A three-dimensional pot lid correction method

Abstract

A three-dimensional pot lid correction method. The basic data used in this method is obtained through the operation and transformation of the pot lid calibration algorithm. The basic data is corrected through three algorithm dimensions: the focal length model, the grayscale model, and the temperature model, and finally high-precision pot lid data suitable for the current physical environment is obtained. The original data is corrected with this pot lid data to obtain a uniform infrared video image.

Description

Technical Field

[0001] The present invention relates to the field of infrared image processing, and particularly to a three-dimensional correction method for infrared images of a pot lid. Background Art

[0002] Due to factors such as design, processing, and assembly of a refrigerated infrared lens, non-uniformity inevitably exists. The most typical non-uniform characteristic is the pot lid effect caused by different transmittances in the central area and the peripheral area.

[0003] For such a pot lid phenomenon, it is generally eliminated by software algorithms. The traditional software method is to align the lens with a uniform surface and collect a frame of image. The gray difference of each pixel shown in this image forms an image matrix, and this matrix can describe the characteristics of the current pot lid. When the image is output, this matrix is applied to the real-time image to eliminate the defect of blackening or brightening in the center caused by the pot lid phenomenon.

[0004] However, the above method has several defects and cannot solve the uniformity problem of a continuously variable zoom lens under all weather conditions. When the focal length changes, the data collected at a single focal length is insufficient to match the entire focal length range. When the temperature changes, the data collected at a single temperature is insufficient to cover the entire temperature range. As a result, although the pot lid phenomenon is weakened, it cannot be completely eliminated, which has a great impact on the recognition rate when observing sky targets for intelligent recognition. Summary of the Invention

[0005] In order to overcome the above-mentioned deficiencies in technology, the present invention provides an effective and reliable pot lid correction method based on multiple groups of calibrations that can adapt to the entire temperature range and large focal length variations.

[0006] The technical solution adopted by the present invention to overcome its technical problems is as follows:

[0007] A three-dimensional pot lid correction method includes the following steps:

[0008] a) Obtain the original video data captured by an infrared detector;

[0009] b) Select M discrete focal length segments for the zoom lens of the infrared detector. The focal length of the x-th focal length segment is f(x), where x ∈ {1, 2,..., M};

[0010] c) Calculate the pot lid data PD f(x) (h, v);

[0011] d) Calculate the gain coefficient k(h, v) and the offset coefficient b(h, v) through the pot lid data PD f(x) (h, v);

[0012] e) The continuous focal length of the zoom lens of the infrared detector is F, and the lid data G of each pixel point at the continuous focal length F is calculated through the focal length correction algorithm f (h, v);

[0013] f) The lid data G after temperature correction is calculated through the temperature correction algorithm fT (h, v);

[0014] g) The lid data G after gray scale correction is calculated through the gray scale correction algorithm fTg (h, v);

[0015] h) Add the lid data G fTg (h, v) after gray scale correction and the original video data captured by the infrared detector to obtain the corrected video data.

[0016] Preferably, in step b), M = 16.

[0017] Furthermore, in step b), each focal length segment equally divides the focal length range of the zoom lens of the infrared detector.

[0018] Furthermore, in step c), through the formula PD f(x) (h, v) = PB f(x) (h, v) - PS f(x) (h, v) to calculate the lid data PD f(x) (h, v) at the focal length f(x), where PB f(x) (h, v) is a frame of image data collected by the zoom lens of the infrared detector aiming at the black body window at room temperature T2, and PS f(x) (h, v) is a frame of image data collected by the zoom lens of the infrared detector with the shutter closed at the shutter temperature T1. h is the number of horizontal pixels of the image, and v is the number of vertical pixels of the image.

[0019] Preferably, the black body temperature is set to 25 °C, the value range of h is 0 - 639, and the value range of v is 0 - 511.

[0020] Furthermore, step d) includes the following steps:

[0021] d - 1) Divide the M discrete focal lengths and their corresponding lid data PD f(x) (h, v) into groups. The data of the mth group is (f(m), PD f(m) (h, v)), where f(m) is the focal length of the mth focal length segment, and PD f(m) (h, v) is the lid data at the focal length f(m). The data of the (M + 1 - m)th group is (f(M + 1 - m), PD f(M+1-m)(h, v)), where f(M + 1 - m) is the focal length of the (M + 1 - m)-th focal length segment, and PD f(M+1-m) (h, v) is the data of the pot lid under the focal length f(M + 1 - m);

[0022] d - 2) is calculated through the formula to obtain the gain coefficient k(h, v) of each group m ;

[0023] d - 3) is calculated through the formula

[0024] to obtain the offset coefficient b(h, v) of each group m ;

[0025] d - 4) is calculated through the formula to obtain the gain coefficient k(h, v);

[0026] d - 5) is calculated through the formula to obtain the offset coefficient b(h, v).

[0027] Furthermore, in step e), the data G f (h, v) of each pixel point of the pot lid under the continuous focal length F is calculated through the formula G f (h, v) = k(h, v)F + b(h, v).

[0028] Furthermore, in step f), the data G fT (h, v) of the pot lid after temperature correction is calculated through the formula G shutter (h, v) = (T env - T f ) / (T1 - T2) * G T (h, v) * L fT , where T shutter is the shutter temperature, T env is the ambient temperature, L T is a typical value related to the type of the lens, and L T ranges from 0.7 to 1.0.

[0029] Preferably, L T = 0.987.

[0030] Furthermore, step g) includes the following steps:

[0031] g - 1) Establish a gray - level distribution composed of N rings. The width of each ring is equal, and the N rings are concentrically nested with each other from the inside out. The N rings form a circular area with a radius of R. The gray - level values of the rings increase sequentially from the inside out. N is a positive integer between 3 and 10;

[0032] g - 2) The average gray value of the nth ring is Gray(r n ), where n is a positive integer between 0 and N - 1, and r n is the radius range of the nth ring.

[0033] g - 3) Through the formula S n = Gray(r n ) / Gray e the ratio S of the average gray value of the nth ring to the average value of the full screen is calculated. n In the formula, Gray e is the gray value of the full screen.

[0034] g - 4) Through the formula G fTg (h, v) = G fT (h, v) / S n the lid data G fTg (h, v) after gray correction is calculated.

[0035] The beneficial effect of the present invention is that the basic data used for the lid correction of the present invention is obtained through the operation and transformation of the lid calibration algorithm. The basic data is corrected through three algorithm dimensions of the focal length model, the gray model, and the temperature model, and finally, high - precision lid data suitable for the current physical environment is obtained. Using this lid data to correct the original data, a uniform infrared video image is obtained. Brief Description of the Drawings

[0036] Figure 1 is the flowchart of the method of the present invention;

[0037] Figure 2 is the schematic diagram of the gray - scale annular diagram of the present invention. Detailed Embodiment

[0038] The following further describes the present invention with reference to the attached Figure 1 , attached Figure 2 .

[0039] The present invention studies three aspects of temperature, focal length, and single - frame data, and designs an effective and reliable lid correction method based on multiple sets of calibrations that can adapt to the full temperature range and large focal - length changes.

[0040] The core content of the present invention is the selection of calibration points, the design of temperature - model coefficients, and the design of focal - length - model coefficients. Specifically, as shown in the attached Figure 1 , a three - dimensional lid correction method includes the following steps:

[0041] a) Obtain the original video data captured by the infrared detector.

[0042] b) Select M discrete focal length segments for the zoom lens of the infrared detector. The focal length of the x-th focal length segment is f(x), where x ∈ {1, 2,..., M}. For the zoom lens, M must be an even number from 9 to 32. In a specific embodiment of the present invention, in step b), M = 16, that is, there are 16 focal length segments, and each focal length segment equally divides the focal length range of the zoom lens of the infrared detector, and the focal lengths are f(1), f(2),..., f(M - 1), f(M).

[0043] c) Calculate the pot lid data PD f(x) (h, v).

[0044] d) Calculate the gain coefficient k(h, v) and the offset coefficient b(h, v) through the pot lid data PD f(x) (h, v).

[0045] e) The continuous focal length of the zoom lens of the infrared detector is F. Calculate the pot lid data G f (h, v) of each pixel point at the continuous focal length F through the focal length correction algorithm.

[0046] f) Calculate the pot lid data G fT (h, v) after temperature correction through the temperature correction algorithm.

[0047] g) Calculate the pot lid data G fTg (h, v) after gray scale correction through the gray scale correction algorithm. The pot lid data G fTg (h, v) after gray scale correction is the real-time pot lid data after three-dimensional correction, that is, the data finally acting on the real-time image correction.

[0048] h) The key links of the invention are the pot lid calibration algorithm process and the three-dimensional correction process. Among them, the three-dimensional pot lid correction process includes three aspects: the focal length model, the gray scale model, and the temperature model. We use the pipeline method to connect the above processes in series to obtain the real-time image. Specifically, add the pot lid data G fTg (h, v) after gray scale correction and the original video data captured by the infrared detector to obtain the corrected video data. Experiments prove that this method effectively improves the accuracy of pot lid correction and achieves the purpose of no overcorrection and good correction.

[0049] The basic data used in the pot lid correction of the present invention is obtained through the operation and transformation of the pot lid calibration algorithm. The basic data is corrected through three algorithm dimensions of the focal length model, the gray scale model, and the temperature model, and finally high-precision pot lid data suitable for the current physical environment is obtained. Using this pot lid data to correct the original data, a uniform infrared video image can be obtained.

[0050] Example 1:

[0051] In step c), the pot cover data PD f(x) (h, v) is calculated by the formula PD f(x) (h, v) = PB f(x) (h, v) - PS f(x) (h, v), where PB f(x) (h, v) is a frame of image data collected when the zoom lens of the infrared detector is aligned with the blackbody window at room temperature T2, and PS f(x) (h, v) is a frame of image data collected when the shutter of the zoom lens of the infrared detector is closed at shutter temperature T1. h is the number of horizontal pixels of the image, and v is the number of vertical pixels of the image. In a specific embodiment of the present invention, the blackbody temperature is set to 25 °C, the value range of h is 0 - 639, and the value range of v is 0 - 511. The pot cover data PD f(x) (h, v) is the pot cover data of each pixel point at the discrete focal length f(x). Different focal lengths f(x) result in different pot cover data. In practical applications, the lens may stay at any focal length segment containing M discrete data. Therefore, next, we need to obtain the pot cover data of the entire focal length segment through the data of M discrete focal length segments.

[0052] Embodiment 2:

[0053] Next, curve fitting is to be performed on each pixel point. The purpose is to analyze and process the pot cover data PD f(x) (h, v) at M discrete focal length segments f(x) obtained in step c) to obtain the corresponding relationship between the pot cover data and the focal length. In this way, we can obtain the pot cover data at any focal length segment according to this corresponding relationship.

[0054] We use F to represent the continuous focal length, and PD F (h, v) to represent the pot cover data at the continuous focal length. Experiments and data prove that the relationship between PD F (h, v) and F conforms to a linear equation, which can be expressed as: PD F (h, v) = k(h, v)F + b(h, v), where k(h, v) is the gain coefficient and b(h, v) is the offset coefficient. These two sets of data are the key data to be solved in this step, and the solution process is the curve fitting process. Specifically, step d) includes the following steps:

[0055] d - 1) Divide the M discrete focal lengths and their corresponding pot cover data PD f(x) (h, v) into groups, each group contains two pairs of discrete focal lengths and the corresponding pot cover data. The data of the mth group is (f(m), PD f(m) (h, v)), where f(m) is the focal length of the m-th focal length segment, PD f(m) (h, v) is the data of the pot lid under the focal length f(m). The data of the (M + 1 - m)-th group is (f(M + 1 - m), PD f(M+1-m) (h, v)), where f(M + 1 - m) is the focal length of the (M + 1 - m)-th focal length segment, PD f(M+1-m) (h, v) is the data of the pot lid under the focal length f(M + 1 - m).

[0056] d - 2) is calculated through the formula to obtain the gain coefficient k(h, v) of each group m .

[0057] d - 3) is calculated through the formula

[0058] to obtain the offset coefficient b(h, v) of each group m .

[0059] d - 4) is calculated through the formula to obtain the gain coefficient k(h, v).

[0060] d - 5) is calculated through the formula to obtain the offset coefficient b(h, v). Further, the gain coefficient k(h, v) can be organized according to the pixel coordinates to form the matrix K(h, v), and the offset coefficient b(h, v) can be organized according to the pixel coordinates to form the matrix B(h, v). The purpose of forming the matrix K(h, v) and the matrix B(h, v) is to facilitate storage in the embedded system.

[0061] Example 3:

[0062] In step e), the data G f (h, v) of each pixel point of the pot lid under the continuous focal length F is calculated through the formula G f (h, v) = k(h, v)F + b(h, v). The purpose of the focal length correction algorithm is to calculate the data of the pot lid of the lens under a specific focal length.

[0063] Example 4:

[0064] The temperature difference T env between the ambient temperature T of the device and the shutter temperature T shutter affects the pot lid correction effect. After a large amount of data analysis and experiments, the final data of the pot lid and the temperature correction coefficient L d have a linear relationship with the ambient temperature and the shutter temperature difference. Specifically, in step f), through the formula G T (h, v) = (T fT (h, v) = (T shutter - Tenv ) / (T1 - T2) * G f (h, v) * L T Calculate the lid data G after temperature correction fT (h, v), where T shutter is the shutter temperature, T env is the ambient temperature, L T is a typical value related to the type of the lens, L T The value range is 0.7 - 1.0. In a specific embodiment of the present invention, L T = 0.987.

[0065] Example 5:

[0066] Step g) includes the following steps:

[0067] g - 1) As shown in the appendix Figure 2 , establish a gray - level distribution composed of N rings. The width of each ring is equal, and the N rings are concentrically nested with each other from the inside out. The N rings form a circular area with a radius of R. The gray - level values of the rings increase sequentially from the inside out. The value of N is a positive integer between 3 and 10. In a specific embodiment of the present invention, the value of N is 5. The gray - level distribution is correlated with the lid data. When the global gray - level difference is large, the lid data can be weakened, and when the global gray - level difference is small, the lid data is appropriately enhanced. According to the characteristics of the lid data, the strength of the lid is distributed in a circular pattern. Therefore, for the gray - level distribution dimension, the adjustment of the lid data adopts a circular statistical method.

[0068] g - 2) The average gray - level value of the nth ring is Gray(r n ), where n takes positive integer values between 0 and N - 1, and r n is the radius range of the nth ring,

[0069] g - 3) Calculate the ratio S n of the average gray - level value of the nth ring to the full - screen average value through the formula S n = Gray(r e ), where Gray n is the full - screen gray - level value. e

[0070] g - 4) Calculate the lid data G fTg (h, v) after gray - level correction through the formula G fT (h, v) = G n (h, v) / S fTg (h, v).

[0071] ​Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A three-dimensional pot lid correction method, characterized in that, It includes the following steps: a) Obtain the original video data captured by the infrared detector; b) Select M discrete focal length segments for the zoom lens of the infrared detector. The focal length of the x-th focal length segment is f(x), where x ∈ {1, 2, …, M}; c) Calculate the pot lid data PD at the focal length f(x) f(x) (h, v); d) Through the lid data PD f(x) (h, v) Calculate the gain coefficient k(h, v) and the offset coefficient b(h, v); e) The continuous focal length of the zoom lens of the infrared detector is F, and the lid data G of each pixel point at the continuous focal length F is calculated through the focal length correction algorithm f (h, v); f) Calculate the lid data G after temperature correction through the temperature correction algorithm fT (h,v); g) The lid data G after gray correction is calculated by the gray correction algorithm fTg (h, v); (h) Add the lid data G after grayscale correction fTg to the original video data captured by the (h, v) and infrared detectors to obtain the corrected video data; Step d) includes the following steps: (d-1) Divide the M discrete focal lengths and their corresponding dish data PD f(x) (h, v) into groups. The data of the m-th group is (f(m), PD f(m) (h, v)), where f(m) is the focal length of the m-th focal length segment, and PD f(m) (h, v) is the dish data at the focal length f(m). The data of the (M + 1 - m)-th group is (f(M + 1 - m), PD f(M+1-m) (h, v)), where f(M + 1 - m) is the focal length of the (M + 1 - m)-th focal length segment, and PD f(M+1-m) (h, v) is the dish data at the focal length f(M + 1 - m); d-2) The gain coefficient k(h, v) of is calculated by the formula for m groups; d-3) Through the formula Calculated offset coefficients b(h, v) for m ; d-4) The gain coefficient k(h, v) is calculated through the formula ; d-5) The offset coefficient b(h, v) is calculated by the formula ​ 2. The three-dimensional pot lid correction method according to claim 1, wherein: In step b), M = 16.

3. The three-dimensional pot lid correction method according to claim 1, wherein: In step b), each focal length segment equally divides the focal length range of the zoom lens of the infrared detector.

4. The three-dimensional pot lid correction method according to claim 1, wherein: In step c), the formula PD f(x) (h,v)=PB f(x) (h,v)-PS f(x) (h,v) calculates the pot cover data PD at focal length f(x) f(x) (h,v), where PB f(x) (h, v) is a frame of image data collected by the infrared detector’s zoom lens aimed at the black body window at room temperature T2. f(x) (h, v) is a frame of image data collected by the infrared detector's zoom lens with the shutter closed at a shutter temperature of T1, h is the number of horizontal pixels in the image, and v is the number of vertical pixels in the image.

5. The three-dimensional pot lid calibration method according to claim 4, wherein: The blackbody temperature is set to 25 °C, the value range of h is 0 - 639, and the value range of v is 0 - 511.

6. The three-dimensional pot lid correction method according to claim 1, wherein: In step e), the lid data G of each pixel point under the continuous focal length F is calculated through the formula G f (h, v) = k(h, v)F + b(h, v). f (h, v).

7. The three-dimensional pot lid correction method according to claim 4, wherein: In step f), the data of the pot lid after temperature correction, G fT (h, v), is calculated through the formula G shutter (h, v) = (T env - T f )(h, v) / (T1 - T2) * G T (h, v) * L fT , where T shutter is the shutter temperature, T env is the ambient temperature, and L T is a typical value related to the type of the lens. The value range of L T is 0.7 - 1.

0.

8. The three-dimensional pot lid correction method according to claim 7, characterized in that: L T ＝0.987。 9. The three-dimensional pot lid correction method according to claim 1, characterized in that Step g) includes the following steps: g-1) Establish a gray-scale distribution composed of N rings. The ring width of each ring is equal. The N rings are concentrically nested with each other from the inside out. The N rings form a circular area with a radius of R. The gray-scale values of the rings increase sequentially from the inside out. The value of N is a positive integer between 3 and 10; g-2) The average gray value of the nth ring is Gray(r n ), where n is a positive integer between 0 and N-1, and r n is the radius range of the nth ring, g-3) Calculate the ratio S of the average gray value of the nth ring and the average value of the full screen through the formula S n = Gray(r n ) / Gray e , where S is the ratio of the average gray value of the nth ring and the average value of the full screen; n , in the formula Gray e is the gray value of the full screen; g-4) Calculate the lid data G after gray correction through the formula G fTg (h, v) = G fT (h, v) / S n where G fTg (h, v) is obtained.

Images