A method and system for simulating vibration changes of remote sensing images
Through the flutter change simulation method of remote sensing images, the impact of matching window size and TDI CCD integration series on satellite flutter detection is analyzed, and the problems of insufficient quantitative analysis and flutter image shift change in the TDI CCD integration process in the prior art are solved. , improves vibration detection accuracy and image quality.
Patent Information
- Application Number
- CN202111048757.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-08
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-09-08
AI Technical Summary
In the flutter detection of satellite platform, the impact of the matching window size on the highest frequency of flutter detection has not been quantitatively analyzed, and the flutter image shift changes with time during the TDI CCD integration process, resulting in the image quality being affected and the error is relatively high. big.
The flutter change simulation method of remote sensing images is adopted, and the disparity value and geometric displacement value are obtained through simulation image generation, image matching and Fourier series calculation, and the influence of matching window size and TDI CCD integral series on flutter detection is analyzed. , construct a flutter curve expression to obtain the flutter frequency and amplitude.
The analysis of the flutter detection frequency and amplitude of the matching window size and integral stages is realized, providing a beneficial reference for flutter detection based on remote sensing images, and improving the detection accuracy and image quality.
Smart Images

Figure CN113870321B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of satellite remote sensing images, and in particular to a method and system for simulating vibration changes of remote sensing images. Background Art
[0002] With the continuous improvement of the spatial resolution of satellite remote sensing images, the impact of platform jitter on imaging quality has become increasingly prominent. Many international satellites have been affected by jitter, such as the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) on the Terra satellite, the High Resolution Imaging Science Experiment (HiRISE) camera on the Mars Reconnaissance Orbiter (MRO), the narrow-angle camera on the Lunar Reconnaissance Orbiter (LRO), Japan's Advanced Land Observing Satellite (ALOS), China's Resources-3 Satellite (ZY-3), Tianhui-1 Satellite, Yaogan-26 Satellite, Gaofen-9 Satellite, etc.
[0003] The accurate detection of satellite platform flutter is the key to eliminating deformation and other problems caused by platform flutter, and plays a very important role in compensating for the impact of flutter and improving image quality. At present, satellite platform flutter detection can be divided into two methods. One is to directly measure platform flutter using high-performance attitude measurement sensors, such as star sensors, gyroscopes, and angular displacement sensors (ADS). Flutter detection based on attitude sensor measurement is not affected by external conditions, is simple to implement and easy to analyze, and has strong versatility, but the detection accuracy mainly depends on the measurement accuracy and frequency of the attitude sensor itself. The second type is based on remote sensing image inversion of platform attitude flutter. Since on-orbit satellites are basically equipped with multispectral or panchromatic sensors, this method has strong versatility and is easy to implement, but the accuracy of dense matching will affect the accuracy of flutter detection.
[0004] High-precision and robust dense matching is the key to image-based flutter detection. Currently, many scholars have developed effective matching methods to obtain the image parallax caused by flutter, and then invert the satellite attitude flutter. In the existing studies on flutter detection based on image matching, it is generally believed that the matching window is equivalent to a low-pass filter, and the size of the window will affect the highest detectable frequency of flutter. However, there are few studies on the quantitative experimental analysis of the matching window size on flutter detection.
[0005] In addition, for high-resolution optical satellites, their imaging payloads mostly use TDI CCD (Time Delay and Integration Charge Coupled Devices) for photoelectric conversion imaging, and the scan lines they finally output are all images synthesized by multi-level CCD sensor integration. During the TDI CCD multi-level integration process, the image shift of each level of imaging changes with time, so the impact of the jitter on the synthesized image is no longer the same as that of the single-level integrated image. However, existing image-based jitter detection and estimation research has not considered the TDI mode, and it is assumed that each image line in the image is acquired within the line exposure time. Based on this assumption, the jitter information obtained by detection and estimation will have errors.
[0006] In order to study the influence of matching window size and TDI CCD integration level on flutter detection, a flutter change simulation technology of remote sensing image is urgently needed for the above research. Summary of the invention
[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a remote sensing image vibration change simulation method and system for studying the influence of matching window size and TDI CCD integration series on vibration detection.
[0008] The purpose of the present invention can be achieved by the following technical solutions:
[0009] A method for simulating vibration changes of remote sensing images comprises the following steps:
[0010] Simulation image generation steps: input an orthophoto image as a reference image, set track parameters, camera parameters and flutter parameters, and generate a simulation image based on a digital simulation model of dynamic imaging;
[0011] Image matching step: performing image registration on the simulation images at adjacent moments;
[0012] The vibration calculation steps based on Fourier series are as follows: matching two simulated images with a fixed imaging time interval to obtain a disparity value, modeling the disparity value using Fourier series, and obtaining the frequency and amplitude of the disparity value;
[0013] According to the differential relationship between the parallax value and the geometric displacement value, the amplitude and frequency of the jitter are obtained to describe the jitter changes of the remote sensing image.
[0014] Furthermore, the Fourier series expansion formula of the disparity value and the geometric displacement value is:
[0015]
[0016]
[0017]
[0018] in, α m =|sinc(N m Tf)|,α n =|sinc(N n Tf)|;
[0019] In the formula, a 0 ,a i ,b i (i=1,2,...k) are the coefficients of the Fourier series expansion formula of the parallax value, k is the order of the Fourier series, f is the frequency of the parallax value, and according to the properties of the same frequency combined vibration, it can be known that the flutter frequency is also f, j tdi (t) is the geometric displacement value caused by attitude flutter on the TDI image plane at time t, j tdi (t+Δt) is the geometric displacement value caused by attitude flutter in the TDI image plane at time t+Δt, t is the imaging time, Δt is the imaging time interval, is the parallax value of the simulated image corresponding to the imaging time t, T is the single-level integration time, N m is the reference image integral series, N n is the integral series of the image to be matched, A i , B i are the coefficients of the Fourier series expansion of the geometric displacement value, i = 1, 2, ... k, and N is the integral series.
[0020] Furthermore, the calculation expression of the coefficient of the Fourier series of the geometric displacement value caused by the flutter is:
[0021]
[0022]
[0023] in, w = 2πf;
[0024] Parameter a i and b i according to The data are obtained through fitting.
[0025] After obtaining the coefficients of the flutter Fourier series, the flutter amplitude can be obtained.
[0026]
[0027] Furthermore, in the image matching step, under matching windows of different sizes, the image registration is performed by normalized cross-correlation and quadratic parabola fitting methods to obtain disparity difference values and build a model, and according to the relationship between the disparity value and the geometric displacement value caused by the vibration, the vibration frequency amplitude is derived and compared with the actual vibration frequency, and the influence of matching windows of different sizes on the vibration detection frequency is obtained.
[0028] Furthermore, the digital simulation model of dynamic imaging adopts a TDI integral imaging model, which is used to simulate a multi-level CCD integral imaging process;
[0029] The remote sensing image vibration change simulation method also includes: according to the TDI integral image imaging principle, the platform vibration of each level of imaging changes with time during the multi-level integration process, constructing an expression of the vibration curve after TDI integration, and obtaining the relationship between the TDI integral and the single-level integral vibration amplitude.
[0030] Furthermore, the expression of the flutter curve of the synthesized image is:
[0031]
[0032] Where, d TDI is the chatter curve of the image after TDI integration, N is the integration level, L is the line number, A is the single-level integration chatter amplitude, f is the frequency, t is the imaging time, is the phase, T is the single-stage integration time;
[0033] The relationship between the TDI integrated vibration amplitude and the single-stage integrated vibration amplitude is expressed as follows:
[0034] A TDI =A|sinc(Ntf)|
[0035] In the formula, A TDI It is the jitter amplitude of the image synthesized by TDI integration.
[0036] Furthermore, the expression of the TDI integral imaging model is:
[0037]
[0038]
[0039] Where I(r,c) is the grayscale value of row r and column c, r is the row number, c is the column number, N is the integration level, a is the sampling interval of CCD in the x and y directions, T is the single-level integration time, v is the push-scan speed during linear array push-scan imaging, t is the imaging time, and y is the integration level of the CCD. jitter(t) is the vibration in the vertical rail direction at time t, A is the single-stage integrated vibration amplitude, f is the frequency, For phase.
[0040] Furthermore, in the image matching step, the image registration specifically includes the following steps:
[0041] Obtain two simulated images at adjacent moments, recorded as image M and image S;
[0042] Use Wallis filtering to enhance images and improve image contrast;
[0043] The image M and the image S are translated to achieve initial registration and eliminate fixed offset;
[0044] For each point to be matched on image M, the pixel coordinates with the largest normalized mutual correlation coefficient on image S and the normalized mutual correlation values of the pixels in the matching window centered on the pixel coordinates are calculated. The normalized mutual correlation values in the matching window are fitted using a quadratic parabola, and the pixel coordinates at the vertex of the quadratic parabola are the sub-pixel matching results.
[0045] Furthermore, the image registration process also includes a step of removing mismatched points:
[0046] Calculate the normalized mutual correlation coefficient between matching windows and remove the areas where the correlation coefficient is less than a preset correlation threshold;
[0047] The disparity of the matching point pairs of the image M and the image S is counted, and the matching points whose offset is greater than the preset disparity threshold are eliminated.
[0048] The present invention also provides a remote sensing image vibration change simulation system, comprising a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the remote sensing image vibration change simulation method as described above.
[0049] Compared with the prior art, the present invention has the following advantages:
[0050] (1) The present invention generates two satellite remote sensing images with a fixed imaging time interval through simulation. In view of the characteristic of parallax observation, the parallax image pair is densely matched line by line to obtain the parallax value, and the frequency and amplitude of the parallax are solved based on Fourier series modeling. Then, the frequency and amplitude of the vibration are obtained according to the differential relationship between the parallax value and the geometric displacement value caused by the vibration. By changing the matching window size and the integral series, the influence of the TDI CCD integral series and the matching window size in image matching on the vibration amplitude and detection frequency can be analyzed.
[0051] (2) The results of the experiment using the remote sensing image vibration change simulation method of the present invention show that the matching window size and the integration series have little effect on the vibration detection frequency. However, as the matching window increases or the vibration frequency increases, the detected vibration amplitude becomes smaller and smaller. At the same time, as the integration series increases, the detected amplitude becomes smaller. The conclusion of the present invention can provide a useful reference for vibration detection based on remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 A flowchart of a method for simulating vibration changes of remote sensing images provided by an embodiment of the present invention;
[0053] Figure 2 is a disparity curve diagram of a sub-pixel matching result when the integration level is 1 in an embodiment of the present invention;
[0054] Figure 3 is a dither curve diagram of a sub-pixel matching result when the integration order is 1 in an embodiment of the present invention;
[0055] Figure 4 is a disparity curve diagram of the sub-pixel matching result when the integration level is 16 in an embodiment of the present invention;
[0056] Figure 5 FIG. 4 is a dither curve diagram of the sub-pixel matching result when the integration level is 16 in an embodiment of the present invention. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0058] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.
[0059] Example 1
[0060] like Figure 1 As shown, this embodiment provides a method for simulating vibration changes of remote sensing images, comprising the following steps:
[0061] Simulation image generation step S1: input an orthophoto image as a reference image, set track parameters, camera parameters and flutter parameters, and generate a simulation image based on a digital simulation model of dynamic imaging;
[0062] Image matching step S2: performing image registration on the simulation images at adjacent moments;
[0063] The flutter calculation step S3 based on Fourier series: matching two simulated images, obtaining a disparity value, and modeling the disparity value using Fourier series;
[0064] Match two simulated images with a fixed imaging time interval to obtain a disparity value, model the disparity value using Fourier series, and obtain the frequency and amplitude of the disparity value;
[0065] According to the differential relationship between the parallax value and the geometric displacement value, the amplitude and frequency of the jitter are obtained to describe the jitter changes of the remote sensing image.
[0066] The Fourier series expansion formula for the parallax value and geometric displacement value is:
[0067]
[0068]
[0069]
[0070] in, α m =|sinc(N m Tf)|,α n =|sinc(N n Tf)|;
[0071] In the formula, a 0 ,a i ,b i (i=1,2,...k) is the coefficient of the Fourier series of the parallax value, k is the order of the Fourier series, f is the frequency of the parallax value, according to the properties of the same frequency combined vibration, it can be known that the flutter frequency is also f, j tdi (t) is the geometric displacement value caused by attitude flutter on the TDI image plane at time t, j tdi (t+Δt) is the geometric displacement value caused by attitude flutter in the TDI image plane at time t+Δt, t is the imaging time, Δt is the imaging time interval, is the parallax value of adjacent simulated images corresponding to the imaging time t, T is the single-level integration time, N m is the reference image integral series, N n is the integral series of the image to be matched, A i , B iare the coefficients of the Fourier series of geometric displacement values, i = 1, 2, ... k, and N is the integral series.
[0072] The calculation expression of the coefficient of the Fourier series of geometric displacement value is:
[0073]
[0074]
[0075] in, w = 2πf;
[0076] Parameter a i and b i according to The data are obtained through fitting.
[0077] After obtaining the coefficients of the flutter Fourier series, the flutter amplitude can be obtained.
[0078]
[0079] The present invention generates two satellite remote sensing images with a fixed imaging time interval by simulation, and obtains the disparity value by densely matching the disparity image pair line by line in view of the characteristic of parallax observation. Then, the frequency and amplitude of the vibration are obtained by expanding the Fourier series analysis of the disparity value, thereby being able to analyze the influence of the TDI CCD integral series and the matching window size in image matching on the vibration amplitude and detection frequency.
[0080] In the image matching step, under matching windows of different sizes, the image registration is performed by normalized cross-correlation and quadratic parabola fitting methods to obtain disparity difference values and build a model. According to the relationship between the disparity value and the geometric displacement value caused by the vibration, the vibration frequency amplitude is derived and compared with the actual vibration frequency to obtain the influence of matching windows of different sizes on the vibration detection frequency.
[0081] As a preferred implementation, the digital simulation model of dynamic imaging adopts a TDI integral image imaging model, which is used to simulate the image synthesized by the integration of a multi-level CCD sensor;
[0082] The method for simulating the vibration change of remote sensing images also includes:
[0083] According to the imaging principle of TDI integral imaging, the platform flutter of each imaging stage changes with time during the multi-stage integration process. The expression of the flutter curve after TDI integration is constructed to obtain the relationship between the TDI integral and the single-stage integral flutter amplitude.
[0084] The expression of the flutter curve of the synthesized image is:
[0085]
[0086] Where, d TDI is the flutter curve of the synthesized image, N is the integration level, L is the line number, A is the single-level integration flutter amplitude, f is the frequency, t is the imaging time, is the phase, c is the column number, and T is the single-stage integration time;
[0087] The expression of the influence of TDI integral on the overall flutter amplitude of remote sensing images is:
[0088] A TDI =A|sinc(Ntf)|
[0089] In the formula, A TDI It is the jitter amplitude of the image synthesized by TDI integration.
[0090] The expression of TDI integral imaging model is:
[0091]
[0092]
[0093] Where I(r,c) is the grayscale value of row r and column c, r is the row number, c is the column number, N is the integration level, a is the sampling interval of CCD in the x and y directions, T is the single-level integration time, v is the push-scan speed during linear array push-scan imaging, t is the imaging time, and y is the integration level of the CCD. jitter (t) is the vibration in the vertical rail direction at time t, A is the single-stage integrated vibration amplitude, f is the frequency, For phase.
[0094] As a preferred implementation, in the image matching step, the image registration specifically includes the following steps:
[0095] Obtain two simulation images at adjacent moments, recorded as image M and image S;
[0096] Use Wallis filtering to enhance images and improve image contrast;
[0097] The images M and S are translated to achieve initial registration and eliminate fixed offset;
[0098] For each point to be matched on image M, the pixel coordinates with the largest normalized mutual correlation coefficient on image S and the normalized mutual correlation values of the pixels in the matching window centered on the pixel coordinates are calculated. The normalized mutual correlation values in the matching window are fitted using a quadratic parabola, and the pixel coordinates at the vertex of the quadratic parabola are the sub-pixel matching results.
[0099] Further as a preferred implementation, the image registration process also includes a step of removing mismatched points:
[0100] Calculate the normalized mutual correlation coefficient between matching windows and remove the areas where the correlation coefficient is less than a preset correlation threshold;
[0101] The disparity of the matching point pairs of the image M and the image S is counted, and the matching points whose offset is greater than the preset disparity threshold are eliminated.
[0102] This embodiment also provides a remote sensing image vibration change simulation system, including a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the remote sensing image vibration change simulation method as described above.
[0103] The specific implementation process of the above remote sensing image vibration change simulation method includes the following steps:
[0104] S1: Generate single-level integral and TDI integral simulated images affected by jitter: Input any orthophoto image as a reference image, then set orbit and camera parameters (satellite orbit altitude, camera focal length, pixel size, line frequency, integral order, array physical spacing, jitter frequency amplitude) and other information, and generate simulated images of different integral orders affected by different jitter characteristics according to the digital simulation model of dynamic imaging;
[0105] S2: Flutter detection based on sub-pixel dense matching: For two multi-temporal simulated images, the dense matching method of normalized cross-correlation and quadratic parabola fitting is used to obtain the precise parallax difference value between the simulated images;
[0106] S3: Flutter estimation based on Fourier series: For the disparity values obtained by matching, the Fourier series is used to accurately model the flutter amplitude, frequency and other characteristics.
[0107] The following is a detailed description of each step:
[0108] 1. Generation of simulated images affected by vibration in step S1
[0109] Assume that the simulated image before CCD sampling is f(x,y), the sampling interval of CCD in the x direction and y direction is equal and equal to a, when the linear array push-scan imaging, assume that the push-scan speed is v, N is the integral series, and the vibration in the vertical track direction is y jitter (t), under the effect of image shift and sampling, the TDI integral image imaging model is as follows:
[0110]
[0111]
[0112] When N=1, This is the dynamic imaging model of an ordinary linear array camera.
[0113] 2. Obtaining the flutter parallax based on sub-pixel matching in step S2
[0114] According to step S1, two simulated images with fixed imaging intervals are generated, denoted as M and S respectively, and Wallis filtering is used to enhance the images and improve the image contrast. The images are translated according to the known parallax to achieve initial registration and eliminate the fixed offset between the two images. Then, dense matching is performed at fixed intervals in the row and column directions of the images. For each point to be matched on image M, the pixel coordinates with the maximum normalized cross-correlation coefficient (NCC) on image S and the NCC value in its 3×3 neighborhood are calculated, and then the NCC value in the window is fitted using a quadratic parabola. The position of the quadratic parabola vertex is the sub-pixel matching result. At the same time, possible mismatched points are eliminated, including the following two steps: (1) Calculate the normalized cross-correlation coefficient between matching windows and remove low-texture or low-correlation areas with correlation coefficients less than a threshold of 0.7; (2) Count the parallax of matching point pairs in the same image row and eliminate matching points with large offsets using three times the mean square error.
[0115] 3. Flutter analysis based on Fourier series
[0116] Based on the relative parallax of the same-name points obtained in step S2, a line-by-line analysis is carried out to obtain the change of relative parallax with imaging time. Finally, the platform vibration is analyzed by Fourier series to obtain the frequency spectrum characteristics of the vibration, such as frequency and amplitude.
[0117] First, consider the flutter model of TDI multi-level integration. Assume that the single-level integration time is T, the single-level integration flutter amplitude is A, the initial integration time of L is t, and the initial phase is The number of integration levels is N. Because the platform vibration of each imaging level changes with time during the TDICCD multi-level integration process, the vibration curve of the synthesized image is as follows:
[0118]
[0119] Therefore, the relationship between the TDI integral and the single-stage integral amplitude is:
[0120] A TDI =A|sinc(Ntf)| (4)
[0121] Therefore, the parallax-based chatter detection method is as shown in formula (6):
[0122]
[0123] Where t is the imaging time, Δt is the imaging time interval, j tdi (t) is the geometric displacement value caused by attitude flutter on the TDI image plane, It is the parallax value of each row of image corresponding to the imaging time t.
[0124]
[0125] where α m =|sinc(N m Tf)|,α n =|sinc(N n Tf)|;
[0126] and j tdi (t) can be expressed in the form of Fourier series as shown below.
[0127]
[0128]
[0129]
[0130] make
[0131]
[0132] The flutter coefficients A and B can be obtained as shown in the following formula.
[0133]
[0134] 4. Experiment and analysis
[0135] In the simulation experiment, it is assumed that the satellite orbit altitude is 537km, the ground resolution is 2m, the pixel size is 7.44μm, the row exposure time of the sensor is 0.000285s, and the physical interval between adjacent line arrays is 100 rows. The reference image used in the present invention is a 2-meter resolution digital orthophoto (DOM) with an image size of 1280*1280 pixels. Figure 1 shown.
[0136] In order to test the influence of different matching window sizes and integration levels on the frequency and amplitude of chatter detection, the window size in the column direction is fixed to 128 pixels, and the window sizes in the row direction are 1, 2, 4, 8, 16, and 32 respectively. At the same time, in order to simulate single-integral images and TDI multi-integral images, the integration levels are set to 1 and 16 respectively. The results are listed in Tables 1 and 2. The results show that in this experiment, for chatter of 305Hz or 485Hz and below, each matching window can stably detect the chatter frequency; for higher frequency chatter, some matching windows cannot continuously and stably detect the chatter frequency, among which the 128×4 matching window has the best robustness. Although the detection frequencies of matching windows of different sizes are slightly different, there is no inverse relationship between the matching window size and the maximum detectable frequency as generally believed.
[0137] Table 1 Chatter detection results under different matching window sizes (integral level is 1)
[0138]
[0139]
[0140] Table 2 Flutter detection results under different matching window sizes (integration level = 16)
[0141]
[0142] In order to further analyze the influence of the matching window on the vibration detection amplitude, we analyzed the vibration amplitude detected in the above experiment. Figure 2 and Figure 3 It can be found that when the frequency is small, the amplitudes of the disparity curves and the chatter curves obtained under different matching windows are slightly different, indicating that the matching window has little effect on the low-frequency chatter detection amplitude; however, as the frequency increases, the impact of the matching window on the amplitude estimation becomes more obvious, and the larger the matching window, the smaller the amplitude of the detection estimation.
[0143] Combining Table 1 and Table 2, it can be found that for TDI integral simulation images, as the flutter frequency increases, the flutter detection results become less ideal and it is difficult to detect flutter. Figure 4 and Figure 5 By comparison, it can be found that when the frequency is less than 100 Hz, the image vibration detection amplitudes of the same frequency and the same matching window and different integration levels are slightly different. However, when the frequency gradually increases, the parallax value amplitude of the TDI multi-level integration image is significantly smaller than the parallax value of the single-level integration image, indicating that the influence of higher-frequency vibration on images with longer integration time is easily obscured, while the influence of low-frequency vibration decays slowly with the integration time.
[0144] 5. Conclusion
[0145] This embodiment verifies the influence of the matching window size and the integration series on the flutter detection through simulation experiments. The experimental results show that the matching window size and the integration series have little influence on the flutter detection frequency. However, as the matching window increases or the flutter frequency increases, the detected flutter amplitude becomes smaller and smaller. At the same time, as the integration series increases, the detected amplitude becomes smaller. The conclusion of this embodiment can provide a useful reference for flutter detection based on remote sensing images.
[0146] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A method for simulating vibration changes of remote sensing images, characterized in that: The following steps are involved: Simulation image generation steps: input an orthophoto image as a reference image, set track parameters, camera parameters and flutter parameters, and generate two simulation images with a fixed imaging time interval based on a dynamic imaging digital simulation model; Image matching step: performing image registration on the simulation image; The vibration calculation steps based on Fourier series are as follows: matching two simulated images with a fixed imaging time interval to obtain a disparity value, modeling the disparity value using Fourier series, and obtaining the frequency and amplitude of the disparity value; According to the differential relationship between the parallax value and the geometric displacement value, the amplitude and frequency of the vibration are obtained to describe the vibration changes of the remote sensing image. The Fourier series expansion formula of the parallax value and the geometric displacement value caused by the vibration is: Among them, a m =|sinc(N m Tf)|,a n =|sinc(N n Tf)|; In the formula, a0, a i ,b i (i=1,2,...k) are the coefficients of the Fourier series expansion formula of the parallax value, k is the order of the Fourier series, f is the frequency of the parallax value, and according to the properties of the same frequency combined vibration, it can be known that the flutter frequency is also f, j tdi (t) is the geometric displacement value caused by attitude flutter on the TDI image plane at time t, j tdi (t+Δt) is the geometric displacement value caused by attitude flutter in the TDI image plane at time t+Δt, t is the imaging time, Δt is the imaging time interval, is the parallax value corresponding to the imaging time t, T is the single-level integration time, N m is the reference image integral series, N n is the integral series of the image to be matched, A i , B i are the coefficients of the Fourier series of geometric displacement values, i = 1, 2, ... k, and N is the integral series; The calculation expression of the Fourier series coefficient of the geometric displacement value caused by the flutter is: in, Parameter a i and b i Obtained by fitting; After obtaining the coefficients of the flutter Fourier series, the flutter amplitude Amp can be obtained; In the image matching step, under matching windows of different sizes, a method of normalized cross-correlation and quadratic parabola fitting is used to perform dense matching, obtain relative disparity of the same-name points and build a model, derive the amplitude of the vibration frequency according to the relationship between the disparity value and the geometric displacement value caused by the vibration, and compare it with the actual vibration frequency to obtain the influence of matching windows of different sizes on the vibration detection frequency; The digital simulation model of dynamic imaging adopts a TDI integral imaging model, which is used to simulate the multi-stage CCD sensor integral imaging process; The remote sensing image vibration change simulation method further includes: according to the TDI integral image imaging principle, the platform vibration of each imaging stage changes with time during the multi-stage integration process, so an expression of the vibration curve after TDI integration is constructed to obtain the relationship between the TDI integral and the single-stage integral vibration amplitude; The expression of the flutter curve of the synthesized image is: Where, d TDI is the chatter curve of the image after TDI integration, N is the integration level, L is the line number, A is the single-level integration chatter amplitude, f is the frequency, t is the imaging time, is the phase, T is the single-stage integration time; The relationship between the TDI integrated vibration amplitude and the single-stage integrated vibration amplitude is expressed as follows: A TDI =A|sinc(Ntf)| In the formula, A TDI is the jitter amplitude of the image synthesized by TDI integration; The expression of the TDI integral imaging model is: Where I(r,c) is the grayscale value of row r and column c, r is the row number, c is the column number, N is the integration level, a is the sampling interval of CCD in the x and y directions, T is the single-level integration time, v is the push-scan speed during linear array push-scan imaging, t is the imaging time, and y is the integration level of the CCD. jitter (t) is the vibration in the vertical rail direction at time t, A is the single-stage integrated vibration amplitude, f is the frequency, For phase.
2. The method for simulating vibration changes of remote sensing images according to claim 1, characterized in that: In the image matching step, the image registration specifically includes the following steps: Obtain two simulated images at adjacent moments, recorded as image M and image S; Use Wallis filtering to enhance images and improve image contrast; The image M and the image S are translated to achieve initial registration and eliminate fixed offset; For each point to be matched on image M, the pixel coordinates with the largest normalized mutual correlation coefficient on image S and the normalized mutual correlation values of the pixels in the matching window centered on the pixel coordinates are calculated. The normalized mutual correlation values in the matching window are fitted using a quadratic parabola, and the pixel coordinates at the vertex of the quadratic parabola are the sub-pixel matching results.
3. The method for simulating vibration changes of remote sensing images according to claim 2, characterized in that: The image registration process also includes the step of removing mismatched points: Calculate the normalized mutual correlation coefficient between matching windows and remove the areas where the correlation coefficient is less than a preset correlation threshold; The disparity of the matching point pairs of the image M and the image S is counted, and the matching points whose offset is greater than the preset disparity threshold are eliminated.
4. A remote sensing image vibration change simulation system, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of any one of the methods according to claims 1 to 3.
Citation Information
Patent Citations
Satellite attitude fluttering detection method based on imaging sensor parallax error
CN103791899A
Satellite platform chattering detection method and system based on multispectral image
CN103983343A
Method for detecting posture flutter of high resolution three linear array satellites along rail direction
CN104864853A
Ground simulation test system for influence of satellite flutter on camera imaging
CN111222544A
Method for reducing image fuzzy degree of TDI-CCD camera
US20160165155A1
