A strip noise detection method and device, and computer storage medium

Abstract

The present application discloses a stripe noise detection method and device, and a computer storage medium, and relates to the field of remote sensing image processing technology. The stripe noise detection method is based on the Fengyun-4B fast imager, and includes: superimposing satellite remote sensing images within a certain period of time, calculating the mean pixel value of each column of pixels in the superimposed remote sensing image, and obtaining the original mean data; using a window to smooth the original mean data to obtain the fitted mean data; calculating the difference between the fitted mean and the original mean of each column of pixels, and differentiating the difference; setting a threshold, and comparing the threshold with the differential value column by column, if the differential value of the column of pixels is greater than the threshold, then there is stripe noise in the column. Through simple steps, the stripe noise in the satellite image can be detected in real time and in batches, and the specific location of the stripe noise in the satellite image can be detected quickly and accurately, making the detection of the stripe noise more efficient, and greatly reducing the workload of calculation and detection.

Description

Technical Field

[0001] The present application relates to the technical field of remote sensing image processing, and more specifically, to a stripe noise detection method and device, and a computer storage medium. Background Art

[0002] The rapid imager is one of the main payloads on board the Fengyun-4B satellite, which can achieve rapid imaging of an area of ​​2000km×2000km with a frequency of 1 minute.

[0003] In the process of remote sensing imaging, various types of noise will inevitably be generated in the collected remote sensing image data. Among them, the most typical one is stripe noise, which appears as linear strips with fixed distribution in the remote sensing image. It not only affects the image observation effect of the satellite remote sensing image collected by the Fengyun-4B rapid imager, but also causes subsequent processing problems.

[0004] In recent years, there have been many studies on the removal of abnormal stripes on satellite remote sensing images, mainly including digital filtering, image grayscale information statistics, variational methods, etc. However, the current research on stripe noise often requires the pre-acquisition of stripe location information. When the distribution of stripe noise does not have any regularity, the use of manual interpretation to mark the location information will consume a lot of human resources. This limitation is particularly evident when dealing with sparsely distributed complex stripe noise.

[0005] There are still some defects in the existing patents related to the detection of stripe noise in images. For example, the patent application "Self-detection and Removal Method of Spectral Domain Noise of Aerial Hyperspectral Remote Sensing Images" (application number 200510027528.9) discloses that the reflectivity of the image can be used to detect the spectral domain noise of the hyperspectral image, but the calculation process is relatively complicated; for example, the patent application "An Improved Algorithm for Removing Horizontal Stripe Noise in Line Array Images Based on Line Tracking" (application number 201610649429.2) discloses a method for determining the location information of horizontal stripe noise through the connection between feature points, but in the process of determining the noise location, it is necessary to repeat multiple comparisons and judgments, which is cumbersome as a whole; for another example, in the patent application "A Radiation Quality Detection Method and System for Multi-Source Satellite Remote Sensing Image Products" (application number: 202210586657.5), although it is mentioned that stripe detection can be performed on multi-source satellite remote sensing image products, it is necessary to extract the frame to be detected in the image during stripe detection, instead of detecting the entire image. Summary of the invention

[0006] In view of this, the present application provides a stripe noise detection method and device, and a computer storage medium. Through simple steps, the stripe noise in satellite images can be detected in real time and in batches, and the specific location of the stripe noise in the satellite images can be detected quickly and accurately, making the detection of stripe noise more efficient, while greatly reducing the workload of calculation and detection.

[0007] In a first aspect, the present application provides a stripe noise detection method. The stripe noise detection method performs stripe noise detection based on the FY-4B fast imager. The fast imager is used to obtain satellite remote sensing images in real time. The stripe noise detection method includes:

[0008] Divide the satellite remote sensing images by time, and superimpose the satellite remote sensing images within a certain period of time to obtain superimposed remote sensing images;

[0009] Calculate the mean pixel value of each column of the stacked remote sensing image to obtain the original mean data;

[0010] The original mean data is smoothed using a window with a preset size to obtain fitted mean data, where the fitted mean data is a set of fitted means of pixels in each column of the stacked remote sensing image;

[0011] Calculate the difference between the fitted mean and the original mean of each column of pixels in the stacked remote sensing image, and perform differential processing on the difference;

[0012] Set a threshold and compare the threshold with the differential result column by column. If the differential result of the pixels in this column is greater than the threshold, there is stripe noise in this column; if the differential result of the pixels in this column is less than or equal to the threshold, there is no stripe noise in this column.

[0013] Among them, the mean pixel value of each column of the superimposed remote sensing image is calculated to obtain the original mean data including:

[0014] The pixel values ​​of each column of the superimposed remote sensing image are summed and then averaged to obtain the original mean value of each column of pixels;

[0015] Arrange the original mean values ​​of each column of pixels into rows according to the position order of the corresponding column of pixels to obtain the original mean value data;

[0016] The window has a center point, and the original mean data is smoothed using a window with a preset size to obtain fitted mean data, which is a set of fitted mean values ​​of each column of pixels in the superimposed remote sensing image, including:

[0017] The center point of the window moves column by column along the row direction on the original mean data. During the movement, the original mean of the overlapping part of the window and the original mean data is fitted, and the fitted value is used as the fitting mean of the pixels in the column where the center point is located in the superimposed remote sensing image. The set of fitting means of pixels in each column of the superimposed remote sensing image is the fitting mean data.

[0018] Optionally, where:

[0019] The default size of the window is 11.

[0020] Optionally, where:

[0021] The remote sensing images acquired by the rapid imager include visible light range images, near-infrared range images and infrared range images. When the remote sensing images are visible light range images and near-infrared range images, the threshold is set to 0.03; when the remote sensing images are infrared range images, the threshold is set to 0.2.

[0022] Optionally, where:

[0023] Before dividing the satellite remote sensing images by time and superimposing the satellite remote sensing images within one hour to obtain the superimposed remote sensing images, the strip noise detection method further includes:

[0024] The satellite remote sensing images acquired by the rapid imager are preprocessed, which includes quality inspection, geolocation, radiometric calibration, and conversion of pixel values ​​in visible light range images and infrared range images into reflectance according to the following formula:

[0025] ρ λ =SCALE×DN+OFFSET

[0026] Where: λ is the preprocessed satellite remote sensing image, SCALE provides the slope for converting pixel values ​​into reflectance, DN is the pixel value of the satellite remote sensing image, and OFFSET provides the intercept for converting pixel values ​​into reflectance.

[0027] In a second aspect, the present application also provides a stripe noise detection device, comprising: a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instruction to implement the stripe noise detection method described in the first aspect.

[0028] In a third aspect, the present application further provides a computer storage medium, in which instructions are stored. When the instructions are executed, the strip noise detection method described in the first aspect is implemented.

[0029] Compared with the prior art, the stripe noise detection method and device, and computer storage medium provided by the present application achieve at least the following beneficial effects:

[0030] The stripe noise detection method provided in the present application performs subsequent processing and detection on the basis of the stacked remote sensing images obtained by stacking, thereby ensuring the accuracy of the detection results. Afterwards, the pixel value mean of the stacked remote sensing images after stacking is calculated along the column direction, and then the mean data obtained is mean fitted using a window with a preset size, and the fitting mean and the original mean of each column of pixels are subjected to difference processing, and the difference result is differentiated to reduce the irregular fluctuation of the data and enhance the display of abnormal data; finally, the abnormal stripes are screened by setting a threshold value, thereby realizing the rapid detection of the specific location of the stripe noise in the Fengyun-4B satellite rapid imager. It can be seen that the present application combines the scanning imaging characteristics of the FY-4B fast imager with the characteristics of strips in satellite remote sensing images, establishes a reliable stripe noise detection method, and greatly improves the real-time on-orbit detection and diagnosis capabilities of strips in FY-4B remote sensing images; at the same time, it can realize real-time batch detection of abnormal stripes in satellite remote sensing images, greatly reducing the workload of calculation and detection, and further improving detection efficiency; in addition, the stripe noise detection method provided by the present application can also accurately identify and mark a single stripe, and the differential result can also reflect the intensity of the stripe noise, which provides assistance for abnormality detection and fault judgment.

[0031] Of course, any product implementing this application does not necessarily need to achieve all of the technical effects described above at the same time.

[0032] Other features and advantages of the present application will become apparent from the following detailed description of exemplary embodiments of the present application with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the application and, together with the description, serve to explain the principles of the application.

[0034] Figure 1 FIG. 1 is a flow chart of a stripe noise detection method provided in an embodiment of the present application;

[0035] Figure 2 Shown is a stripe-free satellite remote sensing image of the visible light channel 1 of the Fengyun-4B rapid imager provided in an embodiment of the present application;

[0036] Figure 3 Shown is a schematic diagram of noise components established according to an ideal strip noise model provided in an embodiment of the present application;

[0037] Figure 4 Shown is a satellite remote sensing image of the visible light channel 1 of the Fengyun-4B fast imager provided in an embodiment of the present application, which includes noise components established according to an ideal strip noise model;

[0038] Figure 5 Shown is the pair Figure 4 The schematic diagram of the original mean value of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown;

[0039] Figure 6 Shown is the pair Figure 4 The schematic diagram of the fitted mean value of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown;

[0040] Figure 7 Shown is the pair Figure 4 The schematic diagram of the difference between the fitted mean and the original mean of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown;

[0041] Figure 8 Shown is the pair Figure 4 The schematic diagram of the difference result of the difference of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown;

[0042] Fig. 9 Shown is the pair Figure 4 The schematic diagram of the threshold screening result obtained after stripe noise detection of the satellite remote sensing image shown;

[0043] Fig.10 Shown is a step-by-step diagram of stripe noise detection for satellite remote sensing images acquired by the Fengyun-4B rapid imager within the 7th hour of visible light channel 1 on September 14, 2021 in an embodiment of the present application;

[0044] Fig.11 (a) and Fig.11 (b) shows two satellite remote sensing images randomly selected from all satellite remote sensing images of visible light channel 1 within the 7th hour acquired by the Fengyun-4B rapid imager on September 14, 2021 in an embodiment of the present application;

[0045] Fig.12 Shown is a schematic diagram of the detection results obtained by performing stripe noise detection on the satellite remote sensing image within the 7th hour of infrared channel 7 acquired by the Fengyun-4B rapid imager on September 14, 2021 in an embodiment of the present application;

[0046] Fig.13 (a) and Fig.13 (b) shows two satellite remote sensing images randomly selected from all satellite remote sensing images within the 7th hour of infrared channel 7 acquired by the Fengyun-4B rapid imager on September 14, 2021 in an embodiment of the present application. DETAILED DESCRIPTION

[0047] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present application.

[0048] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.

[0049] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0050] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0051] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0052] On June 3, 2021, Fengyun-4B was successfully launched at the Xichang Satellite Launch Center. Fengyun-4B is the first operational meteorological satellite of the Fengyun-4 series, the second generation of geostationary meteorological satellites in my country. It will form my country's geostationary meteorological satellite operational observation network together with Fengyun-4A, Fengyun-2H, Fengyun-2G and Fengyun-2F, and realize the commercialization of Fengyun-4 satellites, gradually replacing Fengyun-2 as the main force of the World Meteorological Organization's global meteorological satellite observation system.

[0053] Fengyun-4B carries four main payloads: Advanced Geostationary Radiation Imager (AGRI), Geostationary Interferometric Infrared Sounder (GIIRS), Geo High-speed Imager (GHI) and Space Environment Monitoring Instrument Package (SEP).

[0054] The Global Hierarchical Imager (GHI) is the first such payload to be carried in geostationary orbit. It can achieve rapid imaging of an area of ​​2000km×2000km at a frequency of 1 minute, and its spatial resolution reaches 250m, which is the highest resolution of geostationary meteorological satellites in the world. It further enhances the continuous, flexible and high-resolution observation capability of typhoons, heavy rains and mesoscale disastrous weather.

[0055] As FY-4B flies forward along its orbit, the imager it carries will also move with it. According to the working principle of the FY-4B GHI, after collecting electromagnetic radiation, the charge-coupled device (CCD) linear array detector will scan and image line by line. When the offset and relative gain values ​​of the detector pixels are not equal, stripe noise will be generated. Therefore, the distribution of the stripe noise must be consistent with the scanning direction, in the form of rows or columns along the image.

[0056] At present, there are two main types of remote sensing imaging systems: push-broom imaging systems and cross-track imaging devices with stripes. During the imaging process, the imaging sensor will inevitably generate various types of noise on the collected remote sensing image data, among which stripe noise is more typical. The main reason for its generation is that the aging of the satellite CCD detector leads to uneven response of the detection unit, and the vignetting of the imaging optical system, which makes the image appear as linear noise with fixed direction. Affected by this, the GHI image of Fengyun-4B is not only affected by the image observation effect, but also causes subsequent processing problems, such as classification, target detection, quantitative application, etc. Since there is no correlation between the response of each pixel in the image to the noise, this stripe noise will show strong randomness and unmonitorability on the remote sensing image, and it is difficult to remove it using general calibration methods.

[0057] In recent years, there have been many studies on algorithms for removing abnormal stripes from satellite remote sensing images, mainly digital filtering, image grayscale information statistics, variational methods, etc., but there are fewer studies on abnormal stripe detection on satellite remote sensing images. At present, the study of stripe noise often requires the pre-acquisition of the position information of the stripe. In the prior art, the acquisition of stripe noise position information is often obtained by manual translation, but when the distribution of stripe noise does not have any regularity, the use of manual interpretation to mark the position information will consume a lot of human resources. This limitation is particularly evident when dealing with sparsely distributed complex stripe noise.

[0058] In order to solve the above technical problems, the present application proposes a stripe noise detection method and device, and a computer storage medium. Through simple steps, the stripe noise in satellite images can be detected in real time and in batches, and the specific location of the stripe noise in the satellite images can be detected quickly and accurately, making the detection of stripe noise more efficient, while greatly reducing the workload of calculation and detection.

[0059] The following is a detailed description with reference to the accompanying drawings and specific embodiments.

[0060] Figure 1 The figure is a flow chart of the strip noise detection method provided in the embodiment of the present application.

[0061] like Figure 1 As shown, the embodiment of the present application provides a stripe noise detection method, which performs stripe noise detection based on the FY-4B fast imager, which is used to obtain satellite remote sensing images in real time. The stripe noise detection method includes:

[0062] S100, dividing the satellite remote sensing images according to time, and superimposing the satellite remote sensing images within a certain period of time to obtain a superimposed remote sensing image;

[0063] S200, calculating the mean pixel value of each column of pixels in the stacked remote sensing image to obtain original mean data; wherein calculating the mean pixel value of each column of pixels in the stacked remote sensing image to obtain original mean data includes:

[0064] The pixel values ​​of each column of the superimposed remote sensing image are summed and then averaged to obtain the original mean value of each column of pixels;

[0065] Arrange the original mean values ​​of each column of pixels into rows according to the position order of the corresponding column of pixels to obtain the original mean value data;

[0066] S300, using a window with a preset size to smooth the original mean data to obtain fitted mean data, where the fitted mean data is a set of fitted mean values ​​of each column of pixels in the stacked remote sensing image; wherein the window has a center point, and using a window with a preset size to smooth the original mean data to obtain fitted mean data, where the fitted mean data is a set of fitted mean values ​​of each column of pixels in the stacked remote sensing image, includes:

[0067] The center point of the window moves column by column along the row direction on the original mean data. During the movement, the original mean of the overlapped part of the window and the original mean data is fitted. The fitted value is used as the fitted mean of the pixels in the column where the center point is located in the stacked remote sensing image. The set of the fitted mean of each column of pixels in the stacked remote sensing image is the fitted mean data.

[0068] S400, calculating the difference between the fitted mean and the original mean of each column of pixels in the stacked remote sensing image, and performing differential processing on the difference;

[0069] S500, setting a threshold, and comparing the threshold with the differential result column by column. If the differential result of the pixels in the column is greater than the threshold, there is stripe noise in the column; if the differential result of the pixels in the column is less than or equal to the threshold, there is no stripe noise in the column.

[0070] Based on this, the stripe noise detection method provided in the embodiment of the present application first superimposes multiple satellite remote sensing images acquired by the rapid imager within a certain period of time, and performs subsequent processing and detection on the basis of the superimposed remote sensing images obtained by superposition, thereby ensuring the accuracy of the detection results. Afterwards, the pixel value mean of the superimposed remote sensing images after superposition is calculated along the column direction, and the calculated pixel value mean is arranged in rows column by column, and then the mean data obtained is mean fitted using a window with a preset size to obtain the fitted mean data after smoothing; the fitting mean and the original mean of each column of pixels are subjected to difference processing, and the difference result is subjected to difference processing to reduce the irregular fluctuation of the data and enhance the display of abnormal data; finally, the difference result is compared column by column through the set threshold to filter out abnormal stripes. Since the pixels at the location where the ground objects change in the satellite remote sensing image are generally mixed pixels, the pixel values ​​of adjacent pixels will only change gradually, but not suddenly, so the difference result can be used to determine whether the pixel values ​​of adjacent pixels have suddenly changed, and then determine whether there are stripes. If the differential result is greater than the threshold, it means that the pixel values ​​of adjacent pixels have changed suddenly, which further indicates that there is stripe noise in the column, thus realizing the rapid detection of the specific position of the stripe noise in the FY-4B fast imager. It can be seen that the embodiment of the present application combines the scanning imaging characteristics of the FY-4B fast imager with the characteristics of strips in satellite remote sensing images, and establishes a reliable strip noise detection method with simple steps and efficient methods, which greatly improves the real-time on-orbit detection and diagnosis capabilities of strips in FY-4B remote sensing images; at the same time, the strip noise detection method provided in the embodiment of the present application is based on the image data acquired by the FY-4B fast imager, and takes into account the spatial and temporal statistical characteristics of multiple remote sensing images in a continuous time, so as to realize real-time batch detection of abnormal strips in satellite remote sensing images, greatly reduce the workload of calculation and detection, and further improve the detection efficiency; in addition, the strip noise detection method provided in the embodiment of the present application can not only realize the accurate detection of the position of abnormal strips in satellite remote sensing images, but also can accurately identify and mark single strips, and the differential result can also reflect the intensity of strip noise, which provides help for abnormality detection and fault judgment.

[0071] It should be noted that the pixel value of a pixel in the embodiment of the present application can be understood as the grayscale value of the pixel.

[0072] In some examples, the embodiments of the present application take the abnormal stripes in the satellite remote sensing image obtained by the Fengyun-4B rapid imager as column additive noise. The distribution of the strip noise along the column direction in the satellite remote sensing image is used as an example for illustration, but is not specifically limited.

[0073] In some examples, when dividing satellite remote sensing images by time and superimposing satellite remote sensing images within a certain period of time to obtain superimposed remote sensing images, satellite remote sensing images within one hour can be superimposed. Considering that abnormal bands in satellite remote sensing images are mainly caused by imperfect calibration, which usually lasts for a period of time or even months on the detector, the strip noise detection method provided in the embodiment of the present application can divide the satellite remote sensing images collected by the fast imager along the time into hours, process the image data within one hour each time, superimpose and sum all the image data within the hour, and then take the average along the column direction, instead of processing each image each time, which greatly reduces the workload of calculating and monitoring the actual satellites in orbit while maintaining the accuracy of the detection results.

[0074] In some examples, when the original mean data is smoothed, a window with a preset size can be used to move column by column along the row direction on the data set composed of the original mean values ​​of each column of pixels. When moving, the original mean in the window is mean-fitted, and the fitting result is assigned to a new array as the fitting mean at the position of the window center point. After the smoothing process is completed, the fitting means of all column pixels are arranged in rows column by column, and finally the fitting mean data is obtained, which is convenient for detecting whether there is an abnormal band in the current column based on the difference between the fitting mean and the original mean. It can be seen that the strip noise detection method provided in the embodiment of the present application can realize the accurate detection of abnormal bands through simple steps, without the need to use complex calculations and cumbersome formulas, which greatly reduces the workload of calculation and detection, and further improves the detection efficiency.

[0075] For example, since the embodiment of the present application uses the stripe noise as column additive noise for illustration, the original mean data formed is a data set in which the original mean values ​​of each column of pixels are arranged in rows. Therefore, the window selected in the embodiment of the present application is a window arranged along the row direction; if the stripe noise is row additive noise, a window arranged along the column direction can be selected for mean fitting, and no specific examples are given here.

[0076] Exemplarily, the preset size of the window is 11. Specifically, the window is a window with 11 pixels arranged one by one along the row direction. At this time, the center point is located in the middle position of the window with 11 points, and there are 5 pixels on the left and 5 pixels on the right of the center point. During the smoothing process, the window starts to move along the row direction from the first original mean of the original mean data. Specifically, the center point of the window gradually moves along the row direction from the original mean corresponding to the first column of pixels. When the center point of the window coincides with the original mean position corresponding to a column of pixels, the original mean of the 11 pixels in the window that overlap with the original mean data is averaged, and the averaged value is assigned to the corresponding position of the column of pixels corresponding to the center point in a new array, which is the fitting mean of the column of pixels. When the preset size of the window is 11, a good smoothing effect can be achieved on the basis of retaining key information. If the preset size of the window is too large, key information will be missed; if the preset size of the window is too small, no smoothing effect will be achieved.

[0077] For example, during the smoothing process, if the positions of the pixels of the window do not completely coincide with the original mean data, that is, the positions of some of the pixels of the window do not have the original mean, the window needs to be mirrored. For example, when the center point of the window coincides with the leftmost original mean of the original mean data, the positions of the 5 pixels to the left of the center point do not have the original mean, and the original mean corresponding to the 5 pixels to the right of the center point needs to be mirrored to the left, and the left and right sides of the center point are completely symmetrical.

[0078] In some examples, calculating the difference between the fitted mean and the original mean of each column of pixels in the stacked remote sensing image refers to subtracting the original mean from the fitted mean of each column of pixels in the stacked image to perform the difference.

[0079] In some examples, the remote sensing images acquired by the rapid imager include visible light range images, near infrared range images, and infrared range images. When the remote sensing images are visible light range images and near infrared range images, the threshold is set to 0.03; when the remote sensing images are infrared range images, the threshold is set to 0.2.

[0080] Based on this, when setting a threshold for the differential result to screen abnormal stripes, the threshold can be set according to the experience in actual projects. The specific setting value of the threshold varies depending on the visible light channel, near infrared channel and infrared channel. In actual business and research, the threshold can be set to 0.03 to screen abnormal stripes in the visible light channel and near infrared channel, and the threshold can be set to 0.2 to screen abnormal stripes in the infrared channel. When the differential result is greater than the set threshold, it means that there is strip noise in the column, which can well detect abnormal stripes that affect the application of satellite remote sensing image data, further improving the accuracy of strip noise detection. Exemplarily, the threshold can also be adjusted to a smaller value to detect smaller abnormal stripes.

[0081] At present, most of the stripe noise detection and removal models are to find the ideal clean stripe-free image directly from the measured image. However, considering the good characteristics of stripe noise, if the characteristics of stripe noise can be modeled and constrained, the estimation of stripe components will be more accurate, and the obtained stripe detection results will be more ideal.

[0082] Therefore, in the on-orbit detection of stripe noise of the Fengyun-4B satellite, the embodiment of the present application combines the characteristics of stripe noise in satellite remote sensing images, pre-sets the position and intensity information of the stripe, establishes an ideal stripe noise model on the stripe-free satellite remote sensing image of visible light channel 1, and determines whether the detection result of the embodiment of the present application is consistent with the stripe noise designed in the model, thereby proving the effectiveness of the embodiment of the present application. The specific establishment process of the ideal stripe noise model is as follows.

[0083] First, the embodiment of the present application models the stripe effect in the FY-4B rapid imager image as additive noise, and the degradation process of the stripe-free satellite remote sensing image is expressed as follows:

[0084] X+S=Y

[0085] Among them, the image containing stripes can be expressed as the sum of the original image and the stripe component, where Y represents the degraded satellite remote sensing image, X represents the stripe-free satellite remote sensing image, and S represents the stripe component.

[0086] In order to more accurately design the stripe component to represent the stripe noise of the FY-4B rapid imager on-orbit, the key issue is to make full use of the directionality, low rank and sparse characteristics of the stripe component information and describe them with an appropriate model.

[0087] Therefore, when describing the sparse distribution characteristics of strip components at a global scale, the embodiment of the present application can directly design the following sparse measurement formula to quantify the sparsity of the strip components:

[0088]

[0089] For a solution vector, use x = [x1, x2, ... x n ] represents the sparse coefficient, and N is the column pixel dimension of the satellite remote sensing image. The larger the value of Sparseness(x), the greater the sparsity of x, indicating that the stripes are sparser.

[0090] After adding additive noise in the range of 20%-50% of the image mean to the strip-free satellite remote sensing image, observable stripes appear in the image, where the image mean refers to the result obtained by averaging the pixel values ​​of all pixels in the entire strip-free satellite remote sensing image, that is, the image intensity mean. Therefore, in order to make the subsequent observation of the strip detection results more intuitive, the embodiment of the present application establishes the following strip representation equation from left to right with the strip noise as the weak noise of 20% of the image intensity mean to the strong noise component of 50%:

[0091] y=B i X i

[0092] Among them, X i represents the strip vector of column i, B i represents the noise intensity of the band in the i-th column, B i Satisfy (B i -B i-1 =B i-1 -B i-2 ).

[0093] Figure 2 Shown is a stripe-free satellite remote sensing image of the visible light channel 1 of the Fengyun-4B rapid imager provided in an embodiment of the present application; Figure 3 Shown is a schematic diagram of noise components established according to an ideal strip noise model provided in an embodiment of the present application; Figure 4 Shown is a satellite remote sensing image of the visible light channel 1 of the Fengyun-4B fast imager provided in an embodiment of the present application, which includes noise components established according to an ideal strip noise model; Figure 5 Shown is the pair Figure 4 The schematic diagram of the original mean value of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown; Figure 6 Shown is the pair Figure 4 The schematic diagram of the fitted mean value of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown; Figure 7 Shown is the pair Figure 4 The schematic diagram of the difference between the fitted mean and the original mean of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown; Figure 8 Shown is the pair Figure 4 The schematic diagram of the difference result of the difference of each column of pixels obtained after stripe noise detection of the satellite remote sensing image shown; Fig. 9 Shown is the pair Figure 4 The schematic diagram of the threshold screening results obtained after stripe noise detection on the satellite remote sensing image shown.

[0094] So far, if Figure 3 As shown in the figure, the ideal strip noise model is established. Figure 2 The strip-free satellite remote sensing image of visible light channel 1 is shown in Figure 1. Figure 3 The noise components shown in the figure are combined according to the ideal strip noise model to obtain Figure 4 The strip-free satellite remote sensing image of the visible light channel 1 containing the noise component established according to the ideal strip noise model is shown in FIG. Figures 5 to 9 As shown, the stripe noise detection method provided in the embodiment of the present application is used to perform stripe detection on the satellite remote sensing image of the visible light channel 1 containing the ideal stripe noise model. Figures 5 to 9 In the figure, the horizontal axis represents the number of each column of pixels in the satellite remote sensing image, and the vertical axis represents the value obtained after processing the pixel values ​​of a certain column of pixels, which is represented by I.

[0095] Fig. 9 The vertical axis in represents the difference between the fitted mean and the original mean. The larger the absolute value of the difference, the larger the jump in the pixel value of the column, and the more obvious the stripe noise. Fig. 9 It can be seen from the threshold screening results shown in that the result obtained after the stripe noise detection method provided in the embodiment of the present application is used for stripe detection is Figure 3 The stripe noise components established according to the ideal stripe noise model shown in the figure are consistent, proving that the stripe noise detection method provided in the embodiment of the present application can detect the stripe noise more accurately, and the strength of the stripe noise can be reflected by the size of the difference value. Therefore, by processing the satellite remote sensing images within each hour through the stripe noise detection method provided in the embodiment of the present application, the abnormal columns in the satellite remote sensing images within each hour can be obtained, and then the stripe occurrence rate within the hour can be obtained. The stripe occurrence rate is an index related to the imager scanning column, which can reflect the frequency of stripe occurrence in the satellite remote sensing images acquired by the fast imager and the quality of the image.

[0096] In some examples, before dividing the satellite remote sensing images by time and superimposing the satellite remote sensing images within one hour to obtain the superimposed remote sensing images, the strip noise detection method further includes:

[0097] The satellite remote sensing images acquired by the rapid imager are preprocessed, which includes quality inspection, geolocation, radiometric calibration, and conversion of pixel values ​​in visible light range images and infrared range images into reflectance according to the following formula:

[0098] ρ λ=SCALE×DN+OFFSET

[0099] Where: λ is the preprocessed satellite remote sensing image, SCALE provides the slope for converting pixel values ​​into reflectance, DN is the pixel value of the satellite remote sensing image, and OFFSET provides the intercept for converting pixel values ​​into reflectance.

[0100] Based on this, the acquired FY-4B rapid imaging satellite remote sensing images can be used as level 0 source package data, and the level 0 source package data can be preprocessed, including quality inspection, geographic positioning, and radiation calibration processing. Since the pixel value itself does not have a unit meaning, the satellite remote sensing images of the visible light channel and the infrared channel can be converted from digital quantization values ​​(Digital Number, abbreviated as DN, also known as pixel value) to reflectivity and brightness temperature according to the above formula or lookup table to obtain L1 level remote sensing image data, and the L1 level remote sensing image data is used for subsequent stripe noise detection, which can convert the dimensionless pixel value into a reflectivity with actual physical meaning and eliminate the error caused by the sensor itself.

[0101] Fig.10 Shown is a step-by-step diagram of stripe noise detection for satellite remote sensing images acquired by the Fengyun-4B rapid imager within the 7th hour of visible light channel 1 on September 14, 2021 in an embodiment of the present application; Fig.11 (a) and Fig.11 (b) shows two satellite remote sensing images randomly selected from all satellite remote sensing images of visible light channel 1 within the 7th hour acquired by the Fengyun-4B rapid imager on September 14, 2021 in an embodiment of the present application; Fig.12 Shown is a schematic diagram of the detection results obtained by performing stripe noise detection on the satellite remote sensing image within the 7th hour of infrared channel 7 acquired by the Fengyun-4B rapid imager on September 14, 2021 in an embodiment of the present application; Fig.13 (a) and Fig.13 (b) shows two satellite remote sensing images randomly selected from all satellite remote sensing images of infrared channel 7 within the 7th hour acquired by the Fengyun-4B fast imager on September 14, 2021 in the embodiment of the present application. Fig.10 The detection step-by-step diagram shown refers to an image composed of the original mean diagram of each column of pixels, the fitted mean diagram of each column of pixels, the difference diagram of the fitted mean and the original mean of each column of pixels, and the difference result diagram of the difference of each column of pixels, in the process of stripe noise detection of the satellite remote sensing image acquired by the Fengyun-4B rapid imager in the 7th hour of visible light channel 1 on September 14, 2021; Fig.10 and Fig.12The horizontal axis represents the number of each column of pixels in the satellite remote sensing image, and the vertical axis represents the value obtained after processing the pixel values ​​of a certain column of pixels.

[0102] In order to further verify the effectiveness and accuracy of the stripe noise detection method provided in the embodiment of the present application, Figure 10 to Figure 13 As shown, the satellite remote sensing image acquired by the Fengyun-4B rapid imager on September 14, 2021 was preprocessed to obtain L1-level remote sensing image data. Since the preprocessing process has undergone radiometric calibration, the obtained L1-level remote sensing image data is specifically L1B-level remote sensing image data, and the L1B-level remote sensing image data of the visible light channel 1 and the infrared channel are subjected to abnormal stripe detection according to the stripe noise detection method provided in the embodiment of the present application. Fig.10 and Fig.11 It can be seen that no stripe noise was detected in the satellite remote sensing image of visible light channel 1 within the 7th hour acquired by the Fengyun-4B rapid imager on September 14, 2021. At the same time, stripe noise detection was performed on visible light channels 1 to 6 acquired by the Fengyun-4B rapid imager on September 14, 2021, respectively, showing that the stripe noise occurrence rate of each visible light channel on that day was 0, which is consistent with the actual L1B-level remote sensing image data of the visible light channel; reference Fig.12 and Fig.13 It can be seen that according to the actual experience of the infrared channel, the threshold is set to 0.2 and the stripe noise detection method provided in the embodiment of the present application is used to perform stripe detection on infrared channel 7. It is found that the infrared channel 7 acquired by the Fengyun-4B rapid imager on September 14, 2021 detected abnormal stripes in each hour, among which the number of stripes found in the 7th hour was the largest, which is consistent with the stripe noise situation in the actual satellite remote sensing images.

[0103] Based on the same inventive concept, the present application also provides a stripe noise detection device, including: a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instruction to implement the stripe noise detection method described in the above embodiment.

[0104] Compared with the prior art, the beneficial effects of the stripe detection device are the same as the beneficial effects of the stripe noise detection method described in the above embodiment, which will not be described in detail here.

[0105] Based on the same inventive concept, the present application also provides a computer storage medium, in which instructions are stored. When the instructions are executed, the stripe noise detection method described in the above embodiment is implemented.

[0106] Compared with the prior art, the beneficial effects of the computer storage medium are the same as the beneficial effects of the strip noise detection method described in the above embodiment, which will not be described in detail here.

[0107] In summary, the stripe noise detection method and device, and computer storage medium provided by the present application achieve at least the following beneficial effects:

[0108] The stripe noise detection method provided in the present application performs subsequent processing and detection on the basis of the stacked remote sensing images obtained by stacking, thereby ensuring the accuracy of the detection results. Afterwards, the pixel value mean of the stacked remote sensing images after stacking is calculated along the column direction, and then the mean data obtained is mean fitted using a window with a preset size, and the fitting mean and the original mean of each column of pixels are subjected to difference processing, and the difference result is differentiated to reduce the irregular fluctuation of the data and enhance the display of abnormal data; finally, the abnormal stripes are screened by setting a threshold value, thereby realizing the rapid detection of the specific location of the stripe noise in the Fengyun-4B satellite rapid imager. It can be seen that the present application combines the scanning imaging characteristics of the FY-4B fast imager with the characteristics of strips in satellite remote sensing images, establishes a reliable stripe noise detection method, and greatly improves the real-time on-orbit detection and diagnosis capabilities of strips in FY-4B remote sensing images; at the same time, it can realize real-time batch detection of abnormal stripes in satellite remote sensing images, greatly reducing the workload of calculation and detection, and further improving detection efficiency; in addition, the stripe noise detection method provided by the present application can also accurately identify and mark a single stripe, and the differential result can also reflect the intensity of the stripe noise, which provides assistance for abnormality detection and fault judgment.

[0109] Although some specific embodiments of the present application have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are only for illustration, not for limiting the scope of the present application. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.

Claims

1. A strip noise detection method, characterized in that: The stripe noise detection method is based on the fast imager of the Fengyun-4B satellite to detect stripe noise. The fast imager is used to obtain satellite remote sensing images in real time. The stripe noise detection method includes: Dividing the satellite remote sensing images by time, and superimposing the satellite remote sensing images within a certain period of time to obtain superimposed remote sensing images; Calculating the mean pixel value of each column of pixels in the stacked remote sensing image to obtain original mean value data; Using a window with a preset size to smooth the original mean data, to obtain fitted mean data, wherein the fitted mean data is a set of fitted means of pixels in each column of the stacked remote sensing image; Calculating the difference between the fitted mean and the original mean of each column of pixels in the stacked remote sensing image, and performing differential processing on the difference; A threshold is set, and the threshold is compared with the differential result column by column. If the differential result of the pixels in the column is greater than the threshold, then stripe noise exists in the column; if the differential result of the pixels in the column is less than or equal to the threshold, then stripe noise does not exist in the column; The step of calculating the mean pixel value of each column of pixels of the stacked remote sensing image to obtain the original mean value data includes: The pixel values ​​of each column of pixels of the stacked remote sensing image are summed and then averaged to obtain the original mean value of each column of pixels; Arrange the original mean values ​​of each column of pixels into rows according to the position order of the corresponding column of pixels to obtain the original mean value data; The window has a center point, and the window with a preset size is used to smooth the original mean data to obtain fitting mean data, and the fitting mean data is a set of fitting means of each column of pixels in the superimposed remote sensing image, including: The center point of the window moves column by column along the row direction on the original mean data. During the movement, the original mean of the overlapping part of the window and the original mean data is fitted, and the fitted value is used as the fitting mean of the pixels in the column where the center point of the superimposed remote sensing image is located. The set of the fitting means of the pixels in each column of the superimposed remote sensing image is the fitting mean data.

2. The stripe noise detection method according to claim 1, characterized in that: The preset size of the window is 11.

3. The stripe noise detection method according to claim 1, characterized in that: The remote sensing images acquired by the rapid imager include visible light range images, near infrared range images and infrared range images. When the remote sensing images are visible light range images and near infrared range images, the threshold is set to 0.03; when the remote sensing images are infrared range images, the threshold is set to 0.

2.

4. The stripe noise detection method according to claim 3, characterized in that: Before dividing the satellite remote sensing images by time and superimposing the satellite remote sensing images within one hour to obtain the superimposed remote sensing images, the strip noise detection method further includes: The satellite remote sensing image acquired by the rapid imager is preprocessed, and the preprocessing includes quality inspection, geographic positioning, radiometric calibration processing, and converting the pixel values ​​of pixels in the visible light range image and the infrared range image into reflectance according to the following formula: ρ λ =SCALE×DN+OFFSET Where: λ is the preprocessed satellite remote sensing image, SCALE provides the slope for converting pixel values ​​into reflectance, DN is the pixel value of the satellite remote sensing image, and OFFSET provides the intercept for converting pixel values ​​into reflectance.

5. A strip noise detection device, characterized in that: include: A processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a computer program or instruction to implement the strip noise detection method according to any one of claims 1 to 4.

6. A computer storage medium, characterized in that: The computer storage medium stores instructions, and when the instructions are executed, the stripe noise detection method according to any one of claims 1 to 4 is implemented.

Images