Low-frequency noise blind pixel detection method and system based on segmented noise difference

Abstract

The invention discloses a low-frequency noise blind pixel detection method and system based on segmented noise difference, and the method comprises the steps: enabling all pixels of a detector to observe a uniform and stable black body, so as to obtain the observation data # imgabs0 # f of each pixel, and enabling the observation data # imgabs0 # f to represent a frame number; calculating the noise # imgabs2 # of the observation data # imgabs1 # of each pixel, segmenting the observation data # imgabs3 #, combining the segmented observation data # imgabs3 #, calculating the noise # imgabs4 # of the combined new data, and calculating a first low-frequency stability index LS1i and a second low-frequency stability index LS2i of each pixel based on the noise # imgabs6 # of the observation data # imgabs5 # of each pixel and the noise # imgabs7 # of the combined new data # imgabs5 # of each pixel; and judging the pixels of which the first low-frequency stability index LS1i is lower than a first low-frequency stability index threshold and the second low-frequency stability index LS2i is lower than a second low-frequency stability index threshold as the low-frequency noise blind pixels. And missing detection of low-frequency noise blind pixels can be effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared detector signal processing, and in particular to a low-frequency noise blind pixel detection method and system based on segmented noise difference. Background Art

[0002] With the continuous development of aerospace technology, mobile vehicles such as aircraft pods, drones, and aircraft are increasingly used in modern military reconnaissance, civil mapping, environmental monitoring, and other fields. The infrared cameras carried on these vehicles have become an important means of obtaining target information because they can work normally at night or in severe weather conditions. However, during the imaging process, infrared cameras are prone to blind pixels due to the influence of various factors such as unstable performance of internal components and external environmental interference. Blind pixels, that is, points in the image with abnormal or missing pixel values, will seriously affect the clarity and accuracy of infrared images, reduce the reliability of target recognition, and may even lead to the loss of key information.

[0003] The blind pixel problem in infrared imaging systems manifests itself as a complex non-uniform phenomenon. The causes of blind pixels mainly include problems with the device itself, such as pixel damage, inconsistent response rates, and signal communication problems, such as focal plane readout circuits and communication barriers. The causes of blind pixels are diverse and complex. Regardless of their form, they will affect the imaging quality of infrared images to a certain extent. Therefore, the treatment of blind pixels is particularly important.

[0004] In GB / T 17444-2013, the criterion for blind pixels is that their response rate is less than 1 / 2 of the average response rate, and the criterion for noise pixels is that their noise voltage is greater than 2 times the average noise voltage. Signal value and sensitivity are also often used for judgment in engineering. After removing the blind pixels, the subsequent processing and application can be carried out.

[0005] Low-frequency noise refers to noise whose energy is concentrated in the low-frequency area (less than 10Hz), which often manifests as drift, fluctuation and jump of signal value. The above-mentioned traditional blind pixel detection method has poor accuracy for such low-frequency noise, and cannot timely identify low-frequency noise blind pixels as blind pixels, which further affects the detection of subsequent blind pixels and seriously affects the availability of data. Therefore, a low-frequency noise detection method with high accuracy and low complexity is urgently needed. Summary of the invention

[0006] To solve the above problems, the purpose of the present invention is to provide a low-frequency noise blind pixel detection method and system based on segmented noise difference, which can effectively avoid the missed detection of low-frequency noise blind pixels.

[0007] The technical solution provided by the present invention is: a low-frequency noise blind pixel detection method based on segmented noise difference, comprising the following steps:

[0008] All pixels of the detector observe a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number;

[0009] Calculate the observation data for each pixel Noise

[0010] The observation data After segmentation, combine again and calculate the noise of each new segment of data after combination

[0011] Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ;

[0012] The first low frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels.

[0013] Preferably, the observation data The further combination after segmentation includes:

[0014] The observation data According to the frame number f, it is evenly divided into N segments connected end to end, denoted as Where part is the segment number, ranging from 1 to N, f part It is the frame number of each segment of data;

[0015] Connect each two consecutive data segments end to end to form new data In the formula is the new data after combination, P is the segment number of the new data after combination, ranging from 1 to N-1, f P It is the frame number of each new data segment.

[0016] Preferably, the noise of each new data segment after combination is calculated Further including:

[0017] Calculate each new piece of data Noise

[0018]

[0019] In the formula For data The average value, f P for The frame number, F P for The total number of frames.

[0020] Preferably, the method further comprises discarding the remainder or making the number of frames of each segment of data not completely consistent when F cannot divide N.

[0021] Preferably, based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i Further including:

[0022] Calculate the first low-frequency stability index LS1 for each pixel i :

[0023]

[0024] Calculate the second low-frequency stability index LS2 for each pixel i :

[0025]

[0026] In the formula, represent The minimum value in represent The maximum value in .

[0027] Preferably, the first low-frequency stability index threshold is a preset value or the first low-frequency stability index LS1 of all pixels. i The second low-frequency stability index threshold is a preset value or the second low-frequency stability index LS2 of all pixels. i 0.5 times of.

[0028] Preferably, it is applied to the ground testing phase of the remote sensing satellite optical payload detector, or to its on-orbit operation phase; when the instrument is on-orbit, in the instrument idle phase, observations of the on-board black body for no less than 30 seconds are arranged daily or weekly to update the low-frequency noise blind element list.

[0029] Based on the same concept, the present invention also provides a low-frequency noise blind pixel detection system based on segmented noise difference, comprising:

[0030] The data acquisition module is used to observe all pixels of the detector against a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number;

[0031] Preliminary calculation module, used to calculate the observation data of each pixel Noise

[0032] A reorganization module is used to transform the observed data After segmentation, combine again and calculate the noise of each new segment of data after combination

[0033] Index calculation module, used to calculate the index based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ;

[0034] The blind pixel determination module is used to convert the first low-frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels.

[0035] Based on the same concept, the present invention also provides an electronic device, comprising:

[0036] A memory, the memory being used to store a processing program;

[0037] A processor, wherein when executing the processing program, the processor implements any one of the above-mentioned low-frequency noise blind pixel detection methods based on segmented noise differences.

[0038] Based on the same concept, the present invention also provides a readable storage medium, on which a processing program is stored. When the processing program is executed by a processor, any one of the above-mentioned low-frequency noise blind pixel detection methods based on segmented noise differences is implemented.

[0039] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:

[0040] The technical solution of the present invention also observes the data before using the traditional blind pixel judgment method to perform blind pixel judgment After segmentation, combine again and calculate the noise of each new segment of data after combination Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2i , and then judge the low-frequency noise blind pixel according to the threshold, which can effectively avoid the missed detection of low-frequency noise blind pixel. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The specific embodiments of the present invention are further described in detail below with reference to the accompanying drawings, wherein:

[0042] Figure 1 It is a flow chart of the low-frequency noise blind pixel detection method based on segmented noise difference of the present invention;

[0043] Figure 2 Observation data in the embodiment of the present invention Schematic diagram of segmentation followed by reassembly;

[0044] Figure 3 is the drift characteristic of low-frequency noise;

[0045] Figure 4 is the fluctuation characteristic of low-frequency noise;

[0046] Figure 5 It is the transition characteristic of low frequency noise. DETAILED DESCRIPTION

[0047] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims. It should be noted that the drawings are all in a very simplified form and use non-precise ratios, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0048] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0049] First embodiment

[0050] The detection scheme proposed in the present invention can be applied to the ground test phase of the remote sensing satellite optical payload detector and can also be applied to its on-orbit operation phase. The following takes the ground test phase as an example to explain the low-frequency noise detection method in detail.

[0051] like Figure 1 As shown, a low-frequency noise blind pixel detection method based on segmented noise difference includes the following steps:

[0052] All pixels of the detector observe a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number;

[0053] Calculate the observation data for each pixel Noise

[0054] The observation data After segmentation, combine again and calculate the noise of each new segment of data after combination

[0055] Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ;

[0056] The first low frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels.

[0057] The existing blind pixel detection schemes used in remote sensing satellite infrared optical payload detectors are as follows:

[0058] The detector observes a uniform and stable target black body (temperature T1) for a period of time (more than 2s, more than 1000 frames), and calculates the average value S of each detector pixel signal. i (T1), standard deviation σ i (T1), standard deviation σ i (T1) is the noise i (T1)

[0059] The detector observes a uniform and stable target black body (temperature T2, about 10K different from T1) for a period of time (more than 2s, more than 1000 frames), and calculates the average value S of each detector pixel signal i (T2), standard deviation σ i (T2), standard deviation σ i (T2) is the noise i (T2)

[0060] Based on the above data, the noise equalization temperature difference NEdT under T1 is approximately calculated. i and the response rate R i

[0061]

[0062] R i =S i (T1)-S i (T2)

[0063] According to the above calculation results, the signal value S i (T1) Pixels that do not meet the range of 0.5 to 1.5 times the average signal value of all pixels, response rate R i Pixels and noise that do not meet the range of 0.5 to 1.5 times the average response rate of all pixels i (T1) Pixels with noise equivalent temperature difference NEdT greater than 2 times the average noise of all pixels i Pixels whose (T1) is greater than the index requirement are determined to be blind pixels at temperature T1.

[0064] This judgment method is simple and easy to implement, but it may miss the low-frequency noise blind pixel. Low-frequency noise refers to the noise whose energy is mainly concentrated in the low-frequency area (below 10Hz), which is typically manifested as drift, fluctuation and jump of signal value. The traditional blind pixel detection method has poor accuracy for such low-frequency noise, and cannot identify low-frequency noise blind pixels as blind pixels in time, which further affects the subsequent blind pixel detection and seriously affects the availability of remote sensing data.

[0065] The existing technology directly calculates the noise of each pixel, and in a short observation time (2s), low-frequency noise may not appear. However, even if the observation time is increased, the traditional detection method still cannot accurately identify the detector pixels with low-frequency noise. The reason is that when the detector has low-frequency noise, its overall signal value S i , response rate R i 、noise i Noise correction temperature difference NEdT i It is still possible that the pixel is within the range of normal pixels, which may result in it not being accurately judged as a blind pixel and further affect the blind pixel detection results of other pixels. After segmentation, combine again and calculate the noise of each new segment of data after combination Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i , and then judge the low-frequency noise blind pixel according to the threshold, which can effectively avoid the missed detection of low-frequency noise blind pixel. And the algorithm complexity is low and the inspection is fast.

[0066] Preferably, the observation data The further combination after segmentation includes:

[0067] The observation data According to the frame number f, it is evenly divided into N segments connected end to end, denoted as Where part is the segment number, ranging from 1 to N, f part It is the frame number of each segment of data;

[0068] Connect each two consecutive data segments end to end to form new data In the formula is the new data after combination, P is the segment number of the new data after combination, ranging from 1 to N-1, f P It is the frame number of each new data segment.

[0069] For example, see Figure 2 , showing the observed data Schematic diagram of segmentation and then combination. According to the frame number f, it is evenly divided into N segments connected end to end (N is a positive integer, generally greater than 10), recorded as Where part is the segment number, ranging from 1 to N, f part is the frame number of each segment of data, Divided into There are N segments in total. (When F cannot divide N, the remainder can be discarded, or the number of frames of each segment of data can be made inconsistent.) Connect the two consecutive segments of data end to end to form new data. The specific method is:

[0070]

[0071] In the formula is the new data after combination, P is the segment number of the new data after combination, ranging from 1 to N-1, f P is the frame number of each new data segment, and [a,b] represents connecting data a and b end to end to form new data.

[0072] Preferably, the noise of each new data segment after combination is calculated Further including:

[0073] Calculate each new piece of data Noise

[0074]

[0075] In the formula For data The average value, f P for The frame number, F P for The total number of frames.

[0076] For example, see Figure 3-Figure 5 , which respectively show the drift, fluctuation and jump characteristics of low-frequency noise. Figure 3 and Figure 4 In the example, data 1 is selected and the standard deviation is calculated. The standard deviation of data 1 is much smaller than the standard deviation of all data. Therefore, the value of standard deviation (data 1) / standard deviation (all data) is very small, and the calculated first stability index is very small. Therefore, it can be judged that there is drift or fluctuating low-frequency noise in the data. Figure 5 In the example, we select the area of ​​data 1 and data 2 and calculate their standard deviations. Obviously, the standard deviation of data 1 is much smaller than that of data 2. Therefore, the value of standard deviation (data 1) / standard deviation (data 2) is very small, and the calculated second stability index is very small, which indicates that there is low-frequency noise in the data.

[0077] Preferably, the method further comprises discarding the remainder or making the number of frames of each segment of data not completely consistent when F cannot divide N.

[0078] In some cases, even if F cannot divide N, the consistency and integrity of the data still need to be maintained. In this case, by adjusting the number of frames of each segment of data (i.e., not distributing the data evenly) or discarding the remainder, the uniform distribution of data or the arrangement of data in a specific format can be achieved while ensuring that the overall data volume remains unchanged.

[0079] Preferably, based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i Further including:

[0080] Calculate the first low-frequency stability index LS1 for each pixel i :

[0081]

[0082] Calculate the second low-frequency stability index LS2 for each pixel i :

[0083]

[0084] In the formula, represent The minimum value in represent The maximum value in .

[0085] Calculate the first low-frequency stability index LS1 i and the second low frequency stability index LS2i The higher the values ​​of the two are, the smaller the low-frequency noise of the pixel is and the better the low-frequency stability is.

[0086] Preferably, the first low-frequency stability index threshold is the first low-frequency stability index LS1 of all pixels. i 0.5 times, the second low-frequency stability index threshold is the second low-frequency stability index LS2 of all pixels i 0.5 times of.

[0087] The first low frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels. The first low-frequency stability index threshold and the second low-frequency stability index threshold can be set manually, or they can be defined as 0.5 times the average value of the first low-frequency stability index and the second low-frequency stability index of all pixels.

[0088] Preferably, it is applied to the ground testing phase of the remote sensing satellite optical payload detector, or to its on-orbit operation phase; when the instrument is on-orbit, in the instrument idle phase, observations of the on-board black body for no less than 30 seconds are arranged daily or weekly to update the low-frequency noise blind element list.

[0089] After performing the above method, low-frequency noise blind pixel detection is realized. Then, blind pixel detection is performed according to the traditional method, which can effectively avoid missing the low-frequency noise blind pixel.

[0090] When the instrument is in orbit, the observation time of the onboard black body is about 2 seconds each time, which is not enough for low-frequency noise detection. During the idle phase of the instrument, the observation of the onboard black body for about 30 seconds can be arranged daily or weekly to update the low-frequency noise blind element list.

[0091] Based on the same concept, the present invention also provides a low-frequency noise blind pixel detection system based on segmented noise difference, comprising:

[0092] The data acquisition module is used to observe all pixels of the detector against a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number;

[0093] Preliminary calculation module, used to calculate the observation data of each pixel Noise

[0094] A reorganization module is used to transform the observed data After segmentation, combine again and calculate the noise of each new segment of data after combination

[0095] Index calculation module, used to calculate the index based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ;

[0096] The blind pixel determination module is used to convert the first low-frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixel formula fonts.

[0097] Based on the same concept, the present invention also provides an electronic device, comprising:

[0098] A memory, the memory being used to store a processing program;

[0099] A processor, wherein when executing the processing program, the processor implements any one of the above-mentioned low-frequency noise blind pixel detection methods based on segmented noise differences.

[0100] Based on the same concept, the present invention also provides a readable storage medium, on which a processing program is stored. When the processing program is executed by a processor, any one of the above-mentioned low-frequency noise blind pixel detection methods based on segmented noise differences is implemented.

[0101] If the low-frequency noise blind pixel detection method based on segmented noise difference is implemented in the form of program instructions and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of software, and the computer software is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0102] Those skilled in the art can clearly understand that, for the sake of convenience and brevity in description, the identification content specifically executed by the above-described system and device can refer to the corresponding process in the aforementioned method embodiment.

[0103] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the protection scope of the present invention.

Claims

1. A low-frequency noise blind pixel detection method based on segmented noise difference, characterized in that: The following steps are involved: All pixels of the detector observe a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number; Calculate the observation data for each pixel Noise The observation data After segmentation, combine again and calculate the noise of each new segment of data after combination Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ; The first low frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels.

2. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 1, characterized in that: The observation data The further combination after segmentation includes: The observation data According to the frame number f, it is evenly divided into N segments connected end to end, denoted as Where part is the segment number, ranging from 1 to N, f part It is the frame number of each segment of data; Connect each two consecutive data segments end to end to form new data In the formula is the new data after combination, P is the segment number of the new data after combination, ranging from 1 to N-1, f P It is the frame number of each new data segment.

3. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 2 is characterized in that: Calculate the noise of each new piece of data after combination Further including: Calculate each new piece of data Noise In the formula For data The average value, f P for The frame number, F P for The total number of frames.

4. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 2, characterized in that: The method also includes discarding the remainder or making the number of frames of each segment of data inconsistent when F cannot divide N evenly.

5. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 1, characterized in that: Based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i Further including: Calculate the first low-frequency stability index LS1 for each pixel i : Calculate the second low-frequency stability index LS2 for each pixel i : In the formula, represent The minimum value in represent The maximum value in .

6. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 5, characterized in that: The first low-frequency stability index threshold is a preset value or the first low-frequency stability index LS1 of all pixels. i The second low-frequency stability index threshold is a preset value or the second low-frequency stability index LS2 of all pixels. i 0.5 times of.

7. The low-frequency noise blind pixel detection method based on segmented noise difference according to claim 1, characterized in that: It is applied to the ground test phase of the optical payload detector of the remote sensing satellite, or to its on-orbit operation phase. When the instrument is in orbit, in the idle phase of the instrument, observations of the on-board black body for no less than 30 seconds are arranged every day or every week to update the low-frequency noise blind element list.

8. A low-frequency noise blind pixel detection system based on segmented noise difference, characterized in that: include: The data acquisition module is used to observe all pixels of the detector against a uniform and stable black body to obtain the observation data of each pixel. f represents the frame number; Preliminary calculation module, used to calculate the observation data of each pixel Noise A reorganization module is used to transform the observed data After segmentation, combine again and calculate the noise of each new segment of data after combination Index calculation module, used to calculate the index based on the observation data of each pixel Noise And the noise of each new data after combination Calculate the first low-frequency stability index LS1 for each pixel i and the second low frequency stability index LS2 i ; The blind pixel determination module is used to convert the first low-frequency stability index LS1 i Below the first low frequency stability index threshold and the second low frequency stability index LS2 i Pixels below the second low-frequency stability index threshold are determined as low-frequency noise blind pixels.

9. An electronic device, characterized in that: include: A memory, the memory being used to store a processing program; A processor, wherein when executing the processing program, the processor implements the low-frequency noise blind pixel detection method based on segmented noise difference as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a processing program, and when the processing program is executed by the processor, the low-frequency noise blind pixel detection method based on segmented noise difference as described in any one of claims 1 to 7 is implemented.

Images