A sampling method for non-destructive readout of infrared detectors

Abstract

A sampling method for non-destructive readout infrared detectors, including: testing the response linearity of the non-destructive readout infrared detector, setting the full well detection threshold according to the range of the linear region; in the readout process of a single frame image, reading the storage parameters of the pixel after the sampling of a single pixel is completed, calculating and storing new parameters, and performing the above operations on each pixel row by row and column by column in accordance with the order of the detector output image data; ending the integration after completing the readout of a specified number of frames, and calculating the final output result according to the storage parameters. Compared with the commonly used sampling methods for general infrared detectors such as correlated double sampling, the method of the present invention can further reduce the readout noise; compared with the special sampling methods for non-destructive readout infrared detectors such as standard ramp sampling, this method has the same noise reduction effect as the standard ramp sampling, while also greatly reducing the storage space requirement.

Description

Technical Field

[0001] The invention relates to a sampling method for non-destructive reading of infrared detectors, which can be applied to remote sensing infrared cameras and belongs to the field of image processing technology. Background Art

[0002] With the successful development of non-destructive readout infrared detectors, corresponding readout algorithms have also been developed. Non-destructive readout infrared detectors are usually used in scenarios with ultra-long integration times, such as space telescopes. Under conditions of ultra-long integration time and weak target radiation intensity, the traditional correlated double sampling (CDS) method of general detectors cannot effectively eliminate readout noise due to the long integration time.

[0003] In addition to correlated double sampling, the commonly used sampling algorithms in the world are Fowler sampling and ramp sampling (UTR). In ramp sampling, the data may be processed by multiple accumulation (MACC) and differential multiple accumulation (DMACC). Fowler sampling performs multiple sampling after the reset is completed, and performs the same number of samplings again when the integration is almost completed, and uses the multiple sampling average of the difference between the previous and next image data as the output result.

[0004] Slope sampling is to sample at equal intervals during the entire integration period, and fit the slope through the least squares method and output it. Conventional slope sampling requires the storage of all pixels, which are read out from the storage unit after sampling and fitted. For the area array detector, the amount of image data itself is extremely large. As the number of sampling points increases, the amount of data increases linearly. The standard slope sampling algorithm has extremely high requirements for the size of the storage space and the data throughput rate. Therefore, the image data is usually processed by multiple accumulation or differential accumulation, and the average of the data of several adjacent sampling points is used as a sampling point to reduce the amount of data exponentially. However, multiple accumulation processing usually selects 2, 4 or 8 pixels as a group. Continuing to increase the number of pixels in each group will cause the fitting line to deviate seriously from the actual curve, so the data compression is limited. Taking the Near Infrared Spectrometer (NIRS) in the James Webb Space Telescope (JWST) as an example, its array size is 2K×2K, with a total of 88 sampling points, which are divided into 22 groups using multiple accumulation processing. Assuming that the image data is quantized at 16 bits, it takes 1.477Gb to store all the pixel data in one integration process, which requires a large number of expensive hardware storage devices. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present invention aims to propose a sampling method for non-destructive readout of infrared detectors, which mainly solves the problem that non-destructive readout of infrared detectors with ultra-long integration time occupies a huge amount of data storage space during the process of using the ramp sampling algorithm. To this end, the technical solution adopted by the present invention is to further simplify the least squares calculation formula based on the characteristics of ramp sampling and equal interval sampling, store calculation parameters instead of storing all pixel data during the imaging process, and read out the calculation parameters after the integration stage.

[0006] The technical solution of the present invention is:

[0007] A sampling method for non-destructive readout of infrared detectors includes the following steps:

[0008] 1) Use standard black body test to non-destructively read out invalid pixels of infrared detectors;

[0009] 2) Standard black body test non-destructively reads the nonlinearity of the infrared detector response, and extracts the starting value θ0 and the ending value θ of the linear region where the nonlinearity is ≤1% in the response curve; wherein the data in the response curve does not include the data of invalid pixels;

[0010] 3) Use a non-destructive readout detector to perform n-integral imaging on the scene target area to obtain n+1 images; after removing invalid pixels from the image, use the linear interval range [θ0, θ] to process each pixel in the image respectively, obtain the pixel full well flag F(p, q) corresponding to each pixel, and obtain the algorithm output value V(p, q) corresponding to each pixel as the preprocessing result of the remote sensor; where p and q are both integers, representing the row number and column number of the pixel respectively;

[0011] 4) Use the image receiving system to receive the preprocessing results sent by the remote sensor, perform image post-processing on the preprocessing results according to the pixel full well flag F(p, q) corresponding to each pixel, and obtain the processed image DN value V′(p, q) as the image sampling result.

[0012] Preferably, the method for determining the pixel full well flag F(p, q) of the pixel position of the p-th row and the q-th column in the image is specifically as follows:

[0013] The DN values ​​of all pixels in the pth row and qth column in the n+1 images are taken as a set of data sequences to be screened;

[0014] If there is a data point with a DN value greater than θ in the data sequence to be screened of the p-th row and q-th column pixel, then the pixel full well flag F(p,q)=1 at the p-th row and q-th column pixel position; otherwise, F(p,q)=0.

[0015] Preferably, the method of the algorithm output value V(p,q) corresponding to each pixel is specifically as follows:

[0016] Using the linear interval range [θ0, θ], the data in the to-be-screened data sequence of each pixel are screened and processed respectively, and the data belonging to the linear interval range are obtained to form the valid data sequence of each pixel;

[0017] According to the pixel full well flag of each pixel position and the valid data sequence corresponding to each pixel position, the algorithm output value V(p,q) corresponding to each pixel is determined respectively.

[0018] Preferably, if there are two or more data points in the valid data sequence corresponding to the pixel in the pth row and qth column, and the pixel full well flag F(p,q) of the pixel position is 0, then the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column is as follows:

[0019]

[0020] S i (p,q)=S i-1 (p,q)+R i (p,q)×(2i-m)

[0021] SR i (p,q)=SR i-1 (p,q)+R i (p,q)

[0022] Among them, the m+1 data in the valid data sequence corresponding to the pixel in the p-th row and q-th column are numbered in sampling order starting from 0. When i is 0, S0(p,q)=-m×R0(p,q); SR0(p,q)=R0(p,q); R0(p,q) is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the p-th row and q-th column.

[0023] Preferably, if there are two or more data points in the valid data sequence corresponding to the pixel in the pth row and qth column, and the pixel full well flag F(p,q) of the pixel position is 1, then the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column is as follows:

[0024]

[0025] S i (p,q)=S i-1 (p,q)+R i (p,q)×(2i-m)

[0026] SR i (p,q)=SR i-1 (p,q)+Ri (p,q)

[0027] Among them, the m+1 data in the valid data sequence corresponding to the pixel in the p-th row and q-th column are numbered in sampling order starting from 0. When i is 0, S0(p,q)=-m×R0(p,q); SR0(p,q)=R0(p,q); R0(p,q) is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the p-th row and q-th column.

[0028] Preferably, if there is only one data point in the valid data sequence corresponding to the pixel in the pth row and qth column or the valid data sequence is empty, and the pixel full well flag F(p,q) of the pixel position is an arbitrary value, then the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column is 0.

[0029] Preferably, the method for obtaining the processed image DN value V′(p,q) is specifically as follows:

[0030] When F(p,q)=1, V′(p,q)=m×V(p,q);

[0031] When F(p,q)=0, V′(p,q)=V(p,q).

[0032] Preferably, step 4) further includes adjusting the total integration time of the non-destructive readout infrared detector according to the pixel full well flag F(p, q) corresponding to each pixel and the processed image DN value V′(p, q) to improve the subsequent image output quality.

[0033] Preferably, when there is a pixel with a pixel full well flag F(p, q)=1 in the corresponding pixel of the scene target in the image, the method of adjusting the imaging parameters of the non-destructive readout infrared detector is specifically as follows:

[0034] If there are pixels with a DN value of 0 in the processed image among the corresponding pixels of the scene target, the total integration time is increased;

[0035] If there are pixels in the corresponding pixels of the scene target whose DN value of the processed image is higher than 150% of the saturation value θ, the total integration time is shortened;

[0036] In other cases, there is no need to non-destructively read out the imaging parameters of the infrared detector during the subsequent imaging process.

[0037] Preferably, when the pixel full well flags F(p, q) of all positions in the pixels corresponding to the scene target in the image are 0, the method for adjusting the imaging parameters of the non-destructive readout infrared detector is specifically as follows:

[0038] If the minimum DN value of the processed image corresponding to the pixel of the scene target in the image is greater than 50% of the saturation value θ, there is no need to adjust the imaging parameters; otherwise, the total integration time should be increased so that the minimum DN value of the processed image is greater than 50% of the saturation value θ.

[0039] Compared with the prior art, the advantages of the present invention are mainly reflected in the following aspects:

[0040] (1) The present invention introduces the storage parameter S i (p,q),SR i (p, q) and output value V(p, q), using stored process parameters instead of storing all pixel data, greatly reducing the storage space requirements of the slope sampling algorithm;

[0041] (2) The present invention combines practical engineering applications and sets the start and end values ​​of the linear region of the detector to avoid the influence of the nonlinearity of the detector itself on the linear fitting;

[0042] (3) The present invention combines practical engineering applications and selects V′(p,q) rather than the slope as the final image, so that the physical meaning of the DN value of the processed image is similar to that of the original image of a general detector, and various existing image processing algorithms can be directly used. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the method of the present invention;

[0044] Figure 2 This is a matlab software simulation result diagram of the white noise reduction effect of the present invention. DETAILED DESCRIPTION

[0045] like Figure 1 The present invention provides a sampling method for non-destructive readout of infrared detectors, comprising the following steps:

[0046] (1) According to GB / T 17444-2013, the standard black body test is used to non-destructively read out the invalid pixel coordinates of the infrared detector. The invalid pixel data should not be included in the subsequent nonlinear calculation.

[0047] 1.1 Set the number of integration times per frame of the non-destructive readout detector to 1, use a standard black body for imaging, set the black body temperature to 20°C, save f frames of image data, and f is recommended to be more than 100 frames.

[0048] 1.2 Change the black body temperature to 35°C and save f frames of image data. During the test, other suitable black body temperature ranges can be selected according to the expected radiation power and wavelength characteristics of the observed target.

[0049] 1.3 After the test is completed, calculate the mean DN value increment of all pixels in the M rows and N columns of the f frame image

[0050]

[0051] Among them, the DN values ​​of the pixels in row p and column q at 20℃ and 35℃ are recorded as I T＝20 (p,q,i) and I T＝35 (p,q,i).

[0052] 1.4 Calculate the incremental mean of the DN value of each frame image of a single pixel f

[0053]

[0054] 1.5 Determine the DN value increment of each pixel one by one. If Then the pixel is a dead pixel, and the coordinates of the dead pixel are recorded.

[0055] 1.6 Use the remaining pixel data to calculate the standard deviation STD of each pixel f frame image when the black body temperature is 20℃ T＝20 (p,q), and its mean like If the pixel is an overheated pixel, the coordinates of the overheated pixel are recorded. The dead pixel and overheated pixel data are not involved in the subsequent nonlinear calculation.

[0056] (2) The output voltage value of the detector is not strictly linearly related to the accumulated received energy. The linearity is better when integrated in the middle of the potential well, which is called the linear region. This varies depending on the input unit of the readout circuit. This causes the corresponding relationship between the image data voltage DN value and the actual radiation power to change as the imaging process progresses. Therefore, it is necessary to discard the data points in the nonlinear region when fitting. First, the linear region test of the detector needs to be completed. According to GB / T 17444-2013, a standard black body test is used to non-destructively read out the nonlinearity of the infrared detector response, and the starting value θ0 and the ending value θ of the linear region with a nonlinearity of ≤1% in the response curve are screened out as the linear interval range [θ0, θ];

[0057] 2.1 After removing dead pixels and overheated pixel data, reselect the appropriate blackbody temperature based on the upper and lower limits of the integration time that the system can achieve, so that all pixels are saturated at the upper limit of the integration time and the average pixel DN value is lower than the saturation value by 0.5% at the lower limit of the integration time.

[0058] 2.2 Ensure that the detector gain and black body temperature remain unchanged, gradually change the integration time for imaging, output more than 30 frames of images for each integration time, and use the average of all pixels in all images as the value corresponding to the integration time. The specific integration time point starts with the lower limit of the integration time, and the increment step is 1% to 1.5% of the difference between the upper and lower limits of the integration time. Among them, the 15% to 85% part between the upper and lower limits of the integration time can only select 5 to 10 evenly distributed integration time points.

[0059] 2.3 Draw the relationship curve between the mean DN value of all pixels and the integration time. The specific mean DN value of all pixels is: obtain the mean DN value of all pixels on each picture at each integration point time, add up the mean DN values ​​of all pixels in multiple pictures at each integration point time, and then average them to obtain the mean DN value of all pixels at each integration point time, so as to obtain the relationship curve between the mean DN value of all pixels and the integration time. Use the least squares method to perform linear fitting to obtain the fitting curve, and calculate the nonlinearity NL of the data, that is, the maximum deviation percentage of each measurement point;

[0060]

[0061] Among them, I j is the mean DN value of all pixels at the jth integration point time, I j ′ is the estimated value of the mean DN value of all pixels at the jth integration point of the fitting curve.

[0062] 2.4 Usually, the linearity of the infrared detector is poor in the front and back sections of the integration. Discard the data points with a deviation percentage greater than 1% (if the nonlinearity is greater than 1%, remove the data points with a deviation percentage greater than 1%) and repeat the process of 2.3. Repeat 2.3 to 2.4 until the nonlinearity of the selected test data is ≤1%. Record the image DN values ​​corresponding to the minimum and maximum sampling time, and record them as the DN value θ0 of the starting test point of the linear region and the DN value θ of the ending test point.

[0063] Image preprocessing

[0064] Use a non-destructive readout detector to perform n-integral imaging to obtain n+1 images; use a linear interval range [θ0, θ] to process the pixels in the p-th row and q-th column of the image respectively to obtain the pixel full well flag F(p, q) corresponding to each pixel and the algorithm output value V(p, q) corresponding to each pixel; the steps are as follows:

[0065] (3) Perform n-integral imaging of the scene target using a non-destructive readout detector to obtain n+1 image data; remove invalid pixel data; determine the pixel full well flag F(p, q) corresponding to each pixel position in the image; obtain the pixel that satisfies the linear interval range at each pixel position in the n+1 images as the valid data sequence corresponding to each pixel position;

[0066] 3.1 After completing the linear region test, enter the normal imaging mode and set the sampling times of the non-destructive readout detector to n, that is, a total of n+1 images will be output before each reset. The sampling interval, that is, the integration time each time, is Δt, and the subsequent parameters all start counting from 0.

[0067] 3.2 Receive n+1 image data output by the detector, remove invalid pixel data on the image data, and use the DN values ​​of all pixels in the pth row and qth column in the n+1 images as a group of data sequences to be screened, totaling p×q groups of data sequences to be screened;

[0068] 3.3 If there is a data point whose DN value is greater than θ in the data sequence to be screened of the pixel in the p-th row and q-th column, the pixel full well flag F(p,q)=1 at the pixel position in the p-th row and q-th column; otherwise, F(p,q)=0;

[0069] 3.4 Using the linear interval range [θ0, θ], the data in the to-be-screened data sequence of each pixel are screened and processed respectively, and the data belonging to the linear interval range are obtained to form the valid data sequence of each pixel; the data in the valid data sequence are numbered in sequence according to the sampling order, and the first sampled data point sequence number i is recorded as 0, and the last sampled point sequence number i is recorded as m. m is a positive integer less than or equal to n;

[0070] (4) According to the pixel full well flag of each pixel position and the valid data sequence corresponding to each pixel position, the algorithm output value corresponding to each pixel is determined respectively, and after traversing the valid data sequences corresponding to all pixels, the preprocessing work is completed and output to the image receiving system; then proceed to step (5); Steps (3) and (4) belong to the image preprocessing link and are completed by the remote sensor. The method of determining the algorithm output value V(p,q) corresponding to each pixel in step (4) is specifically as follows:

[0071] If there are two or more undiscarded data points in the valid data sequence corresponding to the pixel in the pth row and qth column of A, and F(p,q)=0 at the pixel position, then enter B; if there are two or more undiscarded data points, and F(p,q)=1, then enter C; if there is only one data point or the valid data sequence is empty, F(p,q) is an arbitrary value, then enter D;

[0072] B numbers the m+1 data in the valid data sequence corresponding to the pixel in the pth row and qth column in sampling order starting from 0, and determines and outputs the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column using the data points in the valid data sequence;

[0073]

[0074] S i (p,q)=S i-1 (p,q)+R i (p,q)×(2i-m)

[0075] SR i (p,q)=SR i-1 (p,q)+Ri (p,q)

[0076] Among them, when i is 0, S0(p,q)=-m×R0(p,q); SR0(p,q)=R0(p,q); R0(p,q) is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the pth row and qth column. This output value is m times the single increment obtained by linear fitting of a total of m+1 data points, and its physical meaning is similar to the difference between the DN value of the mth image and the DN value of the 0th image.

[0077] C numbers the m+1 data in the valid data sequence corresponding to the pixel in the pth row and qth column in sampling order starting from 0, and determines and outputs the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column using the data points in the valid data sequence;

[0078]

[0079] S i (p,q)=S i-1 (p,q)+R i (p,q)×(2i-m)

[0080] SR i (p,q)=SR i-1 (p,q)+R i (p,q)

[0081] Among them, when i is 0, S0(p,q)=-m×R0(p,q); SR0(p,q)=R0(p,q); R0(p,q) is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the pth row and the qth column. This output value is 1 / m of the output value of step B, which means the single increment of adjacent sampling points obtained by fitting. This output value is selected because there is a bit width limit on the output data during the application process, and under the current conditions, the output value of step B will overflow.

[0082] D sets the algorithm output value V(p,q) corresponding to the pixel in the pth row and qth column to 0.

[0083] Image post-processing

[0084] (5) Use the ground image receiving system to obtain the algorithm output value V(p,q) corresponding to each pixel transmitted by the remote sensor, perform image post-processing, and obtain the processed image DN value V′(p,q) corresponding to the pixel in the pth row and qth column; when F(p,q)=1, V′(p,q)=m×V(p,q);

[0085] When F(p,q)=0, V′(p,q)=V(p,q).

[0086] Since the output value of step C and the output value of step B have different physical meanings, the two output values ​​may exist simultaneously in the same frame of image. Therefore, after completing the image data reception, the external image receiving system adjusts the output value of step C so that the physical meaning of the DN value of each pixel in the entire frame of image after processing is the same.

[0087] (6) According to the processed image DN value corresponding to each pixel and the pixel full well flag F(p, q) corresponding to each pixel position, the imaging parameters of the non-destructive readout infrared detector are adjusted. The steps are as follows:

[0088] 6.1 If there is a pixel with F(p, q)=1 in the pixel full well flag of the pixel position corresponding to the scene target in the image, then go to 6.2; if all the pixel full well flags of the pixel position corresponding to the scene target in the image are F(p, q)=0, then go to 6.3;

[0089] 6.2 If there are pixels with a DN value of 0 in the processed image among the corresponding pixels of the scene target in the image, the total integration time should be significantly increased; if the DN values ​​of the processed image are not 0 and are higher than 150% of the saturation value, the total integration time should be appropriately reduced; in other cases, there is no need to adjust the imaging parameters;

[0090] 6.3 If the minimum DN value of the processed image corresponding to the pixel of the scene target in the image is greater than 50% of the saturation value θ, there is no need to adjust the imaging parameters. Otherwise, the total integration time should be slightly increased so that the minimum DN value of the processed image is greater than 50% of the saturation value θ. Increasing the total integration time can be achieved by increasing the single integration time or increasing the number of sampling points.

[0091] The noise reduction effect of the present invention is exactly the same as that of the standard ramp sampling algorithm. Considering the effect of using the least squares method to reduce the readout noise, assuming that the noise contained in each sampling value is independent of each other, the standard deviation of the correlated double sampling method is Using this as a unit, we can get the relationship between the slope sampling algorithm and the correlated double sampling method

[0092]

[0093] The sampling method was simulated and analyzed using MATLAB software. The noise reduction effect of the mixed voltage RMS value of 0.05mV white noise is as follows Figure 2 As the number of sampling points increases, the noise RMS value of the output result decreases continuously. When the number of sampling points is 30, the noise RMS value of the present invention is only one third of that of the correlated double sampling.

[0094] From the perspective of storage resources, compared with the standard slope sampling method, the present invention can greatly reduce the storage space requirement. Assuming that the number of data quantization bits is b, the MACC process takes the cumulative sum of M data as a group, and the storage space required for each pixel using the standard slope sampling, the slope sampling after multiple accumulation processing, and the sampling method of the present invention is shown in Table 1 below.

[0095] Table 1 Storage space required for each pixel

[0096]

[0097] When the number of quantization bits b=14 bits and the total number of sampling points is 8 (i.e., n=7), the storage space required for a single pixel of standard slope sampling is 98 bits, while the present invention only requires 41 bits. As the number of sampling points increases, the storage space required for the standard slope sampling algorithm will increase rapidly by multiples, while the storage space required for the present invention will only increase slightly because it is logarithmically related to the number of sampling points. When the number of quantization bits is 14 bits, the total number of sampling points is 88, and the MACC processes each group of 4 pixels, a single pixel of standard slope sampling requires 308 bits, while a single pixel of the present invention only requires 53 bits.

[0098] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solution of the present invention by using the above disclosed methods and technical contents without departing from the spirit and scope of the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall belong to the protection scope of the technical solution of the present invention. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0099] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A sampling method for non-destructive readout of infrared detectors, characterized in that: include: Use standard black body test to non-destructively read out invalid pixels of infrared detectors; The standard blackbody test non-destructively reads the infrared detector response nonlinearity and extracts the linear region starting value of the response curve with nonlinearity ≤ 1% from the response curve. With the end value ; Among them, the data in the response curve does not include the data of invalid pixels; Use non-destructive readout probes to detect target areas of a scene. Integral imaging, obtain +1 image; after removing invalid pixels from the image, use the linear interval range [ , ], process each pixel in the image separately, obtain the pixel full well flag F (p, q) corresponding to each pixel, and obtain the algorithm output value corresponding to each pixel As the preprocessing result of the remote sensor; where p and q are both integers, representing the row and column numbers of the pixel respectively; Use the image receiving system to receive the preprocessing results sent by the remote sensor, and perform image post-processing on the preprocessing results according to the pixel full well flag F (p, q) corresponding to each pixel to obtain the DN value of the processed image , as the image sampling result; If there are two or more data points in the valid data sequence corresponding to the pixel in the pth row and qth column, and the pixel full well flag F(p,q) of the pixel position is 0, then the algorithm output value corresponding to the pixel in the pth row and qth column is as follows: Among them, starting from 0, the m+1 data in the valid data sequence corresponding to the pixel in the pth row and qth column are numbered in sampling order. When i is 0, ; ; is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the p-th row and q-th column; If there are two or more data points in the valid data sequence corresponding to the pixel in the pth row and qth column, and the pixel full well flag F(p,q) of the pixel position is 1, then the algorithm output value corresponding to the pixel in the pth row and qth column is as follows: Among them, starting from 0, the m+1 data in the valid data sequence corresponding to the pixel in the pth row and qth column are numbered in sampling order. When i is 0, ; ; is the DN value corresponding to the data numbered 0 in the valid data sequence corresponding to the pixel in the p-th row and q-th column; If there is only one data point in the valid data sequence corresponding to the pixel in the pth row and qth column or the valid data sequence is empty, and the pixel full well flag F(p,q) at the pixel position is an arbitrary value, then the algorithm output value corresponding to the pixel in the pth row and qth column is is 0.

2. A sampling method for non-destructive readout of infrared detectors according to claim 1, characterized in that: The method for determining the pixel full well flag F(p, q) at the pixel position of the p-th row and the q-th column in the image is as follows: The DN values ​​of all pixels in the pth row and qth column in the n+1 images are taken as a set of data sequences to be screened; If there is a data point in the to-be-screened data sequence of the pixel in the pth row and qth column with a DN value greater than If the pixel is a data point, then the pixel full well flag F(p,q) at the pixel position of the pth row and the qth column is set to 1, otherwise, set F(p,q) to 0.

3. A sampling method for non-destructive readout of infrared detectors according to claim 2, characterized in that: The algorithm output value corresponding to each pixel The method is as follows: Using the linear range [ , ], respectively screen the data in the to-be-screened data sequence of each pixel, and obtain the data belonging to the linear interval range to form the valid data sequence of each pixel; According to the pixel full well flag of each pixel position and the valid data sequence corresponding to each pixel position, the algorithm output value corresponding to each pixel is determined respectively. .

4. A sampling method for non-destructive readout of infrared detectors according to any one of claims 2 to 3, characterized in that: Get the DN value of the processed image The method is as follows: When F(p,q)=1, ; When F(p,q)=0, .

5. A sampling method for non-destructive readout of infrared detectors according to claim 4, characterized in that: Get the DN value of the processed image The last part also includes the pixel full well flag F (p, q) corresponding to each pixel and the DN value of the processed image. , adjust the total integration time of the non-destructive readout infrared detector to improve the quality of subsequent image output.

6. A sampling method for non-destructive readout of infrared detectors according to claim 4, characterized in that: When there is a pixel with a pixel full well flag F(p, q)=1 in the corresponding pixel of the scene target in the image, the method for adjusting the imaging parameters of the non-destructive readout infrared detector is specifically as follows: If there are pixels with a DN value of 0 in the processed image among the corresponding pixels of the scene target, the total integration time is increased; If the DN value of the processed image in the corresponding pixel of the scene target is higher than the saturation value If the pixel is 150% of the total integration time, the total integration time will be shortened; In other cases, there is no need to non-destructively read out the imaging parameters of the infrared detector during the subsequent imaging process.

7. The sampling method for non-destructive readout of infrared detectors according to claim 5, characterized in that: When the pixel full well flags F(p, q) of all positions in the pixels corresponding to the scene target in the image are all 0, the method for adjusting the imaging parameters of the non-destructive readout infrared detector is specifically as follows: If the minimum DN value of the processed image corresponding to the pixel of the scene target in the image is greater than the saturation value If the image quality is less than 50%, there is no need to adjust the imaging parameters; On the contrary, the total integration time should be increased so that the minimum DN value of the processed image is greater than the saturation value. 50% of.

Images