Gain compensation method of focal plane movement in wide-temperature environment

Abstract

The invention relates to a gain compensation method for a focal plane movement in a wide-temperature environment. The method comprises the following steps: dividing non-uniform temperature subintervals in a wide-temperature range; fitting in each temperature subinterval by adopting a Chebyshev polynomial, establishing a gain prediction model of each temperature subinterval, and determining a gain compensation coefficient calculation relation of each temperature subinterval based on the gain prediction model; acquiring the working temperature of the focal plane array in real time; determining a temperature subinterval to which the working temperature belongs according to the working temperature collected in real time, calling a gain compensation coefficient calculation relation corresponding to the temperature subinterval, and calculating a real-time gain compensation coefficient in combination with a lookup table and linear interpolation; and applying the real-time gain compensation coefficient to an original output signal of the focal plane movement to realize gain compensation. According to the invention, image artifacts and contrast reduction caused by wide temperature change are effectively suppressed, the image uniformity is improved, the temperature measurement precision is improved, and the working stability and output reliability of a focal plane movement in a severe temperature environment are improved.

Description

Technical Field

[0001] This invention relates to the field of infrared semiconductor detectors, and more specifically, to a gain compensation method for a focal plane array mechanism under wide temperature conditions. Background Technology

[0002] In the field of infrared imaging technology, the Infrared Focal Plane Array (IRFPA) serves as the core detector of an infrared system, and its performance directly determines image quality and temperature measurement accuracy. With the widespread application of infrared technology in military, industrial, medical, and security fields, higher demands are placed on the stability and reliability of infrared cameras across a wide temperature range (-40℃ to 70℃). However, the gain characteristics of an infrared focal plane array are easily affected by changes in its operating temperature, leading to significant image drift. This drift primarily manifests as a nonlinear change in the detector's response gain, severely reducing image uniformity, introducing temperature measurement errors, and even affecting the system's practicality in dynamic temperature environments.

[0003] Currently, while various methods have been attempted to compensate for temperature drift, significant limitations remain. Traditional two-point calibration methods (such as calibration at only -40℃ and 70℃) are only effective near specific temperatures and cannot cover gain nonlinear drift over a wide temperature range from -40℃ to 70℃. Some studies employ scene-based calibration algorithms, but these methods rely on scene motion and are prone to over-calibration problems such as "ghosting" in static or slowly moving scenes. Furthermore, although some studies have attempted to introduce polynomial fitting for compensation, high-order polynomial fitting involves large computational costs, poor real-time performance, and does not significantly improve the accuracy of image temperature measurement; while low-order ordinary polynomials cannot adapt to gain nonlinear changes over a wide temperature range, resulting in low engineering application value. Although linear models or simple piecewise strategies can partially alleviate drift, their compensation accuracy and response speed are still insufficient in extreme high and low temperature ranges and when temperatures change rapidly, failing to simultaneously meet the requirements of high accuracy and high real-time performance.

[0004] Therefore, there is an urgent need for a gain compensation method that can achieve high precision, fast response, and scene independence in a wide temperature range of -40℃ to 70℃, so as to improve the overall performance and adaptability of infrared imaging systems. Summary of the Invention

[0005] To overcome the limitations of existing single-temperature-point or simple segmented compensation methods in maintaining high accuracy and fast response across a wide temperature range, and ineffective handling of the nonlinear characteristics of gain variation with temperature, this invention provides a gain compensation method for focal plane arrays under wide temperature conditions. This method addresses the problems of poor image uniformity and decreased temperature measurement accuracy caused by gain nonlinear drift under wide temperature conditions.

[0006] The technical solution of this invention is as follows:

[0007] A gain compensation method for a focal plane array mechanism operating under wide temperature range includes the following steps:

[0008] Step 1: Divide the temperature range into non-uniform sub-regions within a wide temperature range of -40℃ to 70℃;

[0009] Step 2: Fit the Chebyshev polynomial within each temperature sub-interval to establish a gain prediction model for each temperature sub-interval, and determine the calculation relationship of the gain compensation coefficient for each temperature sub-interval based on the gain prediction model.

[0010] Step 3: Real-time acquisition of the operating temperature of the focal plane array;

[0011] Step 4: Determine the temperature sub-interval to which the temperature belongs based on the real-time collected working temperature, call the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, and calculate the real-time gain compensation coefficient by combining the lookup table and linear interpolation.

[0012] Step 5: Apply the real-time gain compensation coefficient to the original output signal of the focal plane array mechanism to achieve gain compensation.

[0013] As a preferred technical solution of the present invention, the basis for dividing the non-uniform temperature sub-intervals in step 1 includes: the slope change points or curvature change points of the gain-temperature curve.

[0014] Optionally, the temperature sub-ranges specifically include: a low temperature range of -40 to -15℃, a medium low temperature range of -15 to 10℃, a normal temperature range of 10 to 45℃, and a high temperature range of 45 to 70℃, with the interval division density of the low temperature range and the high temperature range being greater than that of the normal temperature range.

[0015] As a preferred embodiment of the present invention, the Chebyshev polynomial in step 2 is a second-order Chebyshev polynomial, and the fitting process includes the following steps:

[0016] Step 201: Select at least 4 temperature calibration nodes uniformly within each temperature sub-interval, and measure the actual gain value corresponding to each temperature calibration node;

[0017] Step 202: Map the temperature values ​​of each temperature calibration node from the temperature range [a,b] of its respective temperature sub-interval to the interval [-1,1] to obtain the mapping value X. The mapping formula is: X=[2t-(a+b)] / (ba);

[0018] Where t is the temperature value of the temperature calibration node, a is the lower limit temperature of the temperature sub-interval, and b is the upper limit temperature of the temperature sub-interval.

[0019] Step 203: Based on the basis functions t0(x) = 1, t1(x) = x, t2(x) = 2x of the second-order Chebyshev polynomial. 2 -1, construct a system of linear equations with respect to the polynomial coefficients c0, c1, and c2. Where A is the matrix of values ​​of the basis function at each mapping value x. This is the vector of actual gain values ​​for each temperature calibration node;

[0020] Step 204: Solve the linear equation system using the least squares method to obtain the coefficients c0, c1, and c2 of the second-order Chebyshev polynomial, and then establish the gain prediction model: G pred (t)=c0+c1·t1(x)+c2·t2(x), where G pred (t) represents the predicted gain value corresponding to temperature t.

[0021] Furthermore, after establishing the gain prediction model for each temperature sub-interval in step 2, a calibration and verification step is also included:

[0022] Within each temperature sub-interval, select at least two verification temperature points that are not involved in polynomial fitting, and ensure that the verification temperature points do not coincide with the temperature calibration nodes in step 201.

[0023] Measure and verify the actual gain value G corresponding to the temperature point. actual The gain prediction value G for the corresponding verification temperature point is calculated using the gain prediction model for the temperature sub-interval. pred ;

[0024] Calculate the gain error rate σ = |(G) at each verification temperature point. pred -G actual ) / G actual If the gain error rate σ of all verification temperature points is less than or equal to 2%, the gain prediction model of the temperature sub-interval is deemed to be qualified. If there are verification temperature points with a gain error rate σ > 2%, the number of temperature calibration nodes in the temperature sub-interval is increased, and the fitting process in step 2 is repeated until the model meets the error requirements.

[0025] Furthermore, the gain compensation coefficient calculation relationship in step 2 is as follows: the gain compensation coefficient C(t) is equal to the gain prediction value G output by the gain prediction model. pred The reciprocal of (t), i.e., C(t) = 1 / G pred (t).

[0026] As a preferred embodiment of the present invention, the specific method for real-time acquisition of the operating temperature of the focal plane array in step 3 includes:

[0027] A high-precision digital temperature sensor is mounted close to the substrate of the focal plane array chip. The high-precision digital temperature sensor is a PT100 sensor or a high-precision thermistor. The output signal of the high-precision digital temperature sensor is converted into a digital temperature signal by an analog-to-digital converter and then transmitted to a microcontroller or field-programmable gate array with a hardware floating-point unit for signal processing.

[0028] As a preferred embodiment of the present invention, the process of generating the lookup table in step 4 includes:

[0029] Before system startup, multiple temperature points t are selected for each temperature sub-range with a temperature step size of 0.1℃ to 0.5℃. i ;

[0030] Based on the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, calculate t for each temperature point. i The corresponding gain compensation coefficient C(t) i );

[0031] Temperature point t i With the corresponding gain compensation coefficient C(t) i The corresponding data is stored in the system's flash memory or random access memory, forming a lookup table for each temperature sub-range.

[0032] As a preferred embodiment of the present invention, step 4, which combines lookup table and linear interpolation to calculate the real-time gain compensation coefficient, specifically includes the following steps:

[0033] Step 401: Record the real-time operating temperature t current By comparing the boundary temperatures of each temperature sub-interval, t is determined. current The temperature sub-range to which it belongs;

[0034] Step 402: Find t in the lookup table corresponding to the temperature sub-interval. current Two adjacent temperature points t l and t h and the corresponding gain compensation coefficient C l and C h , where t l <t current <t h ;

[0035] Step 403: Calculate t using the linear interpolation formula. current The corresponding real-time gain compensation coefficient C current The linear interpolation formula is: C current =C l +(C h -C l )·[(t current -tl ) / (t h -t l )).

[0036] As a preferred embodiment of the present invention, the specific steps for applying the real-time gain compensation coefficient in step 5 include:

[0037] Read the raw voltage signal V output from the focal plane array raw ;

[0038] The original voltage signal V raw With real-time gain compensation coefficient C current Multiplying them together yields the corrected voltage signal V. comp ;

[0039] The corrected voltage signal is transmitted to the image processing module for image generation and display.

[0040] According to the above-described solution, the beneficial effects of this invention are as follows:

[0041] This invention uses non-uniform temperature sub-divisions based on the nonlinear characteristics of gain and fits the results using Chebyshev polynomials within each temperature sub-division. This approach can approximate the actual drift curve of the gain with extremely high accuracy, thereby greatly improving the linearity of the output signal and the accuracy of radiation measurement across the entire temperature range, and thus improving image uniformity.

[0042] This invention achieves millisecond-level compensation response speed by combining lookup tables with real-time calculations of linear interpolation. It can quickly track changes in the working temperature of the focal plane array and does not rely on scene motion, thus avoiding the problem of "ghosting" over-correction and meeting the needs of dynamic applications.

[0043] The compensation method of the present invention effectively suppresses image artifacts and contrast reduction caused by wide temperature variations, while improving image uniformity, enhancing temperature measurement accuracy, and improving the working stability and output reliability of the focal plane array mechanism under harsh temperature environments. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0045] To better understand the purpose, technical solution, and technical effects of this invention, the invention will be further explained and described below in conjunction with the accompanying drawings and embodiments. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. It is also stated that the embodiments described below are only for explaining this invention and are not intended to limit this invention.

[0046] like Figure 1 As shown, a gain compensation method for a focal plane array mechanism under wide temperature conditions includes the following steps:

[0047] Step 1: Divide the temperature range into non-uniform sub-regions within a wide temperature range of -40℃ to 70℃;

[0048] The non-uniform partitioning in step 1 is crucial, making the partitioning of intervals with strong gain nonlinearity denser and the partitioning of intervals with good linearity sparser. For example, the high and low temperature regions are partitioned more densely and the normal temperature region is partitioned more sparsely. This avoids the problem of insufficient fitting accuracy in strong nonlinear intervals and redundant calculations in linear intervals caused by traditional uniform segmentation.

[0049] Step 2: Fit the Chebyshev polynomial within each temperature sub-interval to establish a gain prediction model for each temperature sub-interval, and determine the calculation relationship of the gain compensation coefficient for each temperature sub-interval based on the gain prediction model.

[0050] In step 2, a low-order Chebyshev polynomial is used to approximate the nonlinear curve with high accuracy, while avoiding the computational redundancy of high-order polynomials; the calculation relationship of the gain compensation coefficient transforms the predicted gain into a coefficient rule that can be used for correction, so as to realize the key transformation from the law of "temperature affects gain" to the correction rule of "using coefficients to offset the effect".

[0051] Step 3: Real-time acquisition of the operating temperature of the focal plane array;

[0052] Since the direct cause of gain drift is the temperature change of the focal plane array itself, collecting the operating temperature of the focal plane array itself, rather than the ambient temperature, provides an accurate temperature input for targeted calculation of the compensation coefficient. This step 3 can avoid compensation deviation caused by ambient temperature interference.

[0053] Step 4: Determine the temperature sub-interval to which the real-time collected working temperature belongs, call the fitting model of the temperature sub-interval to ensure compensation for the nonlinear characteristics of the current temperature; call the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, and calculate the real-time gain compensation coefficient by combining lookup table and linear interpolation. The lookup table method can avoid the calculation delay of real-time fitting, and linear interpolation fills the blanks of discrete points in the lookup table to ensure the accuracy of coefficients at any real-time temperature; this step 4 helps to solve the problems of large calculation volume and high latency in traditional real-time fitting.

[0054] Step 5: Apply the real-time gain compensation coefficient to the original output signal of the focal plane array mechanism to achieve gain compensation;

[0055] Step 5 can counteract the nonlinear drift of gain with temperature and output a corrected signal; and the corrected signal is ultimately used for image generation.

[0056] By employing the above method, the compensation temperature range is clearly defined as -40℃ to 70℃, precisely matching the actual application scenarios of infrared focal plane arrays, such as high and low temperature environments in the military, infrared monitoring scenarios for high-temperature equipment in industry, outdoor day-night temperature differences in the security field, and low-temperature cold chain infrared temperature measurement scenarios in medical applications. This overcomes the limitations of traditional compensation methods, which are only applicable to narrow temperature ranges (0℃~50℃) and cannot cope with extreme temperature environments. This invention overcomes the bottleneck of nonlinear compensation by combining non-uniform temperature sub-interval division and Chebyshev polynomial fitting rules, improving the compensation accuracy across the entire temperature range and significantly enhancing the linearity of the output signal within the -40℃ to 70℃ range. This invention achieves a balance between real-time performance and computational accuracy through lookup table pre-storage and linear interpolation calculation, meeting dynamic application requirements and ensuring continuous and accurate compensation coefficients at any real-time temperature. Ultimately, it achieves millisecond-level compensation response speed, enabling rapid tracking of dynamic changes in the operating temperature of the focal plane array. The method of this invention, through temperature zoning, model establishment, real-time temperature acquisition, coefficient calculation, and compensation application, focuses on offsetting gain nonlinear drift in each step, effectively suppressing problems such as image drift, artifacts, and contrast reduction caused by wide temperature variations, and improving the working stability and output reliability of the focal plane array in harsh temperature environments.

[0057] In this invention, the basis for dividing the non-uniform temperature sub-intervals in step 1 includes: the slope change point or curvature change point of the gain-temperature curve. The slope change point of the gain-temperature curve refers to the rate at which the gain changes with temperature, i.e., the temperature node where the curve slope changes significantly. For example, through actual measurement, it was found that the slope of the gain-temperature curve of the focal plane array mechanism increases sharply from 0.03V / ℃ to 0.09V / ℃ at -30℃; this -30℃ is the slope change point, which is used as a reference when dividing the intervals. The curvature change point is the temperature node where the degree or direction of curvature of the gain-temperature curve changes. For example, actual measurement shows that the curvature value of the curve changes from positive to negative at -15℃; therefore, the curvature change point at -15℃ becomes the basis for dividing the low-temperature region from -40℃ to -15℃ and the medium-low-temperature region from -15℃ to 10℃. The partitioning criteria ensure that the division of non-uniform sub-intervals is based entirely on the objective laws of gain change or actual accuracy requirements. This allows for fine partitioning of high and low temperature regions with drastic gain changes due to the dense concentration of slope and curvature change points, ensuring that subsequent Chebyshev polynomial fitting can accurately capture the gain change laws. At the same time, it allows for a wider range of intervals in the room temperature region with gentle gain changes, avoiding unnecessary interval subdivision that would increase the storage of lookup tables and computational redundancy. This improves the flexibility of the method and effectively avoids fitting deviations caused by unreasonable partitioning.

[0058] In one specific embodiment, the temperature sub-intervals specifically include: a low-temperature region of -40 to -15°C, a medium-low-temperature region of -15 to 10°C, a normal-temperature region of 10 to 45°C, and a high-temperature region of 45 to 70°C. The interval division density of the low-temperature and high-temperature regions is greater than that of the normal-temperature region. These four temperature sub-intervals are determined based on the nonlinear characteristics of the focal plane array's gain-temperature curve. The low-temperature region (-40 to -15°C) and the high-temperature region (45 to 70°C) are areas where the gain changes drastically with temperature. The gain changes in the medium-low-temperature region (-15 to 10°C) and the normal-temperature region (10 to 45°C) gradually become more gradual. In particular, the gain characteristics of the focal plane array are most stable and exhibit the lowest nonlinearity in the normal-temperature region (10 to 45°C). On the one hand, the low-temperature and high-temperature regions are the areas with the most severe nonlinear gain drift. A denser division in these regions can ensure the collection of sufficient temperature-gain data, providing richer samples for Chebyshev polynomial fitting. This allows for the accurate capture of subtle gain variation patterns and avoids the fitting model failing to reflect the true drift characteristics due to sparse data points, thereby effectively improving the compensation accuracy under extreme high and low temperature environments. On the other hand, the room-temperature region, as a region with gentle gain changes, can be significantly reduced in terms of the number of calibration nodes, lookup table storage, and computational load by using a sparser division.

[0059] In this invention, the Chebyshev polynomial is a second-order Chebyshev polynomial, and the fitting process includes the following steps:

[0060] Step 201: Select at least 4 temperature calibration nodes uniformly within each temperature sub-interval, and measure the actual gain value corresponding to each temperature calibration node;

[0061] Step 202: Map the temperature values ​​of each temperature calibration node from the temperature range [a,b] of its respective temperature sub-interval to the interval [-1,1] to obtain the mapping value X. The mapping formula is: X=[2t-(a+b)] / (ba);

[0062] Where t is the temperature value of the temperature calibration node, a is the lower limit temperature of the temperature sub-interval, and b is the upper limit temperature of the temperature sub-interval.

[0063] Step 203: Based on the basis functions t0(x) = 1, t1(x) = x, t2(x) = 2x of the second-order Chebyshev polynomial. 2 -1, construct a system of linear equations with respect to the polynomial coefficients c0, c1, and c2. Where A is the matrix of values ​​of the basis function at each mapping value x. This is the vector of actual gain values ​​for each temperature calibration node;

[0064] In one specific embodiment, A is an n-row, 3-column vector, where n is the number of temperature calibration nodes selected within the temperature sub-interval, and n is greater than or equal to 4. Each row contains elements representing the t0, t1, and t2 values ​​at the corresponding temperature calibration node's mapping value x. T denotes the transpose sign, and this vector represents the coefficients of the second-order Chebyshev polynomial to be solved; vector G2, ..., G n ] T T represents the transpose symbol, and is a column vector consisting of the actual gain values ​​corresponding to each temperature calibration node.

[0065] Step 204: Solve the linear equation system using the least squares method to obtain the coefficients c0, c1, and c2 of the second-order Chebyshev polynomial, and then establish the gain prediction model: G pred (t)=c0+c1·t1(x)+c2·t2(x), where G pred (t) represents the predicted gain value corresponding to temperature t.

[0066] Taking the medium-low temperature range (-15~10℃, a=-15, b=10) as an example, four calibration nodes of -15℃, -5℃, 5℃, and 10℃ can be uniformly selected within this temperature sub-range. Experimental measurements show that the actual gain values ​​for each node are 0.80, 1.00, 1.20, and 1.30, respectively. Using the formula X=[2t-(a+b)] / (ba), we calculate that a+b=-15+10=-5, ba=25. Therefore, the mapping value X corresponding to -15℃ is -1.0, the mapping value X corresponding to -5℃ is -0.2, the mapping value X corresponding to 5℃ is 0.6, and the mapping value X corresponding to 10℃ is 1.0, thus completing the correspondence between temperature and mapping value.

[0067] Execute step 203: Based on the basis functions t0(x) = 1, t1(x) = x, t2(x) = 2x of the second-order Chebyshev polynomial. 2 -1, substitute the mapping value X of each calibrated node into the basis function to obtain the values, forming an n x 3 matrix A. When the first node X = -1.0, t0 = 1, t1 = -1.0, t2 = 1, so the first row of A is [1, -1.0, 1]; when the second node X = -0.2, the second row of A is [1, -0.2, 0.92], and so on to form the complete matrix A; vector The coefficients to be determined are a vector. The column vector consisting of the actual gain values ​​of each node is [0.80, 1.00, 1.20, 1.30]. T Thus, a system of linear equations is constructed:

[0068]

[0069] Step 204 involves solving the linear equation system using the least squares method, yielding the following results: c0≈1.05, c1≈0.25, c2≈-0.15. This leads to the gain prediction model G for the low-temperature region. pred (t)=1.05+0.25·t1(x)-0.15·t2(x), where x is the mapping value of temperature t.

[0070] As can be seen in step 2, the second-order polynomial significantly reduces the computational load compared to traditional high-order polynomials, allowing model calculations to be completed without a high-performance processor. This is particularly suitable for embedded systems commonly used in focal plane arrays, such as MCUs with FPUs, avoiding the real-time performance issues caused by high-order fitting. The minimum-maximum error characteristic of the Chebyshev polynomial, combined with temperature mapping operations, makes its approximation accuracy of the gain-temperature curve in each sub-interval far superior to that of ordinary polynomials of the same order. Taking the aforementioned low-temperature region as an example, the error between the -10℃ gain prediction calculated by this model and the actual measured value can be controlled within 2%, far better than the 3% or even 5% error of ordinary second-order polynomials. This effectively solves the problem of difficulty in accurately fitting the nonlinear drift of gain over a wide temperature range, especially in high and low temperature sections. Simultaneously, the combined use of at least four calibration nodes and the least squares method effectively cancels out the measurement noise of a single calibration node, making the fitting model more robust. Even if there are slight deviations in the data of individual nodes, the overall prediction accuracy can still be guaranteed. Different focal plane array mechanisms can quickly establish a suitable gain prediction model by simply adjusting the temperature and gain data of the calibration nodes according to this process, without the need to redesign the fitting logic, thus improving the versatility and portability of the solution.

[0071] In a preferred embodiment, after establishing the gain prediction model for each temperature sub-interval in step 2, a calibration and verification step is also included:

[0072] Select at least two verification temperature points that are not involved in polynomial fitting within each temperature sub-interval, and ensure that the verification temperature points do not coincide with the temperature calibration nodes in step 201; avoid using data that has already been fitted for verification to ensure the objectivity of the verification results. If the verification point coincides with the calibration node, the model's predicted value for that point will inevitably be close to the actual value, and the model will not be able to detect the fitting deviation of other temperature points within the temperature sub-interval.

[0073] Measure and verify the actual gain value G corresponding to the temperature point. actual The gain prediction value G for the corresponding verification temperature point is calculated using the gain prediction model for the temperature sub-interval. pred ;

[0074] Calculate the gain error rate σ = |(G) at each verification temperature point. pred -G actual ) / G actualIf the gain error rate σ of all verification temperature points is ≤ 2%, it is determined that the gain prediction model of the temperature sub - interval is qualified; if there is a verification temperature point with a gain error rate σ > 2%, increase the number of temperature calibration nodes in the temperature sub - interval and re - execute the fitting process in step 2 until the model meets the error requirements.

[0075] Taking the aforementioned medium - low temperature range (-15 to 10 °C) as an example, if the temperature calibration nodes selected in step 201 are -15 °C, -5 °C, 5 °C, and 10 °C, the verification temperature points can be selected as -10 °C and 2 °C, neither of which coincides with the calibration nodes; through experimental measurement, the actual gain value G corresponding to -10 °C actual = 1.10, and the G corresponding to 2 °C actual = 1.15. Then, using the gain prediction model already established for this temperature sub - interval, the calculated G corresponding to -10 °C pred = 1.09, and the G corresponding to 2 °C pred = 1.14. Substitute them into the error rate formula respectively to calculate: the error rate σ of the verification point -10 °C is approximately 0.9%, and the error rate σ of the verification point 2 °C is approximately 0.8%, both ≤ 2%. It is determined that the gain prediction model of this medium - low temperature range is qualified. If the error rate σ of the two verification points is > 2%, add two new temperature calibration nodes -12 °C and 3 °C in this medium - low temperature range, re - execute the fitting process from step 201 to step 204, and verify again until the error rate meets the standard.

[0076] It can be seen that through the setting of independent verification points in the calibration verification step, the risk that the model only adapts to the calibration nodes and has a large prediction deviation for other temperature points within the temperature sub - interval is effectively avoided, ensuring that the gain prediction model has uniform high precision throughout the temperature sub - interval, which may cause obvious artifacts in the subsequent compensated image. After verification, the error of such non - calibration nodes can be controlled within 2%, ensuring the uniformity of full - temperature - range imaging and the stability of temperature measurement accuracy. On the other hand, this verification mechanism enables the model to have an adaptive optimization ability. Without replacing the fitting algorithm or adjusting the temperature sub - interval division, it can cope with the differences in the gain - temperature characteristics of different types of focal - plane cores, such as InGaAs and mercury cadmium telluride, and improve the model adaptability by supplementing calibration nodes.

[0077] In the present invention, the calculation relationship of the gain compensation coefficient in step 2 is: the gain compensation coefficient C(t) is equal to the reciprocal of the gain prediction value G pred (t) output by the gain prediction model, that is, C(t) = 1 / G pred (t). When the gain of the focal - plane array drifts due to temperature changes, a compensation coefficient is required to offset this drift so that the corrected output signal returns to stability; specifically, if the gain of the focal - plane array in the ideal state is set to 1, then when the actual gain prediction value at the current temperature is G predWhen (t), the original output signal will be amplified or reduced by G due to this gain drift. pred (t) times, at which point the original signal is multiplied by G. pred The reciprocal of (t) can cancel out the effect of gain drift, resulting in a correction signal close to the ideal state. The predicted gain value G at 25℃ is calculated using the fitted model. pred Taking (25℃)=1.15 as an example, the actual gain at this temperature is 15% higher than the ideal gain. Therefore, according to C(25℃)=1 / G pred The calculation yields C(25℃)≈0.87, when the focal plane array mechanism outputs the original voltage signal V at 25℃. raw When V = 2.3V, V raw Multiplying by C (25℃) yields the corrected voltage signal V. comp =2.3V×0.87≈2.0V, the output value under the ideal gain is restored, and the gain drift is accurately canceled.

[0078] As can be seen, this step can obtain the compensation coefficient from the gain prediction value through only one reciprocal operation, without the need for complex matrix operations or iterative solutions; secondly, the calculation relationship of the gain compensation coefficient is deeply synergistic with the gain prediction model, because G pred (t) is a reliable value fitted by Chebyshev polynomial. Its reciprocal can accurately match the degree of gain drift at the current temperature. Compared with traditional fixed coefficient compensation or empirical formula compensation, it can further control the compensated gain error within 2%, effectively improving the problem of large temperature measurement deviation caused by inaccurate compensation. Especially in the extreme range of high and low temperatures, it can more accurately offset the severe nonlinear gain drift.

[0079] In one specific embodiment, the method for real-time acquisition of the operating temperature of the focal plane array includes:

[0080] A high-precision digital temperature sensor, either a PT100 sensor or a high-precision thermistor, is mounted in close proximity to the substrate of the focal plane array chip. The output signal of the high-precision digital temperature sensor is converted into a digital temperature signal by an analog-to-digital converter and then transmitted to a microcontroller or field-programmable gate array (FPGA) with a hardware floating-point unit for signal processing. The gain drift of the focal plane array is directly determined by its own operating temperature, not the ambient temperature. Placing the sensor in close proximity to the substrate minimizes temperature conduction losses and avoids interference from ambient temperature drift on the temperature measurement results. The PT100 sensor or high-precision thermistor is chosen because both are adaptable to a wide temperature range of -40℃ to 70℃ and possess high stability: the PT100 sensor achieves a measurement accuracy of ±0.1℃ across the entire temperature range, making it suitable for scenarios requiring extremely high temperature accuracy.

[0081] From the perspective of signal processing, the original signal output by the sensor is an analog signal, which needs to be converted into a digital temperature signal by an analog-to-digital converter. To avoid distortion caused by electromagnetic interference during transmission, an analog-to-digital converter with a resolution of 16 bits or higher is usually selected to ensure the digital accuracy of the temperature signal.

[0082] In this invention, the lookup table mentioned in step 4 is pre-generated and stored offline before the system officially starts running. The generation process includes:

[0083] Before system startup, multiple temperature points t are selected for each temperature sub-range with a temperature step size of 0.1℃ to 0.5℃. i This temperature step range balances compensation accuracy with storage resources.

[0084] Based on the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, calculate t for each temperature point. i The corresponding gain compensation coefficient C(t) i );

[0085] Temperature point t i With the corresponding gain compensation coefficient C(t) i The corresponding values ​​are stored in the system's flash memory or random access memory, forming a lookup table for each temperature sub-interval. Because the gain prediction model and the relationship between gain and compensation coefficient calculation differ between different temperature sub-intervals, interval storage ensures that subsequent real-time calls can accurately match the temperature sub-interval to which the current temperature belongs, avoiding compensation deviations caused by mixing coefficients across intervals.

[0086] Since the gain compensation coefficient has been calculated offline before the system starts, it is only necessary to call the lookup table data through temperature matching. There is no need to perform Chebyshev polynomial fitting and gain compensation coefficient calculation in real time. The calculation process that originally required tens of microseconds is shortened to nanosecond level reading. The system can achieve millisecond level response and effectively cope with the scenario of rapid temperature change of the focal plane array.

[0087] Step 4, which combines lookup tables and linear interpolation to calculate the real-time gain compensation coefficient, includes the following specific steps:

[0088] Step 401: Record the real-time operating temperature t current By comparing the boundary temperatures of each temperature sub-interval, t is determined. current The temperature sub-interval to which it belongs; for example, t current =28℃, by comparison we can know that it belongs to the normal temperature range of 10~45℃, and the corresponding lookup table is accurately matched; as mentioned above, the lookup tables of different sub-intervals are generated based on their respective gain prediction models, avoiding coefficient deviation caused by calling the lookup table across sub-intervals;

[0089] Step 402: Find t in the lookup table corresponding to the temperature sub-interval. current Two adjacent temperature points t l and t h and the corresponding gain compensation coefficient C l and C h , where t l <t current <t h The lookup table stores temperature points and coefficients in discrete step sizes of 0.1℃ to 0.5℃, with real-time temperature t. current It is highly unlikely that the value will completely coincide with the discrete point in the lookup table; the value of t needs to be located in the lookup table of the corresponding temperature sub-interval. current The two most recent discrete temperature points t l and t h And extract the corresponding compensation coefficient;

[0090] Step 403: Calculate t using the linear interpolation formula. current The corresponding real-time gain compensation coefficient C current The linear interpolation formula is: C current =C l +(C h -C l )·[(t current -t l ) / (t h -t l Discrete C l and C h Smooth transition to continuous C current This enables precise integration between discrete lookup tables and continuous temperature data. First, linear interpolation effectively overcomes the limitations of discrete storage in lookup tables. Second, linear interpolation involves only basic addition, subtraction, multiplication, and division operations, eliminating the need for complex polynomial fitting or matrix operations. It can complete calculations in microseconds, meeting the millisecond-level compensation requirements of focal plane array mechanisms.

[0091] In this invention, the specific steps for applying the real-time gain compensation coefficient in step 5 include:

[0092] Read the raw voltage signal V output from the focal plane array raw The signal is unprocessed and truly reflects the nonlinear gain drift caused by temperature changes.

[0093] The original voltage signal V raw With real-time gain compensation coefficient C current Multiplying them together yields the corrected voltage signal V. comp When V raw When the gain is too high and therefore too large, C current It can be pulled back to the ideal range; when V rawWhen the gain is too low, C current This can be enhanced to achieve normalized correction of the gain;

[0094] The corrected voltage signal is transmitted to the image processing module for image generation and display. The corrected effective signal can be converted into a uniform and clear infrared image, completing the transformation from electrical signal correction to visual imaging.

[0095] In summary, the focal plane array gain compensation method of this invention, operating under wide temperature conditions, systematically addresses the core pain points of traditional compensation methods in terms of wide temperature range adaptability, nonlinear compensation accuracy, and real-time response speed through temperature non-uniform partitioning, model construction, real-time temperature acquisition, coefficient calculation, and signal correction. Step 1, based on the non-uniform partitioning of the gain-temperature curve characteristics, lays the interval foundation for subsequent accurate fitting, avoiding the contradiction between accuracy and efficiency in uniform partitioning. Step 2 employs second-order Chebyshev polynomial fitting combined with the calibration verification in step 5, achieving high-precision approximation of nonlinear gain drift with low computational complexity, while ensuring model generalization ability through independent verification points, avoiding the risk of overfitting. Step 3's high-precision close-fitting temperature acquisition design ensures the authenticity of temperature input, providing a reliable basis for compensation coefficient calculation. Step 4, through the combination of offline pre-generated lookup tables and online linear interpolation, significantly reduces real-time calculation latency while filling the accuracy gaps in discrete data, achieving millisecond-level response. Finally, step 5's signal multiplication correction transforms all previous calculation results into an improvement in actual imaging quality, ensuring that the corrected signal can directly generate uniform and accurate infrared images.

[0096] This invention not only accurately covers a wide temperature range from -40℃ to 70℃, adapting to the practical needs of various fields such as military, industry, security, and medical, but also possesses strong engineering adaptability. Whether it's the differences in gain characteristics between different focal plane array mechanisms or the varying requirements for accuracy and cost in different scenarios, it can be flexibly adapted by adjusting partition density, lookup table step size, and hardware selection, without requiring reconstruction of the core logic. Overall, this invention achieves a balance between high precision and high real-time performance in wide-temperature gain compensation through deep collaboration among its various steps, effectively improving the operational stability and imaging reliability of focal plane array mechanisms in harsh temperature environments.

[0097] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0098] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A gain compensation method for a focal plane array mechanism under wide temperature range, characterized in that, Includes the following steps: Step 1: Divide the temperature range into non-uniform sub-regions within a wide temperature range of -40℃ to 70℃; Step 2: Fit the Chebyshev polynomial within each temperature sub-interval to establish a gain prediction model for each temperature sub-interval, and determine the calculation relationship of the gain compensation coefficient for each temperature sub-interval based on the gain prediction model. Step 3: Real-time acquisition of the operating temperature of the focal plane array; Step 4: Determine the temperature sub-interval to which the temperature belongs based on the real-time collected working temperature, call the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, and calculate the real-time gain compensation coefficient by combining the lookup table and linear interpolation. Step 5: Apply the real-time gain compensation coefficient to the original output signal of the focal plane array mechanism to achieve gain compensation.

2. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The criteria for dividing non-uniform temperature sub-intervals in step 1 include: the points where the slope or curvature of the gain-temperature curve changes.

3. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The temperature sub-ranges specifically include: a low temperature range of -40 to -15℃, a medium low temperature range of -15 to 10℃, a normal temperature range of 10 to 45℃, and a high temperature range of 45 to 70℃, with the interval division density of the low temperature range and the high temperature range being greater than that of the normal temperature range.

4. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The Chebyshev polynomial mentioned in step 2 is a second-order Chebyshev polynomial, and the fitting process includes the following steps: Step 201: Select at least 4 temperature calibration nodes uniformly within each temperature sub-interval, and measure the actual gain value corresponding to each temperature calibration node; Step 202: Map the temperature values ​​of each temperature calibration node from the temperature range [a,b] of its respective temperature sub-interval to the interval [-1,1] to obtain the mapping value X. The mapping formula is: X=[2t-(a+b)] / (ba); Where t is the temperature value of the temperature calibration node, a is the lower limit temperature of the temperature sub-interval, and b is the upper limit temperature of the temperature sub-interval. Step 203: Based on the basis functions t0(x) = 1, t1(x) = x, t2(x) = 2x of the second-order Chebyshev polynomial. 2 -1, construct a system of linear equations with respect to the polynomial coefficients c0, c1, and c2. Where A is the matrix of values ​​of the basis function at each mapping value x. This is the vector of actual gain values ​​for each temperature calibration node; Step 204: Solve the linear equation system using the least squares method to obtain the coefficients c0, c1, and c2 of the second-order Chebyshev polynomial, and then establish the gain prediction model: G pred (t)=c0+c1·t1(x)+c2·t2(x), where G pred (t) represents the predicted gain value corresponding to temperature t.

5. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 4, characterized in that, After establishing the gain prediction model for each temperature sub-interval in step 2, the process also includes a calibration and verification step: Within each temperature sub-interval, select at least two verification temperature points that are not involved in polynomial fitting, and ensure that the verification temperature points do not coincide with the temperature calibration nodes in step 201. Measure and verify the actual gain value G corresponding to the temperature point. actual The gain prediction value G for the corresponding verification temperature point is calculated using the gain prediction model for the temperature sub-interval. pred ; Calculate the gain error rate σ = |(G) at each verification temperature point. pred -G actual ) / G actual If the gain error rate σ of all verification temperature points is less than or equal to 2%, the gain prediction model of the temperature sub-interval is deemed to be qualified. If there are verification temperature points with a gain error rate σ > 2%, the number of temperature calibration nodes in the temperature sub-interval is increased, and the fitting process in step 2 is repeated until the model meets the error requirements.

6. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 4, characterized in that, The gain compensation coefficient calculation relationship in step 2 is as follows: the gain compensation coefficient C(t) is equal to the gain prediction value G output by the gain prediction model. pred The reciprocal of (t), i.e., C(t) = 1 / G pred (t).

7. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The specific methods for real-time acquisition of the focal plane array's operating temperature in step 3 include: A high-precision digital temperature sensor is mounted close to the substrate of the focal plane array chip. The high-precision digital temperature sensor is a PT100 sensor or a high-precision thermistor. The output signal of the high-precision digital temperature sensor is converted into a digital temperature signal by an analog-to-digital converter and then transmitted to a microcontroller or field-programmable gate array with a hardware floating-point unit for signal processing.

8. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The process of generating the lookup table in step 4 includes: Before system startup, multiple temperature points t are selected for each temperature sub-range with a temperature step size of 0.1℃ to 0.5℃. i ; Based on the gain compensation coefficient calculation relationship corresponding to the temperature sub-interval, calculate t for each temperature point. i The corresponding gain compensation coefficient C(t) i ); Temperature point t i With the corresponding gain compensation coefficient C(t) i The corresponding data is stored in the system's flash memory or random access memory, forming a lookup table for each temperature sub-range.

9. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, Step 4, which combines lookup tables and linear interpolation to calculate the real-time gain compensation coefficient, includes the following specific steps: Step 401: Record the real-time operating temperature t current By comparing the boundary temperatures of each temperature sub-interval, t is determined. current The temperature sub-range to which it belongs; Step 402: Find t in the lookup table corresponding to the temperature sub-interval. current Two adjacent temperature points t l and t h and the corresponding gain compensation coefficient C l and C h , where t l <t current <t h ; Step 403: Calculate t using the linear interpolation formula. current The corresponding real-time gain compensation coefficient C current The linear interpolation formula is: C current =C l +(C h -C l )·[(t current -t l ) / (t h -t l )).

10. The gain compensation method for a focal plane array mechanism under wide temperature conditions according to claim 1, characterized in that, The specific steps for applying the real-time gain compensation coefficient in step 5 include: Read the raw voltage signal V output from the focal plane array raw ; The original voltage signal V raw With real-time gain compensation coefficient C current Multiplying them together yields the corrected voltage signal V. comp ; The corrected voltage signal is transmitted to the image processing module for image generation and display.

Images