A statistical characteristics analysis method for infrared clutter background images

Abstract

The invention belongs to the field of image processing technology, and discloses a statistical characteristic analysis method for infrared clutter background images. The method comprises the following steps: selecting an infrared scene image sequence set for analysis, and establishing a mathematical model and reasonable assumptions of the infrared image based on the spatiotemporal continuity of the image sequence; calculating the mean value and sample autocovariance of all samples, constructing a relationship curve between the sample mean, autocovariance and pixel position, and analyzing the stationary characteristics of the infrared image; using a goodness of fit test method to test the spatial distribution characteristics of the infrared clutter background image; and estimating the correlation length of the infrared background image under a wide-sense stationary condition. The invention can deeply analyze the stationary characteristics, spatial distribution characteristics and spatial correlation characteristics of the infrared clutter background, and can calculate the correlation length range of the clutter background, which is conducive to selecting appropriate filter parameters and is applicable to various backgrounds.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and relates to analyzing the statistical characteristics of an infrared clutter background image by using mathematical statistics theory and an image processing method, and constructing a correlation length calculation model of the infrared clutter background image. Background Art

[0002] Due to the characteristics of infrared imaging equipment with compact structure, high sensitivity and strong resolution, infrared detection technology has been widely used in many fields. At present, infrared imaging information processing has received widespread attention and people have conducted in-depth research on it. In infrared imaging information processing, one of the most basic studies is the description of infrared scene images in images. With the continuous improvement and development of infrared detector manufacturing technology, the internal noise of infrared detectors has been reduced to a fairly low level. In some applications, background fluctuation interference has become an obstacle affecting infrared detection and recognition. For example, in the process of infrared automatic target recognition, when the target is far away from the sensor, the contrast is low, and the target only occupies one or several pixels in the image, almost submerged in the undulating background.

[0003] Infrared images from the atmosphere often contain a lot of clutter caused by clouds, which is called background clutter in infrared image processing. Its statistical characteristics are related to the working band, spatial resolution and working conditions (such as working temperature and working bias voltage) of the infrared sensor. In the research of infrared image processing and recognition, it is crucial to analyze the statistical characteristics of the clutter background in the image and establish the corresponding mathematical model for small target detection and recognition.

[0004] According to the different working environments of infrared sensors, the background of infrared images can be divided into ground background, sky background (or atmosphere background), sea and sky background and space background. Different types of scenes have different statistical characteristics. The present invention only analyzes the statistical characteristics of infrared images formed by the clutter background of sky clouds, but its analysis method is also applicable to other situations such as sea and sky scenes and space scenes.

[0005] The analysis method of statistical characteristics of clutter background based on infrared images is a technical problem that needs to be solved urgently. Currently, no public research data directly related to this problem has been found. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a statistical characteristic analysis method for infrared fluctuating scene images, to provide technical support for infrared small target detection under atmospheric cloud background conditions, to select infrared scene image sequence sets for analysis, and to establish a mathematical model and reasonable assumptions for infrared images based on the spatiotemporal continuity of image sequences; then, using probability theory and mathematical statistics theory, to calculate the mean value and sample autocovariance of all samples, to analyze the stationary characteristics of infrared images; using χ 2 The goodness of fit test method is used to test the spatial distribution characteristics of the infrared clutter background image. Finally, under the condition of wide-sense stationarity, the correlation length of the infrared background image is estimated.

[0007] Specifically, the statistical characteristic analysis method for infrared clutter background images disclosed in the present invention comprises the following steps:

[0008] Select infrared scene image sequence set for analysis, and establish mathematical model and reasonable assumptions of infrared images based on the spatiotemporal continuity of image sequence;

[0009] Calculate the mean and sample autocovariance of all samples, construct the relationship curve between sample mean μ(m), autocovariance C(m; k) and pixel position, and analyze the stationary characteristics of infrared images;

[0010] Using χ 2 The goodness of fit test method is used to test the spatial distribution characteristics of the infrared clutter background image;

[0011] Estimate the correlation length of infrared background image under wide sense stationary conditions.

[0012] Furthermore, the infrared scene image sequence set is:

[0013] X={x i (m)|m=(x,y)}

[0014] Among them, x i (m) represents the i-th frame image in the sequence image set, the vector m = (x, y) represents the position of the pixel, i = 1, 2, ..., I, I is the total number of image frames in the image set, and x, y represent the horizontal and vertical coordinates in the image coordinate system respectively.

[0015] Furthermore, the following assumptions are made for the infrared sequence images:

[0016] (i) Any frame image in the sequence of images can be represented as a finite region in a two-dimensional discrete random field; the two-dimensional random field is denoted by x(m), -∞<|m|<∞;

[0017] (ii) Since any two frames in the image set are non-overlapping, it is assumed that the pixels in different image frames are uncorrelated and are independent and identically distributed; that is, they satisfy the relationship:

[0018]

[0019] Among them, x j (n) represents the j-th frame image in the sequence image set.

[0020] Furthermore, the calculation of the mean value and sample autocovariance of all samples includes:

[0021] Sequence image x i The optimal estimates of the expectation and autocovariance of (m) are the average values ​​of all samples in the image set. and the sample autocovariance Right now:

[0022]

[0023] random variable It must be a Gaussian distribution, and its mathematical expectation is μ(m) and its variance is:

[0024]

[0025] Likewise, the autocovariance is a Gaussian distribution, whose mathematical expectation is C(m;k) and variance is

[0026]

[0027] Furthermore, the mathematical expectation and autocovariance estimates of all pixel positions of the image are calculated; their 2σ error intervals are calculated respectively, namely:

[0028]

[0029] Finding position invariants within a 2σ error interval and The position invariant is independent of the pixel position, and the estimated values ​​of the grayscale mean and autocovariance of all pixels vary around the position invariant.

[0030] Furthermore, the analysis and verification of the stationary characteristics of the clutter background image includes:

[0031] When the number of sequence image samples is greater than 50 and the frequency of the position invariant exceeding the estimated value 2σ error interval is ≤5%, the clutter background image obeys a wide-sense stationary random process. If the frequency is much greater than 5%, the clutter background image is considered to be non-stationary.

[0032] Furthermore, the pixel sample selection and grayscale interval division based on the image sequence includes:

[0033] Select multiple pixels with equal pixel intervals in the horizontal and vertical directions from each frame of the image set as test samples;

[0034] The grayscale value of each pixel is between 0 and 255. i (i=0,1,…,255) represents the observation frequency of its value i in all samples, np i (i=0,1,…,255) represents the expected frequency of pixel gray value i when the image follows Gaussian distribution.

[0035] Furthermore, the test statistic χ 2 The expression is as follows:

[0036]

[0037] The hypothesis test formula is:

[0038] H0:f(x)=f0(x)

[0039] H1:f(x)≠f0(x)

[0040] Among them, f0(x) is the Gaussian distribution function, that is

[0041]

[0042] μ and σ 2 The optimal estimates of and The image grayscale is represented by X, then the hypothesis test formula is equivalent to:

[0043] H0: X follows a normal distribution

[0044] H1: X does not follow a normal distribution;

[0045] Calculate the observed frequency and expected frequency when the hypothesis H0 is established, and calculate χ 2 , and compare χ 2 Is it less than χ 2 Find the quantile in the distribution table on the upper side of the distribution.

[0046] Furthermore, the estimating the correlation length of the infrared clutter background image includes:

[0047] Under the condition of wide-sense stationarity of infrared clutter background image, calculate the grayscale average of all pixels in the i-th frame image in the infrared image sequence

[0048]

[0049] Using the spatial averaging method, the autocovariance of each frame in the infrared image sequence is calculated.

[0050]

[0051] Where A(k) represents the number of all pixel points with offset value k in the i-th frame image;

[0052] Calculate the autocovariance of all images in an infrared image sequence The average

[0053]

[0054] in is the autocovariance function of the infrared image sequence, The relationship with k is approximately a decaying exponential curve;

[0055] Use two-dimensional exponential function to analyze the autocovariance function The correlation length L of the infrared clutter background image is obtained by fitting. The isotropic two-dimensional exponential function is used for fitting as follows:

[0056]

[0057] Where L is the correlation length value of the undulating background in the image, A is a constant, k is the two-dimensional offset on the image plane, and a and b represent the sizes of the sensor spatial resolution unit in the horizontal and vertical directions, respectively.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The present invention uses a combination of statistical methods and numerical calculations to deeply analyze the stability, spatial distribution and spatial correlation characteristics of infrared clutter background for infrared sequence images. The research results are of great significance to the selection of filter models and filter parameters for image noise filtering, target detection, etc. The present invention has the following obvious characteristics and advantages:

[0060] (1) Through image processing and pixel calculation, the wide-sense stationary characteristics and spatial distribution characteristics of specific infrared clutter background images can be deeply analyzed and verified;

[0061] (2) For infrared sequence images of specific scenes, by establishing a correlation model between clutter background pixels, the correlation length range of the clutter background can be calculated, which is conducive to selecting appropriate filter parameters during noise suppression and target signal detection.

[0062] (3) The infrared image characteristic analysis method of the present invention has strong adaptability and is applicable to sky background, sea and sky background, space background, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is the overall flow chart of the present invention;

[0064] Figure 2 Sampling mean along the middle row of the image The relationship curve with the pixel position m;

[0065] Figure 3 Sampling the mean along the middle column of the image The relationship curve with the pixel position m;

[0066] Figure 4 The relationship between the estimated autocovariance and the pixel position m along the middle row of the image (the image set is

[0067] IR1, hysteresis value k = (2,2));

[0068] Figure 5 The relationship between the estimated autocovariance and the pixel position m along the middle column of the image (the image set is

[0069] IR1, hysteresis value k = (2,2));

[0070] Figure 6 The relationship between the estimated autocovariance and the pixel position m along the middle row of the image (the image set is

[0071] IR1, hysteresis value k = (25,25));

[0072] Figure 7 The relationship curve between the estimated autocovariance and the pixel position m along the middle column of the image (the image set is

[0073] IR1, hysteresis value k = (25,25));

[0074] Figure 8 The three-dimensional surface diagram of the autocovariance function of the measured infrared background under wide-sense stationary conditions;

[0075] Fig. 9 Distribution diagram of the autocovariance function of the measured infrared background in the image row direction under wide sense stationary conditions;

[0076] Fig.10 Distribution diagram of the autocovariance function of the measured infrared background in the image column direction under wide-sense stationary conditions. DETAILED DESCRIPTION

[0077] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.

[0078] according to Figure 1 As shown, the whole process is divided into the following steps:

[0079] S1: Establishing a mathematical model for infrared sequence images

[0080] S11 defines the infrared sequence image set as

[0081] X={x i (m)|m=(x,y)} (1)

[0082] Among them, x i (m) represents the i-th frame image in the sequence image set, the vector m = (x, y) represents the position of the pixel, i = 1, 2, ..., I, I is the total number of image frames in the image set, x, y represent the horizontal coordinate and vertical coordinate in the image coordinate system, that is, x = 0, 1, ..., M-1 and y = 0, 1, ..., N-1.

[0083] S12 In order to facilitate quantitative analysis, the following reasonable assumptions are made for the infrared sequence images

[0084] (i) Any frame image in the sequence of images can be represented as a finite region in a two-dimensional discrete random field; the two-dimensional random field is denoted by x(m), -∞<|m|<∞;

[0085] (ii) Since any two frames in the image set are non-overlapping, it is assumed that the pixels in different image frames are uncorrelated and are independent and identically distributed; that is, they satisfy the relationship:

[0086]

[0087] S2: Analysis and verification of the stability characteristics of infrared clutter background images

[0088] S21 calculates the sample mean μ(m) and autocovariance in the image set Variance

[0089] According to statistical theory, when the number of samples of the sequence image I is large (here, I is the number of frames of the sequence image, usually I ≥ 100, which can be considered a large sample), the sequence image x i The optimal estimates of the expectation and autocovariance of (m) are the average values ​​of all samples in the image set. and the sample autocovariance Right now

[0090]

[0091] From formula (2), we can see that when I is large enough, can be represented by enough independent and identically distributed random variables x i (m), and each random variable The effect of is small, so according to the central limit theorem, the random variable It must be a Gaussian distribution, and its mathematical expectation is μ(m) and its variance is:

[0092]

[0093] It can also be shown that the autocovariance is a Gaussian distribution, whose mathematical expectation is C(m;k) and variance is

[0094]

[0095] S22 constructs sample mean μ(m), autocovariance Relationship curve with pixel position

[0096] For the infrared sequence images to be analyzed, the mathematical expectation and autocovariance estimation of all pixel positions of the image are first calculated according to equations (2) and (3), respectively. Then, their 2σ error intervals are calculated according to equations (4) and (5), respectively, namely:

[0097]

[0098] Figures 2 to 7 The experimental results of a certain infrared image sequence set IR1 are given. Figure 2 and 3 The sampling mean of each pixel along the middle row and middle column of the image is plotted respectively. (As shown by the solid line in the figure). Figure 4 , 5 and Figure 6 , 7 The estimated autocovariance of each pixel along the middle row and middle column of the image for different values ​​of lag k is given. (There is no particular significance in taking the middle row and middle column of the image for the experiment here; the same result can be obtained by using other rows or columns).

[0099] Depend on Figures 2 to 7 It can be seen that the estimated values ​​of the grayscale mean and autocovariance of all pixels vary around a fixed value, which is independent of the pixel position and is therefore called a position invariant (denoted as and ). It can be found that almost all pixels in the image are and All of them are within the 2σ error range.

[0100] S23 Analysis and verification of the stationary characteristics of clutter background images

[0101] Based on the numerical calculation results of step S22, it can be known from statistical theory that when the number of sequence image samples is greater than 50 and the frequency (number of times) that the position invariant exceeds the estimated value 2σ error interval is ≤5%, the clutter background image obeys a wide-sense stationary random process. If the frequency is much greater than 5%, the clutter background image is considered to be non-stationary.

[0102] S3, Test of spatial distribution characteristics of infrared images based on goodness of fit test method

[0103] The basic idea of ​​goodness of fit test is to test whether the overall distribution obeys a certain distribution when the overall distribution is unknown, that is, to use the sample observation value to infer the overall distribution and make an answer whether to approve the original hypothesis. Its essence is to test the goodness of fit of the sample to the theoretical distribution. This test is called goodness of fit test. 2 The goodness of fit test method is used to analyze the spatial distribution characteristics of the infrared background.

[0104] S31 Pixel sample selection and grayscale interval division based on image sequence

[0105] (i) Select 4 pixels with equal pixel intervals in the horizontal and vertical directions from each frame of the image set as test samples, with a total of 4×4=16 pixels. Then, the number of samples that meet the above conditions in the 50-frame image set (taking IR1 as an example) is n=16×50=800.

[0106] (ii) The grayscale value of each pixel (sample) is between 0 and 255. Let D i (i=0,1,…,255) represents the observation frequency of its value i in all samples, np i (i=0,1,…,255) represents the expected frequency (or theoretical frequency) of pixel gray value i when the image obeys Gaussian distribution.

[0107] (iii) In practical applications, in order to reduce the amount of calculation, the gray levels can be appropriately merged. The following seven non-overlapping intervals can be defined: [0, 35], [36, 65], [66, 95], [96, 125], [126, 155], [156, 185], [186, 255].

[0108] Hypothesis test of grayscale distribution characteristics of S32 pixels

[0109] Based on the pixel samples and grayscale interval division in step S31, the test statistic χ can be obtained: 2 The expression

[0110]

[0111] Hypothesis Testing

[0112] H0:f(x)=f0(x)

[0113] H1:f(x)≠f0(x) (7)

[0114] Among them, f0(x) is the Gaussian distribution function, that is

[0115]

[0116] From step S22, we know that μ and σ 2 The optimal estimates of and Assume that the image grayscale is represented by X, then equation (7) is equivalent to

[0117] H0: X follows a normal distribution

[0118] H1: X does not follow a normal distribution

[0119] S33 Analysis and verification of image background spatial characteristics based on test statistics

[0120] Based on the hypothesis test in step S32, the observed frequency, expected frequency and other related quantities are calculated when the hypothesis H0 is established. Taking the image set IR1 as an example, the observed frequency and expected frequency results are shown in the following table

[0121] Table 1 Observed frequency and expected frequency and other related quantities (n = 800)

[0122] Interval number <![CDATA[D i ]]> <![CDATA[D i 2 ]]> <![CDATA[np i ]]> <![CDATA[D i 2 / e.g i ]]> 1 48 2304 33.12 69.565 2 80 6400 86.16 74.280 3 160 25600 172.32 148.561 4 256 65536 216.8 302.288 5 128 16384 172.32 95.079 6 96 9216 86.16 106.964 7 32 1024 33.12 30.918

[0123] Given a significance level of α = 0.05, then

[0124] From χ 2 The distribution table on the upper side of the distribution finds the quantile That is, the critical region is (9.488,∞). Calculate the value of the test statistic

[0125]

[0126] Due to χ 2 <9.488, so assuming H0 is true, it can be considered that the background images in image set IR1 obey Gaussian distribution.

[0127] S4 estimates the correlation length of infrared clutter background images

[0128] S41 Under the condition of wide-sense stationarity of infrared clutter background image, the grayscale average value of all pixels in the i-th frame image of the infrared image sequence can be calculated using formula (8):

[0129]

[0130] S42 uses the spatial averaging method and uses formula (9) to calculate the autocovariance of each frame in the infrared image sequence:

[0131]

[0132] Where A(k) represents the number of all pixel points with an offset value of k in the i-th frame image.

[0133] S43 Calculates the autocovariance of all images in the infrared image sequence The average

[0134]

[0135] say is the autocovariance function of the infrared image sequence, The relationship with k is approximately a decaying exponential curve, see Figure 8-Figure 10 .

[0136] S44 uses a two-dimensional exponential function to calculate the autocovariance function Fitting is performed to obtain the correlation length L of the infrared clutter background image. The fitting adopts an isotropic two-dimensional exponential function as shown in formula (11):

[0137]

[0138] Where L is the correlation length value of the undulating background in the image. Where A is a constant, k is the two-dimensional offset on the image plane, and a and b represent the size of the sensor spatial resolution unit in the horizontal and vertical directions, respectively.

[0139] Compared with the prior art, the present invention has the following beneficial effects:

[0140] The present invention uses a combination of statistical methods and numerical calculations to deeply analyze the stability, spatial distribution and spatial correlation characteristics of infrared clutter background for infrared sequence images. The research results are of great significance to the selection of filter models and filter parameters for image noise filtering, target detection, etc. The present invention has the following obvious characteristics and advantages:

[0141] (1) Through image processing and pixel calculation, the wide-sense stationary characteristics and spatial distribution characteristics of specific infrared clutter background images can be deeply analyzed and verified;

[0142] (2) For infrared sequence images of specific scenes, by establishing a correlation model between clutter background pixels, the correlation length range of the clutter background can be calculated, which is conducive to selecting appropriate filter parameters during noise suppression and target signal detection.

[0143] (3) The infrared image characteristic analysis method of the present invention has strong adaptability and is applicable to sky background, sea and sky background, space background, etc.

[0144] As used herein, the word "preferred" is intended to be used as an example, instance, or illustration. Any aspect or design described herein as "preferred" is not necessarily to be construed as being more advantageous than other aspects or designs. On the contrary, the use of the word "preferred" is intended to present concepts in a specific way. The term "or" as used in this application is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" means any one of the naturally included permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.

[0145] Moreover, although the present disclosure has been shown and described with respect to one or implementations, those skilled in the art will think of equivalent variations and modifications based on the reading and understanding of this specification and the accompanying drawings. The present disclosure includes all such modifications and variations, and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the above-mentioned components (such as elements, etc.), the terms used to describe such components are intended to correspond to any component (unless otherwise indicated) that performs the specified function of the component (such as it is functionally equivalent), even if the structure is not equivalent to the disclosed structure of the function in the exemplary implementation of the present disclosure shown herein. In addition, although the specific features of the present disclosure have been disclosed with respect to only one of several implementations, such features can be combined with one or other features of other implementations that may be desired and advantageous for a given or specific application. Moreover, insofar as the terms "including", "having", "containing" or their variations are used in specific embodiments or claims, such terms are intended to be included in a manner similar to the term "comprising".

[0146] The functional units in the embodiments of the present invention may be integrated into a processing module, or each unit may exist physically separately, or multiple or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc. The above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.

[0147] To sum up, the above embodiment is an implementation mode of the present invention, but the implementation mode of the present invention is not limited by the embodiment. Any other changes, modifications, substitutions, combinations, and simplifications that deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A statistical characteristic analysis method for infrared clutter background images, characterized in that: The following steps are involved: Select infrared scene image sequence set for analysis, and establish mathematical model and reasonable assumptions of infrared images based on the spatiotemporal continuity of image sequence; The infrared scene image sequence set is: X={x i (m)|m=(x,y)} Among them, x i (m) represents the i-th frame image in the sequence image set, the vector m = (x, y) represents the position of the pixel, i = 1, 2, ..., I, I is the total number of image frames in the image set, x, y represent the horizontal coordinate and vertical coordinate in the image coordinate system respectively; The following assumptions are made for the infrared sequence images: (i) Any frame image in the sequence of images can be represented as a finite region in a two-dimensional discrete random field; the two-dimensional random field is denoted by x(m), -∞<|m|<∞; (ii) Since any two frames in the image set are non-overlapping, it is assumed that the pixels in different image frames are uncorrelated and are independent and identically distributed; that is, they satisfy the relationship: Among them, x j (n) represents the jth frame image in the sequence image set; Calculate the mean and sample autocovariance of all samples, and construct the sample mean μ(m), autocovariance The relationship curve between m and pixel position is used to analyze the stability characteristics of infrared images, where m is the pixel position and k is the lag value; Sequence image x i The optimal estimates of the expectation and autocovariance of (m) are the average values ​​of all samples in the image set. and the sample autocovariance Right now: random variable It must be a Gaussian distribution, and its mathematical expectation is μ(m) and its variance is: Likewise, the autocovariance is a Gaussian distribution, whose mathematical expectation is C(m;k) and variance is: Using χ 2 The goodness of fit test method is used to test the spatial distribution characteristics of the infrared clutter background image; Under the condition of wide sense stability, the correlation length of the infrared background image is estimated, including: Under the condition of wide-sense stationarity of infrared clutter background image, calculate the grayscale average of all pixels in the i-th frame image in the infrared image sequence M and N are the maximum values ​​of the horizontal and vertical coordinates in the image, respectively; Using the spatial averaging method, the autocovariance of each frame in the infrared image sequence is calculated. Where A(k) represents the number of all pixel points with offset value k in the i-th frame image; Calculate the autocovariance of all images in an infrared image sequence The average in is the autocovariance function of the infrared image sequence, The relationship with k is approximately a decaying exponential curve; Use two-dimensional exponential function to analyze the autocovariance function The correlation length L of the infrared clutter background image is obtained by fitting. The isotropic two-dimensional exponential function is used for fitting as follows: Where L is the correlation length value of the undulating background in the image, A is a constant, k is the two-dimensional offset on the image plane, a, b represent the size of the sensor spatial resolution unit in the horizontal and vertical directions, respectively, and x, y represent the abscissa and ordinate in the image coordinate system, respectively.

2. The statistical characteristic analysis method for infrared clutter background images according to claim 1 is characterized in that: Calculate the mathematical expectation and autocovariance estimates of all pixel positions in the image; calculate their 2σ error intervals respectively, namely: Finding position invariants within a 2σ error interval and The position invariant is independent of the pixel position, and the estimated values ​​of the grayscale mean and autocovariance of all pixels vary around the position invariant.

3. The statistical characteristic analysis method for infrared clutter background images according to claim 2 is characterized in that: The analysis and verification of the stationary characteristics of the clutter background image includes: When the number of sequence image samples is greater than 50 and the frequency of the position invariant exceeding the estimated value 2σ error interval is ≤5%, the clutter background image obeys a wide-sense stationary random process. If the frequency is much greater than 5%, the clutter background image is considered to be non-stationary.

4. The statistical characteristic analysis method for infrared clutter background images according to claim 3 is characterized in that: The pixel sample selection and grayscale interval division based on the image sequence includes: Select multiple pixels with equal pixel intervals in the horizontal and vertical directions from each frame of the image set as test samples; The grayscale value of each pixel is between 0 and 255. i (i=0,1,…,255) represents the observation frequency of its value i in all samples, np i (i=0,1,…,255) represents the expected frequency of pixel gray value i when the image follows Gaussian distribution.

5. The statistical characteristic analysis method for infrared clutter background images according to claim 4 is characterized in that: Test statistic χ 2 The expression is as follows: The hypothesis test formula is: H0:f(x)=f0(x) H1:f(x)≠f0(x) Among them, f0(x) is the Gaussian distribution function, that is μ and σ 2 The optimal estimates of and The image grayscale is represented by X, then the hypothesis test formula is equivalent to: H0: X follows a normal distribution H1: X does not follow a normal distribution; Calculate the observed frequency and expected frequency when the hypothesis H0 is established, and calculate χ 2 , and compare χ 2 Is it less than χ 2 Find the quantile in the distribution table on the upper side of the distribution.

Images