A method of short-wave infrared movement core assembly non-uniformity correction and computer product

Abstract

This invention discloses a method and computer product for non-uniformity correction of short-wave infrared core components, belonging to the field of infrared image non-uniformity correction technology. The method includes two stages: offline calibration and online correction. During offline calibration, with a fixed gain and uniform background, continuous images at different integration times are acquired at multiple preset temperature points. After time-domain averaging, slope and intercept plots showing the linear change of correction parameters with integration time are established pixel by pixel. Then, polynomial fitting is performed on the slope and intercept of each pixel with temperature, and the fitting coefficients are stored. During online correction, the current core temperature and integration time are obtained, and the correction parameters are calculated in real time using the stored fitting coefficients, and single-point non-uniformity correction is performed. This invention achieves adaptive non-uniformity correction of short-wave infrared core components under continuous operating temperature and continuous integration time, significantly reducing the storage space of correction parameters while ensuring high correction accuracy, and is easy to implement in real time on embedded platforms.

Description

Technical Field

[0001] This invention belongs to the field of infrared image non-uniformity correction technology, and relates to a method for non-uniformity correction of short-wave infrared core components and a computer product, particularly to a method for non-uniformity correction of short-wave infrared core components with adaptive integration time and temperature and a computer product. Background Technology

[0002] As is well known, non-uniformity refers to the fact that, under ideal conditions, the output response of each effective detection unit on the focal plane of the short-wave infrared core component should theoretically be consistent under uniform illumination. However, due to the influence of the manufacturing process of the short-wave infrared focal plane detector and the external environment, the output response of each effective detection unit is not completely consistent, which leads to the non-uniformity of the short-wave infrared image.

[0003] Non-uniformity correction is crucial for shortwave infrared images because it can eliminate brightness differences caused by uneven sensor response and illumination variations, thereby improving image quality. This process not only enhances the distinguishability of object surface features but also ensures the accuracy and comparability of data, providing a more reliable foundation for subsequent image analysis and processing.

[0004] Current non-uniformity correction techniques fall into two categories: calibration-based methods and two-point correction algorithms. Two-point correction is a simple and easy-to-implement calibration-based algorithm that shows good results in correcting gain and bias non-uniformity in area-array staring short-wave infrared cameras. Two-point correction achieves non-uniformity correction with fixed parameters by acquiring image data from high-temperature and low-temperature reference sources and calculating the gain and offset of each pixel. However, the two-point correction method relies on only two fixed temperature points and cannot adapt to dynamic operating conditions such as changes in the temperature of the infrared camera module, adjustments to integration time, and device aging. The correction accuracy decreases significantly with scene changes; storing multiple sets of temperature-corresponding parameters also consumes a large amount of storage space.

[0005] Another type of non-uniformity correction technique is scene-based correction methods. These methods do not require a reference source and dynamically update correction parameters by analyzing scene information in infrared images. However, scene-based correction methods are highly dependent on the scene. When the image is static, has a simple texture, or the target occupies too large a proportion, parameter estimation is prone to distortion, leading to correction failure. They also have high algorithm complexity, poor real-time performance, and some methods require storing historical frame data, occupying additional storage space.

[0006] Because shortwave detectors involve numerous operating states, including combinations of different gains, temperatures, and integration times, traditional calibration algorithms struggle to balance calibration accuracy with storage space requirements when performing non-uniform calibration. Storing a separate set of calibration parameters for each discrete temperature point and each discrete integration time results in linearly increasing storage space with the number of state combinations, leading to high costs. Interpolation methods, on the other hand, suffer from limited accuracy. Therefore, it is necessary to improve traditional non-uniform calibration methods for shortwave infrared core components, providing a method that enables adaptive calibration at continuous temperatures and continuous integration times with extremely low storage overhead. Summary of the Invention

[0007] To address the problems existing in the background technology, the present invention provides a method for non-uniformity correction of short-wave infrared core components that is adaptive to integration time and temperature.

[0008] The technical solution of this invention is:

[0009] A method for correcting non-uniformity of shortwave infrared core components includes an offline calibration stage and an online calibration stage.

[0010] Offline calibration stage:

[0011] Step 1: Fix the gain mode of the shortwave infrared sensor assembly and the uniform background brightness of the integrating sphere. At M different preset sensor temperature points, acquire multiple consecutive images at N different integration times; where M≥4 and N≥2. Preferably, M is 4 or more, covering the expected operating temperature range of the sensor (e.g., 0℃, 20℃, 40℃, 60℃); N is 2 or more, and the integration time range covers the range supported by the sensor (e.g., 0-10ms, selecting 1ms, 2ms, ..., 10ms). When using the integrating sphere as the uniform background, the brightness of the integrating sphere needs to be controlled at a fixed value so that the images at each integration time do not overflow at this brightness.

[0012] Step 2: For each integration time at each preset temperature point, perform time-domain averaging on multiple consecutive images to obtain a standard image for that integration time at that temperature point, in order to eliminate random noise.

[0013] Step 3: For each preset temperature point, using the standard image corresponding to different integration times at that temperature point, establish a pixel-by-pixel linear relationship model of the correction parameter B as a function of integration time t: .in, This is a slope graph. The intercept plot is shown. To obtain more robust results, it is preferable to use the least squares method to linearly fit the standard image pixel values ​​at N≥2 integration time points to obtain the slope plot C1(T) and the intercept plot C2(T).

[0014] Step 4: Obtain the slope value for each pixel at M different temperature points. By performing a polynomial fit with the corresponding temperature value, the fitting coefficient of the slope of the pixel as a function of temperature is obtained; similarly, the intercept values ​​of each pixel at M different temperature points are used to obtain the slope of the pixel as a function of temperature. A polynomial fit is performed with the corresponding temperature value to obtain the fitting coefficients of the pixel intercept as a function of temperature. This invention preferably uses a cubic polynomial fit, and the fitting formula is as follows:

[0015]

[0016]

[0017] Where (a3,a2,a1,a0) and (b3,b2,b1,b0) are the slope fitting coefficient and intercept fitting coefficient for each pixel, respectively. The fitting coefficients of all pixels are stored as correction parameters. The storage size of these parameters is independent of the number of preset temperature points M, and only depends on the total number of pixels and the polynomial order, greatly compressing storage space.

[0018] Online calibration phase:

[0019] Step 5: During actual operation, read the temperature sensor value of the short-wave infrared sensor in real time to obtain the current operating temperature. And obtain the currently set points time. .

[0020] Step 6: For each pixel, calculate the current temperature based on its stored slope fitting coefficient and intercept fitting coefficient. The slope value below and intercept value :

[0021]

[0022]

[0023] Then, the correction parameters under the current operating state are calculated:

[0024]

[0025] Step 7: Use the correction parameters obtained in Step 6 Single-point non-uniformity correction is performed on the currently acquired infrared image. The correction formula is as follows:

[0026]

[0027] in, Let (i,j) be the output value of pixel (i,j) before correction at time n. The corrected output value. These are the correction parameters for that pixel. This is the average value of all pixels in the current correction parameter map.

[0028] Furthermore, for different gain modes (low gain, medium gain, high gain) of the shortwave infrared core component, the above offline calibration stage is repeated to generate and store the fitting coefficients for each gain mode. During online calibration, the corresponding fitting coefficients are selected and calculated based on the current gain mode.

[0029] A computer product comprising a non-transient readable storage medium and a computer program, the computer program being tangibly stored on the non-transient readable storage medium, the computer program implementing the steps of the short-wave infrared core component non-uniformity correction method of the present invention when executed by a processor within an FPGA, DSP, ARM, or other computer.

[0030] The beneficial effects of this invention are:

[0031] (1) Adaptive non-uniformity correction of shortwave infrared core under continuous temperature range and continuous integration time was realized, which solved the defect that the traditional calibration method can only work at discrete state points.

[0032] (2) Significantly reduces the storage space for calibration parameters. Traditional methods require storing M_t × M_i sets of parameters to accurately calibrate M_t temperature points and M_i integration time points. This invention only requires storing 8 polynomial coefficients for each pixel (for cubic polynomials), and the storage amount is independent of the number of preset calibration points, making it particularly suitable for resource-constrained embedded systems.

[0033] (3) High correction accuracy. The nonlinear change of the real physical response with temperature is approximated by polynomial fitting, and multi-integral time point fitting is used in the linear part, so the overall accuracy is better than the two-point or multi-point discrete interpolation method.

[0034] (4) The algorithm has low computational complexity and is easy to implement in hardware. The core operations are polynomial evaluation and linear operations, which are easy to implement in real time on platforms such as FPGA and DSP. Attached Figure Description

[0035] Figure 1 This is a general flowchart of the method of the present invention.

[0036] Figure 2 This is a flowchart illustrating the offline calibration stage of the present invention.

[0037] Figure 3 This is a flowchart illustrating the online calibration phase of the present invention.

[0038] Figure 4(a), (b), and (c) are the images before correction (a), the two-point corrected image (b), and the non-uniform correction image (c) of the present invention, respectively. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to embodiments. Those skilled in the art will understand that the following embodiments are for illustrative purposes only and should not be considered as limiting the scope of the invention. Where specific techniques or conditions are not specified in the embodiments, they are performed according to the techniques or conditions described in the literature in the art or according to the product manual. Where the manufacturers of the materials or equipment used are not specified, they are all conventional products that can be obtained by purchase.

[0040] This embodiment takes a certain type of shortwave infrared core component with a resolution of 640×512 as an example. Its operating temperature range is 0℃~60℃, the integration time range is 0~10ms, and the gain mode includes three levels: high, medium, and low.

[0041] The following detailed explanation uses the high-gain mode as an example.

[0042] See Figures 1-3 As shown, a method for non-uniformity correction of short-wave infrared core components with adaptive integration time and temperature includes:

[0043] I. Offline Calibration Stage

[0044] 1. Set the integrating sphere to a uniform light source and adjust its brightness so that the detector response does not saturate at the maximum integration time of 10ms.

[0045] 2. Select four preset temperature points: T1=0℃, T2=20℃, T3=40℃, T4=60℃, and use TEC to precisely control the movement temperature to stabilize at the preset value.

[0046] 3. At each temperature point, the integration time t is set sequentially to 1ms, 2ms, 3ms, …, 10ms (a total of 10 integration times). 100 frames of images are continuously acquired at each integration time.

[0047] 4. Perform time-domain averaging on 100 frames of images at each temperature point and at each integration time to obtain 4×10=40 standard images.

[0048] 5. For each temperature point (k=1..4), using the standard images corresponding to 10 integration times at this temperature point, perform least-squares linear fitting pixel by pixel to obtain the slope map at this temperature point. and intercept plot The fitting formula is: .

[0049] 6. For each pixel (i,j), calculate the slope values ​​at the four temperature points. The coefficients are obtained by performing a cubic polynomial fit with the corresponding temperature. Similarly, a cubic polynomial fit is performed on the intercept values ​​to obtain the coefficients. The eight coefficients of all pixels are stored as a calibration parameter file. Each coefficient image is 640 × 512 × 16 bits = 5,242,880 bits in size. Based on 1 Mb = 1,048,576 bits, this is 5.0 Mb (megabits), equivalent to 5.0 ÷ 8 = 0.625 MB (megabytes). This invention requires storing eight coefficient images, for a total storage capacity of 8 × 5.0 Mb = 40 Mb, or 5.0 MB. In comparison, the traditional two-point calibration method, if it needs to cover 10 different movement temperature points and 10 different integration times, requires storing 10 × 10 = 100 sets of calibration parameters. Each set of calibration parameters typically contains two images (slope image K and intercept image B), requiring a total of 200 images. Each image is also stored with 16-bit precision, with a size of 5.0 Mb. Therefore, the total storage capacity required by the traditional method is 200 × 5.0 Mb = 1000 Mb, or 125 MB. The present invention only requires 5.0 MB of storage, which is about 4% of the storage capacity of the traditional method (5.0 / 125 = 0.04), saving 96% of the storage space. Moreover, the present invention can achieve adaptive correction under continuous temperature and continuous integration time, while the traditional method can only work at discrete state points.

[0050] II. Online calibration phase

[0051] 1. After the mechanism is powered on, read the current temperature sensor value. (e.g., 30℃) and current integration time setting (e.g., 8ms), and obtain the current gain mode (high gain).

[0052] 2. For each pixel (i,j), calculate based on the stored coefficients:

[0053]

[0054]

[0055]

[0056] 3. For the acquired raw images (such as...) Figure 4 As shown in (a), there is obvious non-uniform noise. Apply the correction formula:

[0057]

[0058] in For all pixels in the current frame The mean.

[0059] 4. To verify the actual correction effect of the present invention, short-wave infrared images of real scenes were acquired in a working mode with high gain, core temperature of 30°C, and integration time of 8ms. The images were then processed using the traditional two-point correction method and the non-uniformity correction method of the present invention, and compared from both objective indicators and subjective visual perspectives.

[0060] For objective evaluation, standard deviation (STD) and non-uniformity (NU) were used as metrics. The original image had an STD of 693.49 and a NU of 9.50%. After processing with the two-point correction method, the STD decreased to 294.89 and the NU decreased to 3.84%. After processing with the method of this invention, the STD further decreased to 113.94 and the NU decreased to 1.46%. Compared with the original image, the method of this invention reduced the STD by approximately 83.57% and the NU by 84.63%. Compared with the two-point correction method, the method of this invention further reduced the STD by 61.36% and the NU by 61.98%. These data show that the method of this invention is significantly superior to the traditional two-point correction method in terms of correction accuracy.

[0061] In terms of subjective visual aspects, such as Figure 4 As shown (where (a) is the original image, (b) is the image after two-point correction, and (c) is the image after correction using the method of this invention), the original image exhibits significant non-uniform noise, with details almost completely obscured by the noise; the image after two-point correction shows some improvement in uniformity, but the correction effect is limited; while using the method of this invention, the image becomes clean and clear, the non-uniform noise is significantly reduced, and the detailed information is fully presented. In summary, the method of this invention can adaptively generate correction parameters based on the actual working temperature and integration time, effectively solving the problem of decreased correction accuracy caused by deviations from calibration conditions in traditional calibration methods.

[0062] III. Processing of Other Gain Modes

[0063] For medium-gain and low-gain modes, repeat the offline calibration process described above to generate their respective independent coefficient sets. In actual use, simply load the corresponding coefficient set based on the current gain mode.

[0064] IV. Hardware Implementation

[0065] The computer program product of this invention can be embedded in an FPGA or DSP. In the FPGA implementation, multiple hardware multipliers and adders can be used to calculate the polynomial value of each pixel in parallel, and the pipeline design can achieve real-time processing requirements (e.g., 100Hz frame rate).

[0066] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for correcting non-uniformity of a shortwave infrared core assembly, characterized in that, It includes an offline calibration phase and an online calibration phase; The offline calibration phase includes: Step 1: Fix the gain mode of the shortwave infrared core component and the uniform background brightness of the integrating sphere, and acquire multiple continuous images at N different integration times at M different preset core temperature points; Step 2: For each integration time at each preset temperature point, perform time-domain averaging on multiple consecutive images to obtain a standard image for that integration time at that temperature point; Step 3: For each preset temperature point, using the standard image corresponding to different integration times at that temperature point, establish a pixel-by-pixel linear relationship model of the change of correction parameter B with integration time t: ,in This is a slope graph at that temperature point. This is the intercept plot at that temperature point; Step 4: Combine the slope values ​​obtained for each pixel at M different temperature points with... The corresponding temperature values ​​are subjected to polynomial fitting to obtain the fitting coefficients of the slope of the pixel as a function of temperature; similarly, the intercept values ​​of each pixel at M different temperature points are used to obtain the fitting coefficients. Perform a polynomial fit with the corresponding temperature value to obtain the fitting coefficient of the pixel intercept as a function of temperature; store the fitting coefficients of all pixels as correction parameters. The online calibration phase includes: Step 5: Obtain the actual operating temperature of the current shortwave infrared sensor. and points time ; Step 6: For each pixel, calculate the current temperature based on its stored slope fitting coefficient and intercept fitting coefficient. The slope value below and intercept value Then, the correction parameters under the current working state are calculated. ; Step 7: Use the correction parameters obtained in Step 6 Perform single-point non-uniformity correction on the currently acquired infrared image and output the corrected image.

2. The positive method according to claim 1, characterized in that: The preset core temperature point M in step 1 is greater than or equal to 4 and covers the expected operating temperature range of the short-wave infrared core component; the number of different integration times N is greater than or equal to 2 and the integration time range covers the integration time interval supported by the core component.

3. The method according to claim 1, characterized in that: In step 3, when establishing the linear relationship model of the correction parameter B with the integration time t, the least squares method is used to linearly fit the standard image pixel values ​​at N≥2 integration time points to obtain the slope map C1(T) and the intercept map C2(T).

4. The method according to claim 1, characterized in that: In step 4, the polynomial fitting uses a cubic polynomial, and the fitting formula is as follows: Where (a3,a2,a1,a0) and (b3,b2,b1,b0) are the slope fitting coefficient and intercept fitting coefficient for each pixel, respectively.

5. The method according to claim 1, characterized in that: In step 1, for the low-gain, medium-gain, and high-gain gain modes of the short-wave infrared core component, steps 1 to 4 are repeated to generate and store the fitting coefficients for each gain mode; during online calibration, the corresponding fitting coefficients are selected and calculated according to the current gain mode.

6. The method according to claim 1, characterized in that: The formula for single-point non-uniformity correction in step 7 is as follows: in, Let (i,j) be the output value of pixel (i,j) before correction at time n. The corrected output value. The correction parameters for this pixel are calculated in step 6. This is the average value of all pixels in the current correction parameter map.

7. A computer product comprising a non-transient readable storage medium storing a computer program that, when executed by an FPGA, DSP, or ARM processor, implements the steps of the non-uniformity correction method for short-wave infrared core components according to any one of claims 1-6.

Images