048] S1: The remote sensing image data is segmented and grouped according to the grayscale value (DN), and a remote sensing image DN value grouping dataset {GroupDN} is constructed. When the remote sensing image data is segmented and grouped according to the grayscale value (DN), the segmentation and grouping design covers the sensor SNR design index range and includes the low-end and high-end parts of the DN value; the DN value here refers to the brightness value of the remote sensing image pixel, which records the grayscale value of the ground object. It is the input data for calculating the SNR. For image data containing multiple bands, each band is processed separately;

[0049] S2: Referring to the remote sensing image DN value grouping dataset {GroupDN}, according to the sensor design indicators and radiation transfer model, combined with the prior knowledge of radiation calibration, the surface reflectance grouping dataset {GroupReflectance} is constructed. The typical reflectance grouping can be set to 5-10%, 20%, 30%, 40%, 50%, 60%;

[0050] S3: Based on the surface reflectivity grouping data, typical sample classification design and selection are carried out to construct a typical sample grouping data set {GroupSamples}. Typical samples include artificial targets and relatively stable lakes, deserts, grasslands and other targets with good uniformity;

[0051] S4: According to the surface reflectance grouping and typical sample classification design, the ground pseudo-invariant calibration field is selected preferentially, and the uniformity and stability of the field reflectance and the atmospheric conditions are required to be good, so as to obtain the preliminary pseudo-invariant calibration field data set {AllPICS0}. Domestic remote sensing comprehensive verification field includes Baotou Remote Sensing Comprehensive Verification Field, Dunhuang Radiation Calibration Field, Qinghai Lake Radiation Calibration Field, Gongger Grassland Radiation Calibration Field in Inner Mongolia, etc.; foreign remote sensing comprehensive experimental fields include White Sands Radiation Calibration Field in the United States, Railroad Valley Radiation Calibration Field, La Crau Radiation Calibration Field in France, Lake Frome Calibration Field in Australia, and Algeria 3, Algeria 5, Libya 1, Libya 4, Mauritania 1 and Mauritania2 calibration fields in Africa;

[0052] S5: Use high calibration accuracy image data to evaluate the uniformity and stability of the pseudo-invariant calibration field data, such as using Sentinel-2A / B, Landsat-8 / 9OLI and other data, and further screen and verify the pseudo-invariant calibration field according to the set threshold to obtain the optimized pseudo-invariant calibration field dataset {AllPICS1}; the pseudo-invariant calibration field dataset {AllPICS1} is stored as follows:

[0053] PICS1ID, center position, calibration field range (*.shp), focus target range (*.shp), reflectivity, ...

[0054] PICS1ID, center position, calibration field range (*.shp), focus target range (*.shp), reflectivity, ...

[0055] …;

[0056] S6: For the ground targets of the pseudo-invariant calibration field, a combination of existing data screening and programmatic data acquisition is used to collect data and construct a remote sensing image dataset {AllImages0}; for existing data screening, it can be queried according to the given pseudo-invariant calibration field range, cloud parameters, seasonal characteristics and other constraints to collect the existing calibration field data; for programmatic data acquisition, it is necessary to arrange the sensor data acquisition task according to the sensor transit time forecast parameters, the geographic coordinates of the ground pseudo-invariant field targets, and the solar altitude angle, observation angle, meteorological factors, etc.;

[0057] S7: Based on the remote sensing image dataset {AllImages0}, atmospheric parameters, solar altitude parameters and observation geometry parameters are introduced, and the data are screened again according to the set threshold to obtain the optimized remote sensing image dataset {AllImages1}. The atmospheric parameters used for remote sensing screening image datasets can be extracted from the products of the European Center for Medium-Term Weather Forecasts ECMWF. The solar altitude parameters can be calculated based on the local latitude, solar declination and hour angle. The sensor observation geometry parameters are usually given in the image data auxiliary parameters.

[0058] S8: Based on the remote sensing image dataset {AllImages1}, according to the geographic scope of the pseudo-invariant calibration field dataset {AllPICS1}, the image area used to calculate the SNR is extracted according to the spatial geometric relationship of the image data, and the remote sensing image data sub-block set {AllImageTiles0} is constructed. In this way, the geographic location alignment is achieved for multiple revisited data. In some cases, ground control point data is needed to obtain more accurate ground overlap;

[0059] S9: further evaluate the spatial uniformity of the remote sensing image data based on the remote sensing image data sub-block set {AllImageTiles0}, optimize the image data sub-block set and its range, and obtain the optimized remote sensing image data sub-block set {AllImageTiles1}. Generally, the sliding window method can be used to evaluate the spatial uniformity of the image data. It is necessary to set the window size and uniformity threshold, wherein the uniformity threshold needs to consider the relative radiation correction accuracy of the remote sensing image data;

[0060] S10: Calculate the SNR of the remote sensing image data sub-block set {AllImageTiles1} to obtain the remote sensing image sub-block signal-to-noise ratio data set {AllSNRs}, using the following formula

[0061]

[0062] In the above formula, mean is the grayscale mean of the image data sub-block, stddev is the standard deviation of the image data sub-block, and the unit of SNR is decibel (db);

[0063] S11: Analyze and verify the SNR of the image data of the same sensor. When the main imaging state parameters of the same sensor have not changed, select a shorter statistical period (such as within one year), take the grayscale mean of the image data sub-block as the independent variable, and analyze the change law of SNR with grayscale value.

[0064] In general, sensor noise is an additive noise, which can be expressed as follows:

[0065] ρ s (λ)=ρ t(λ)+ε(λ)

[0066] Where λ is the wavelength, ρ s represents the TOA reflectivity recorded by the sensor, ρ t is s Theoretically true value, ε is the sensor noise;

[0067] The signal-to-noise ratio will increase with the increase of the incident radiation signal, that is, the signal-to-noise ratio will increase with the increase of the DN value, and the SNR of the target with high surface reflectivity will be larger;

[0068] S12: SNR analysis and verification of time series image data of the same sensor. When the main imaging state parameters of the same sensor remain unchanged, the SNR data obtained each year are compared to show the change in sensor performance. Generally, a good sensor has stable performance and no obvious degradation during its design life.

[0069] S13: SNR analysis and verification of image data from different sensors. For image data obtained by different sensors, it is necessary to use the radiation transfer model to convert the image data to the atmospheric surface reflectance (Top of AtmosphereReflectance). Based on this, the SNR variation law is analyzed, the SNR is compared horizontally, and the SNR performance of different sensors is comprehensively evaluated.

[0070] The verification obtained during the specific implementation of the present invention is as follows:

[0071] Main performance indicators of remote sensing image data:

[0072] Image data bands: rear-view panchromatic image, front-view panchromatic image, rear-view multispectral image

[0073] Image data grayscale range: 0~2047

[0074] Image data pixel size: Panchromatic image: 0.65~0.8m; Multispectral image: 2.6m

[0075] When the solar altitude angle is 70° and the ground reflectivity is 0.65, SNR ≥ 48dB;

[0076] When the solar altitude angle is 20° and the ground reflectivity is 0.05, SNR is ≥ 23dB.

[0077] This embodiment selects the rear-view panchromatic image, front-view panchromatic image, and rear-view multispectral image of the Gaofen-7 satellite as experimental objects. The image data is acquired by programming during the in-orbit operation of the Gaofen-7 satellite and is located in the Baotou remote sensing comprehensive verification field. It includes artificial targets of three reflectivities as pseudo-invariant calibration field data, and experimental statistics are performed on the signal-to-noise ratio of the remote sensing image data of this embodiment.

[0078] Signal-to-noise ratio of remote sensing image data

[0079] For the selected data, the signal-to-noise ratio comprehensive evaluation method of remote sensing image data taking into account the background noise proposed in this embodiment is used to conduct a comprehensive evaluation of the signal-to-noise ratio of the artificial target image data acquired by the Gaofen-7 satellite. According to statistical analysis, when the DN of the rear-view panchromatic image and the front-view panchromatic image is around 700 (low end), the SNR is better than 35dB, and when the DN is around 1900 (high end), the SNR exceeds 48dB. The SNR of the DN values ​​of the four bands of the rear-view multispectral image at the low and high ends are better than 40dB and 48dB.

[0080] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0081] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A comprehensive evaluation method for signal-to-noise ratio of remote sensing image data taking into account the influence of background noise, characterized in that: The following steps are involved: S1: Segment and group the remote sensing image data according to the grayscale value (DN) to construct a remote sensing image DN value grouping dataset {GroupDN}; S2: Referring to the remote sensing image DN value grouping dataset {GroupDN}, according to the sensor design indicators and radiation transfer model, combined with the prior knowledge of radiation calibration, the surface reflectance grouping dataset {GroupReflectance} is constructed; S3: Based on the surface reflectance grouping data, typical sample classification design and selection are performed to construct a typical sample grouping data set {GroupSamples}; S4: Select the ground pseudo-invariant calibration field according to the surface reflectance grouping data set and the typical sample grouping data set, and construct the preliminary pseudo-invariant calibration field data set {AllPICS0}; S5: Use high calibration accuracy image data to evaluate the uniformity and stability of the pseudo-invariant calibration field data, and obtain the optimized pseudo-invariant calibration field data set {AllPICS1}; S6: For the ground targets in the pseudo-invariant calibration field, a combination of existing data screening and programmatic data acquisition is used to collect data and construct a remote sensing image dataset {AllImages0}; S7: Based on the remote sensing image dataset {All Images0}, atmospheric parameters, solar altitude angle parameters and observation geometry parameters are introduced, and the data are screened again according to the set threshold to obtain the optimized remote sensing image dataset {All Images1}; S8: Based on the remote sensing image dataset {AllImages1}, according to the geographic scope of the pseudo-invariant calibration field dataset {AllPICS1}, the image area used to calculate the SNR is extracted according to the spatial geometric relationship of the image data, and the remote sensing image data sub-tile set {AllImageTiles0} is constructed; S9: further evaluating the spatial uniformity of the remote sensing image data based on the remote sensing image data sub-block set {AllImageTiles0}, optimizing the image data sub-block set and its range, and obtaining the optimized remote sensing image data sub-block set {AllImageTiles1}; S10: Calculate the SNR of the remote sensing image data sub-block set {AllImageTiles1} to obtain the remote sensing image sub-block signal-to-noise ratio data set {AllSNRs}; S11: Analyze and verify the SNR of the image data of the same sensor, take the grayscale mean of the image data sub-block as the independent variable, and analyze the variation of SNR with grayscale value; S12: Analyze the SNR of the time series image data of the same sensor, detect the change trend of sensor performance, and verify the reliability of SNR; S13: Analyze and verify the SNR of image data from different sensors, and comprehensively evaluate the SNR performance of different sensors.

2. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: When the remote sensing image data is segmented and grouped according to the grayscale value (DN) in S1, the segmented grouping design covers the sensor SNR design index range and includes the low-end and high-end parts of the DN value. For image data containing multiple bands, the bands are processed separately.

3. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: When constructing the surface reflectivity grouping data set in S2, typical reflectivity groups may be set to 5-10%, 20%, 30%, 40%, 50%, and 60%.

4. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: When designing and selecting the typical sample grouping data set in S3, artificial targets, lakes, deserts and grasslands are selected to serve as targets and as typical samples.

5. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: In the above S6, the existing data is screened by querying with the given pseudo-invariant calibration field range, cloud cover parameters and seasonal characteristics as constraint conditions to collect the existing calibration field data; for programmed data acquisition, it is necessary to arrange the sensor data acquisition task according to the sensor transit time forecast parameters, the geographic coordinates of the ground pseudo-invariant field target, and the solar altitude angle, observation angle and meteorological factors.

6. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: The grayscale mean of the image data sub-block in S10 is mean, the standard deviation of the image data sub-block is stddev, and the calculation formula of SNR is:

7. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: In the above S11, when the main imaging state parameters of the same sensor have not changed, a statistical time period within one year is selected, and the grayscale mean of the image data sub-block is used as the independent variable to analyze the variation of SNR with grayscale value.

8. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: In the S12, when the main imaging state parameters of the same sensor remain unchanged, the SNR data acquired in each year are compared to detect the change trend of the sensor performance.

9. The method for comprehensive evaluation of signal-to-noise ratio of remote sensing image data taking into account the influence of background noise according to claim 1, characterized in that: When performing SNR analysis and verification on image data from different sensors in S13, the image data is converted to the atmospheric apparent reflectivity using a radiation transfer model to analyze the SNR variation law and conduct a horizontal comparison analysis of the SNR.

Images