Calibration methods for infrared detectors, infrared imaging equipment, and storage media

Abstract

This application provides a calibration method, an infrared imaging device, and a storage medium for an infrared detector. The method includes: acquiring the original response values ​​of each pixel of the infrared detector; determining the current response value interval of each pixel based on the original response values ​​and the boundaries of multiple pre-calibrated response value intervals of the infrared detector, wherein multiple response value intervals are pre-calibrated based on the characteristics of the response curves of each integration channel of the infrared detector; acquiring the correction coefficient corresponding to the current response value interval of each pixel based on the pre-calibrated correction coefficients corresponding to each pixel in different response value intervals; correcting the original response values ​​of each pixel based on the correction coefficients; and outputting the corrected response values ​​of each pixel. This method effectively improves the non-uniformity correction effect of the image, especially in scenes with large temperature differences, and can significantly suppress horizontal stripe residue caused by nonlinear response.

Description

Technical Field

[0001] This application relates to the field of infrared imaging technology, and in particular to a calibration method for an infrared detector, an infrared imaging device, and a storage medium. Background Technology

[0002] Uncooled infrared detectors, which require no cooling device and can operate at room temperature, offer advantages such as low cost, small size, and low power consumption, leading to their widespread application in infrared imaging. It is worth noting that in high-speed dynamic imaging scenarios, such as real-time fault detection in industrial production lines, low-altitude reconnaissance by drones, and sudden obstacle recognition in autonomous driving, the system frame rate typically needs to reach above 100Hz to avoid motion blur. However, simply increasing the frame rate by shortening the single-frame integration time leads to a decrease in detector signal gain and an increase in noise bandwidth, thereby reducing the image signal-to-noise ratio.

[0003] To balance high frame rate and signal-to-noise ratio, uncooled infrared detectors often employ a parallel readout architecture with multiple row integrator circuits to extend the effective integration time. However, while retaining common pixel-level response nonuniformity, this design also introduces row response nonuniformity with periodic characteristics, which manifests as "striped noise" in the image.

[0004] Existing technologies typically employ conventional non-uniformity correction methods, such as two-point correction, to suppress this type of horizontal stripe noise. However, these traditional correction methods have limited effectiveness in suppressing horizontal stripe noise in real-world scenarios, especially those with large temperature differences. Significant noise residue remains in the images, making it difficult to meet the requirements of high-precision infrared imaging. Summary of the Invention

[0005] To address the existing technical problems, this application provides a correction method for an infrared detector, an infrared imaging device, and a storage medium to improve the non-uniformity correction effect of images.

[0006] Firstly, a calibration method for an infrared detector is provided, the method comprising: Obtain the raw response values ​​of each pixel of the infrared detector; Based on the original response value and the boundaries of multiple response value intervals of the infrared detector pre-calibrated, the current response value interval of each pixel is determined, wherein multiple response value intervals are pre-calibrated based on the characteristics of the response curves of each integration channel of the infrared detector. Based on the pre-calibrated correction coefficients corresponding to each pixel in different response value intervals, obtain the correction coefficients corresponding to the current response value interval of each pixel; The original response value of each pixel is corrected according to the correction coefficient, and the corrected response value of each pixel is output.

[0007] In a second aspect, an infrared imaging device is provided, including an infrared detector, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the infrared detector calibration method described in the above embodiments.

[0008] Thirdly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the infrared detector calibration method described in the above embodiments.

[0009] The infrared detector calibration method provided in the above embodiments divides the full-range response of the detector into multiple response value intervals and calibrates the correction coefficients corresponding to each pixel in different response value intervals. The division of response value intervals is based on the characteristics of the response curves of each integral channel, thereby achieving a precise modeling of response differences under actual working conditions. During real-time calibration, the response value interval of a pixel is determined based on its current original response value, and the correction coefficients for that pixel within that interval are used for calibration. Because the division of different response value intervals is based on the actual shape of the response curve, and the calibration process fully considers the nonlinear characteristics under actual working conditions, the calibration application process can accurately match the true linear characteristics of the detector in each small segment of the response value interval. This breaks the limitations of traditional methods that rely on conventional temperature difference calibration and are based on the assumption of full-range linearity, reducing the systematic errors caused by fitting the entire nonlinear curve with a fixed linear model. This method effectively improves the non-uniformity correction effect of images, especially in scenarios with large temperature differences, and can significantly suppress horizontal stripe residue caused by nonlinear response.

[0010] The infrared imaging device and computer-readable storage medium provided in the above embodiments belong to the same concept as the corresponding infrared detector calibration method embodiments, and thus have the same technical effects as the corresponding infrared detector calibration method embodiments, which will not be repeated here. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of a dual-row integrator circuit architecture in one embodiment.

[0012] Figure 2 This is a flowchart of a calibration method for an infrared detector in one embodiment.

[0013] Figure 3 This is a flowchart illustrating a method for pre-calibrating the response value range and correction coefficients in one embodiment.

[0014] Figure 4 This is a flowchart illustrating a method for pre-calibrating the response value range and correction coefficients in another embodiment.

[0015] Figure 5 This is a flowchart of the steps for generating response curves for each integration channel of an infrared detector in one embodiment.

[0016] Figure 6 This is a schematic diagram of infrared data acquisition and processing in one embodiment.

[0017] Figure 7 This is the response curve of the detection unit corresponding to the four rows of integrator circuits in one embodiment.

[0018] Figure 8 This is a schematic diagram showing the sorting of the interval division points of the four-row integrator circuit in one embodiment.

[0019] Figure 9 This is a flowchart illustrating the steps for calibrating the correction coefficients of each pixel within different response value ranges in one embodiment.

[0020] Figure 10 This is a flowchart illustrating the steps of calibrating the correction coefficients for each pixel within different response value ranges in its respective integration channel, as described in one embodiment.

[0021] Figure 11 This is a flowchart of the steps in one embodiment to construct a uniform response value range and to calibrate the correction coefficients for each pixel within each response value range.

[0022] Figure 12 This is an infrared image obtained using the two-point correction method.

[0023] Figure 13 The image is an infrared image obtained using the correction method of this application. Detailed Implementation

[0024] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] In the following description, the phrase "some embodiments" refers to a subset of all possible embodiments. It should be noted that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0027] In the following description, the terms "first, second, and third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0028] Uncooled infrared detectors, which do not require cooling devices and can operate at room temperature, have advantages such as low cost, small size, and low power consumption, and have been widely used in the field of infrared imaging.

[0029] The integrating circuit is the core module of the infrared detector's readout circuit. Its function is to accumulate the weak photocurrent generated by the pixel after receiving infrared radiation and convert it into a voltage signal that is easy to measure and quantize later. Its operation consists of two stages: Integration Phase: When a pixel is selected, the weak current generated by the infrared radiation emitted by the pixel charges the integrating capacitor. As the charging time continues, the voltage across the integrating capacitor rises linearly, converting the accumulated charge into a voltage value. The duration of this process is called the integration time.

[0030] Readout and Reset Phase: After integration is complete, the voltage value on the integrating capacitor is sampled and sent to the subsequent circuit for reading. Then, the reset switch is closed, clearing the charge from the integrating capacitor and preparing for the next integration.

[0031] In a traditional line-by-line integration architecture, only one line of pixels is processed at a time. Assuming there are 288 lines of pixels, the reading time for each line is... So the time required to read all the lines of one frame is It is worth noting that in high-speed dynamic imaging scenarios, such as real-time fault detection in industrial production lines, low-altitude reconnaissance by drones, and sudden obstacle recognition in autonomous driving, the system frame rate typically needs to reach above 100Hz to avoid motion blur. With higher frame rate requirements, the integration time per line is forced to shorten, leading to a decrease in signal gain, an increase in noise bandwidth, and consequently a reduction in the image signal-to-noise ratio.

[0032] To balance high frame rate and signal-to-noise ratio, uncooled infrared detectors often employ a multi-row integrator circuit architecture, where multiple rows of pixels share an integrator circuit to achieve parallel readout. Figure 1Taking the dual-line integrator circuit shown as an example, the readout circuit is divided into two parts: odd and even lines. Channel A processes odd-numbered lines, and channel B processes even-numbered lines. Both channels can perform integration and readout simultaneously. This means that within one line period, two lines are being integrated simultaneously. Therefore, with the total frame period remaining constant, the integration time per line can theoretically be extended to nearly twice the integration time per line under a progressive integration architecture. In some embodiments, a four-line integrator circuit can also be used, with four independent integration channels, each processing line number... With a pixel size of 1000 pixels, the parallel operation of four channels theoretically allows the available integration time for each row to be up to 4 times that of line-by-line integration. Overall, for the same pixel array size and frame rate, the integration time per row of pixels in the N-row integration architecture is N times that of the line-by-line integration architecture.

[0033] However, while multi-line integration architecture improves integration time, it also introduces new non-uniformity issues. Due to slight manufacturing variations in the analog circuitry (such as integrating capacitors and amplifiers) between different channels, the response characteristics of each channel are not entirely consistent. This strictly periodic brightness difference caused by hardware variations manifests as periodic line response non-uniformity in the image, i.e., visual "stripes noise".

[0034] Existing technologies typically employ conventional non-uniformity correction methods, such as two-point calibration, to suppress this type of horizontal ripple noise. However, these traditional correction methods only calibrate parameters for the detector under typical operating conditions with normal temperature differences, failing to fully consider the actual response differences of the detector under large temperature variations. In practical applications with significant temperature differences, traditional correction methods have limited effectiveness in suppressing horizontal ripple noise, leaving noticeable noise residue in the image and failing to meet the requirements of high-precision infrared imaging.

[0035] Regarding this issue, such as Figure 2 As shown, this application provides a calibration method for an infrared detector, comprising: Step 202: Obtain the raw response values ​​of each pixel of the infrared detector.

[0036] During real-time imaging, the infrared detector converts the received target radiation into an electrical signal. After integration and analog-to-digital conversion, a digital image is output. This step yields raw image data without any correction, containing the direct response values ​​of all pixels to the current scene irradiance. ， The raw data retains the periodic line response characteristics caused by the hardware differences in the multi-line integrator circuit, i.e., ripple noise.

[0037] Step 204: Determine the current response value range of each pixel based on the original response value of the pixel and the boundaries of multiple response value ranges of the infrared detector that have been pre-calibrated. Multiple response value ranges are pre-calibrated based on the characteristics of the response curves of each integration channel of the infrared detector.

[0038] Specifically, the response values ​​of the pixels processed by each integration channel of the infrared detector are pre-processed to obtain the response curve of each integration channel. In one embodiment, the characteristic of the response curve, namely the inflection point on the response curve, can be determined by analyzing mathematical characteristics such as the rate of change of the slope or the curvature of the curve. Based on the inflection points of the response curves of each integration channel, multiple response value intervals are defined.

[0039] In practical applications, for each pixel, its original response value is... Each pixel is compared with the boundary of each response value interval to determine the current response value interval in which the pixel is located.

[0040] In this embodiment, by segmenting the full range of responses, including nonlinear ones, according to the characteristics of the response curve, the limitations of the linear assumption of traditional methods are broken, laying the foundation for applying different correction coefficients in each response value interval, thereby effectively suppressing horizontal stripe residue in scenarios with large temperature differences.

[0041] Step 206: Based on the pre-calibrated correction coefficients corresponding to each pixel in different response value intervals, obtain the correction coefficients corresponding to the current response value interval of each pixel.

[0042] Through pre-calibration, the correction coefficients corresponding to each pixel in different response value intervals were also determined. That is, in this embodiment, response value intervals were divided based on the characteristics of the response curve, and the correction coefficients for each pixel in different response value intervals were calibrated. Therefore, in the application stage, the response value interval in which the pixel is located is first determined, and then the correction coefficients to be applied to different pixels are determined based on the correction coefficients for that response value interval. Thus, for different pixels, if their current response value intervals are different, the correction coefficients applied will also be different. This allows for precise matching of the detector's actual linear characteristics within each small segment of the response value interval, reducing the systematic errors caused by fitting the entire nonlinear curve with a fixed linear model.

[0043] Step 208: Correct the original response value of each pixel according to the correction coefficient, and output the corrected response value of each pixel.

[0044] In this embodiment, for the full-frame image, the correction coefficient corresponding to each pixel is determined, and the original response value of each pixel is corrected using the correction coefficient to obtain the corrected response value of each pixel.

[0045] In this embodiment, pixel-level, segmented adaptive real-time computation ensures that whether the target in the scene is a low-temperature sky or a high-temperature object, it can be mapped to the correct response value range and corrected using coefficients that match the true response characteristics of that response value range.

[0046] The calibration method for this infrared detector divides the detector's full-range response into multiple response value intervals and calibrates the correction coefficients corresponding to each pixel within different intervals. The division of response value intervals is based on the characteristics of the response curves of each integration channel, thus achieving a precise modeling of response differences under actual operating conditions. During real-time calibration, the response value interval of a pixel is determined based on its current original response value, and the correction coefficients for that pixel within that interval are used for calibration. Because the division of different response value intervals is based on the actual shape of the response curve, and the calibration process fully considers the nonlinear characteristics under actual operating conditions, the calibration application process can accurately match the detector's true linear characteristics within each small segment of the response value interval. This overcomes the limitations of traditional methods that rely on conventional temperature difference calibration and the assumption of full-range linearity, reducing the systematic errors caused by fitting the entire nonlinear curve with a fixed linear model. This method effectively improves the non-uniformity correction effect of images, especially in scenarios with large temperature differences, and can significantly suppress horizontal stripe residue caused by nonlinear response.

[0047] In one embodiment, the method of pre-calibrating the response value range and correction coefficient is as follows: Figure 3 As shown, it includes the following steps Step 302: Obtain response value data of the infrared detector under multiple radiation conditions, and generate response curves for each integration channel of the infrared detector based on the response value data.

[0048] Step 304: Identify the characteristics of the response curves of each integration channel.

[0049] Step 306: Based on the characteristics of the response curves of each integration channel, calibrate multiple continuous response value ranges for the infrared detector.

[0050] Step 308: Calibrate and store the correction coefficients for each pixel in different response value ranges.

[0051] In this embodiment, multiple continuous response value intervals are pre-calibrated for the infrared detector based on the characteristics of the response curves of each integration channel. This allows the response curves of each integration channel to be normalized to the same reference curve, achieving non-uniformity correction based on response value segmentation. In application, the actual response value of a pixel is compared with the boundaries of the multiple response value intervals of the infrared detector to determine the current response value interval of the pixel. This method is relatively simple to implement in engineering; it only requires storing a set of response value interval boundaries for the infrared detector, and the intervals between response values ​​can be determined by comparing the response values.

[0052] In one embodiment, the multiple response value intervals of the infrared detector include multiple response value intervals corresponding to each integration channel of the infrared detector. In application, the row address of the pixel is obtained to determine the integration channel to which the pixel belongs; based on the pixel's original response value and the pre-defined boundaries of the multiple response value intervals of the integration channel to which the pixel belongs, the current response value interval of each pixel is determined.

[0053] Correspondingly, such as Figure 4 As shown, the method for pre-calibrating the response value range and correction coefficients includes the following steps: Step 402: Obtain response value data of the infrared detector under multiple radiation conditions, and generate response curves for each integration channel of the infrared detector based on the response value data.

[0054] Step 404: Identify the characteristics of the response curves of each integration channel.

[0055] Step 406: Based on the characteristics of the response curves of each integration channel of the infrared detector, calibrate multiple continuous response value intervals for each integration channel of the infrared detector.

[0056] Step 408: Calibrate and store the correction coefficients for each pixel within the response value range of its respective integration channel.

[0057] This approach, based on the characteristics of the response curve of each integration channel, individually sets the corresponding response value range for each integration channel, so that the division of the response value range fits the characteristics of the response curve of each integration channel, preserving the hardware characteristics of each integration channel, and theoretically can achieve higher correction accuracy.

[0058] In one embodiment, such as Figure 5 As shown, response data of the infrared detector under multiple radiation conditions are acquired, and response curves for each integration channel of the infrared detector are generated based on the response data, including: Step 502: Obtain the response value sequence of different integration channels of the infrared detector under multiple radiation conditions.

[0059] Specifically, such as Figure 6 As shown in the schematic diagram of infrared data acquisition and processing, during the calibration process, a target blackbody is used as the standard radiation source to provide uniform and controllable infrared radiation. By setting the blackbody to different temperatures (e.g., from low to high temperatures with fixed intervals), a known and stable irradiance input can be provided to the detector, thereby acquiring the detector's response data under different radiation conditions. The irradiance condition can be either temperature or irradiance.

[0060] The optical system (typically composed of infrared lenses) is responsible for collecting the infrared radiation emitted by the target blackbody and focusing it onto the photosensitive surface of an infrared detector array. The infrared detector array converts the received infrared radiation (light signal) into an electrical signal. This array consists of a large number of tiny detection units (pixels), each pixel responding independently to the incident irradiance.

[0061] The analog signals output from the detector array are typically small in amplitude and may contain noise, requiring amplification and filtering by the analog signal processing module. Subsequently, the analog-to-digital converter converts the conditioned analog signal into a digital signal and performs preliminary digital processing. The data acquisition unit is responsible for acquiring and outputting this raw digital image data for subsequent calibration or correction.

[0062] For the response data acquired by the infrared detector under different target blackbody radiation conditions (such as temperature / irradiation), these response data are split according to the integration channel to which the pixel belongs, based on the readout circuit architecture of the infrared detector. For example, for an N-row integration circuit architecture, it can be split into N independent subsets, each subset corresponding to a sequence of response values ​​for an integration channel under all radiation conditions.

[0063] Step 504: Generate the response curves for each integration channel based on the response value sequence of each integration channel.

[0064] For each integration channel, its average response under different radiation conditions is plotted on the ordinate, and the radiation condition (e.g., irradiation) on the abscissa to create a response curve for that channel. Due to the response characteristics of uncooled infrared detectors, this curve typically exhibits an S-shape. Figure 7 The response curves of the four integral channels shown indicate the presence of a nonlinear segment in the low irradiance region, an approximately linear segment in the middle, and a nonlinear segment in the high irradiance region.

[0065] In this embodiment, the analog circuits (such as integrating capacitors and amplifiers) of different integration channels have manufacturing deviations, and the shape of their response curves (such as the inflection point position and the slope of the linear region) will inevitably have slight differences. By generating the response curves of each integration channel separately based on the response value sequences of each integration channel under multiple radiation conditions, these differences can be preserved, preventing the true characteristics of each channel from being masked by data mixing and averaging, and providing a basis for subsequent accurate positioning of the linear interval of each channel.

[0066] In one embodiment, the characteristics of the response curve include inflection points. For the response curve of each integral channel, the point on the curve where the curvature changes most drastically is determined by analyzing its mathematical characteristics (e.g., calculating the second derivative to find the point of maximum curvature change, analyzing the curvature change pattern, or using curve fitting to identify points of curve shape change). These inflection points correspond to the boundary points where the response curve transitions from curvature to linearity or from linearity to curvature.

[0067] Determining the inflection point relies on the response data of each integration channel of the detector, representing a refined analysis of the response characteristics of each integration channel. By finding the inflection point of each channel itself, the effective linear interval of that integration channel can be accurately defined.

[0068] After obtaining the inflection points of all integration channels, the response value range is determined for the infrared detector.

[0069] In one embodiment, by combining the characteristics of the response curves of each integration channel of the infrared detector, the range of the detector's response value is divided into multiple continuous response value intervals. This allows the response curves of each integration channel within each interval to be normalized to the same reference curve, achieving non-uniformity correction based on response value segmentation. This method is relatively simple to implement in engineering; it only requires storing a set of response value interval boundaries, and the intervals between response values ​​can be determined by comparing the response values.

[0070] In one embodiment, based on the characteristics of the response curves of each integration channel of the infrared detector, a set of continuous response value intervals is defined for each integration channel.

[0071] In this embodiment, the detector's own integration channels are used as analysis units. By analyzing the inflection points of the response curves of each integration channel, the periodic row response non-uniformity introduced by the multi-row integration circuit is specifically analyzed, resulting in the curve characteristics of the response curves. Then, based on the response characteristics of each integration channel, the response value range corresponding to each integration channel of the infrared detector is determined. This process achieves personalized calibration of the infrared detector, thereby eliminating periodic horizontal ripple noise caused by hardware differences in the integration channels at its source.

[0072] In one embodiment, such as Figure 5 As shown, step 502 specifically includes: Step 5021: Obtain the response data of the infrared detector under different target blackbody radiation conditions.

[0073] Specifically, the target blackbody is set as a standard radiation source, and its temperature is controlled to gradually increase from a low temperature range (e.g., -40℃) to a high temperature range (e.g., +80℃), with multiple temperatures set at fixed intervals (e.g., every 10℃). At each temperature, after the blackbody temperature stabilizes, the infrared detector is activated to continuously acquire multiple frames of images.

[0074] Radiation conditions include temperature or irradiance. According to Planck's blackbody radiation law, under fixed optical system conditions, there is a definite correspondence between the blackbody temperature and the irradiance received by the detector image surface. Therefore, during calibration, the irradiance of the detector image surface can be uniquely determined by setting the blackbody temperature.

[0075] Taking radiation conditions, including temperature, as an example, since single-frame images may contain random noise, multiple original images acquired at the same temperature can be averaged in the time domain to suppress random noise and extract stable response values. The averaged image data serves as representative response data for that temperature. By iterating through all preset temperatures, a dataset of temperature-response values ​​can be obtained.

[0076] Step 5023: Based on the integration circuit architecture of the infrared detector, the response value data is split according to the integration channel to which the pixel belongs, and the response value data of each integration channel is obtained.

[0077] In this embodiment, the response value data of each integration channel is split according to the integration circuit architecture of the infrared detector. For example, if the infrared detector adopts a dual-row integration circuit architecture, the response value data is split into response value data of two integration channels. If the infrared detector adopts a four-row integration circuit architecture, the response value data is split into response value data of four integration channels.

[0078] Taking a four-line integrating circuit as an example, the detector's readout circuit contains four independent integrating channels: channel 1, channel 2, channel 3, and channel 4. Each channel is responsible for processing a set of pixels in a specific row (for example, channel 1 processes rows 1, 5, 9..., channel 2 processes rows 2, 6, 10..., and so on).

[0079] For the response data at each temperature obtained in step 5021, they are classified into their respective integration channels according to the row address of the pixels, forming four independent subsets.

[0080] The analog circuits (such as integrating capacitors and amplifiers) of different integration channels have slight deviations in their manufacturing process, resulting in differences in their response characteristics. By separating the channels, these differences are completely preserved in their respective datasets, preventing the true characteristics of each channel from being masked by data mixing and averaging. Since the root cause of horizontal ripple noise is the inconsistency in the response between different integration channels, this step splits the data by channel, allowing subsequent analysis to accurately correspond to each physical hardware path, providing a data foundation for reducing or eliminating periodic line noise.

[0081] Step 5025: Based on the radiation conditions of each target blackbody, calculate the average value of the response values ​​of all pixels in each integration channel to obtain the response value sequence of each integration channel under multiple radiation conditions.

[0082] For each points channel and each temperature Take the response values ​​of all pixels in the integration channel at this temperature and calculate their arithmetic mean: in, For channel The total number of pixels contained. For channel The Middle Each pixel in temperature The original response value.

[0083] By measuring all temperatures By repeating the above calculations, the channel can be obtained. The temperature-average response value sequence is obtained. Plotting this sequence on a coordinate system yields the response curve for the integral channel.

[0084] Taking the average value of all pixels within a channel effectively suppresses random noise and minor anomalies in individual pixels, highlighting the overall response trend inherent in the integration channel due to its hardware design. This allows subsequent curve analysis to focus on the hardware characteristics of the channel itself, rather than the random fluctuations of individual pixels.

[0085] In this embodiment, the response value sequence of each integration channel is generated based on the detector's own data and is calculated independently for each integration channel, which fully reflects the deep personalized calibration of the detector's hardware architecture and individual differences.

[0086] In one embodiment, identifying the characteristics of the response curves of each integration channel includes: By analyzing the mathematical characteristics of the response curves, the inflection points of the linear and nonlinear regions in the response curves of each integral channel can be identified. The mathematical characteristics include the curvature changes of the response curves, the rate of change of the first derivative, the extreme points of the second derivative, or the changes in the curve fitting residuals.

[0087] Within a large irradiation range, the response characteristics of the detector elements corresponding to different integrator circuits all exhibit varying degrees of nonlinearity, and are approximately as follows: Figure 7 The diagram shows a set of S-shaped curves. A significant characteristic is the obvious nonlinear curvature in the low and high irradiance regions, while an approximately linear region exists in the middle. The "inflection point" of the curve is precisely the boundary between the linear and nonlinear regions, possessing clear physical meaning and mathematical characteristics. From a mathematical perspective, the inflection point corresponds to the critical position where the degree of curvature of the curve changes significantly.

[0088] In one embodiment, the inflection point of the response curve of each integration channel can be determined based on the curvature change. Curvature is a direct measure of the degree of curvature bending. The greater the curvature, the more "bent" the curve. The inflection point is often where the curvature rapidly increases from a small value (small curvature in the linear region) (entering the nonlinear region) or rapidly decreases from a large value; that is, the point where the curvature change is most drastic.

[0089] In one embodiment, the inflection point of the response curve for each integration channel can be determined based on the rate of change of the first derivative. The first derivative (slope) represents the detector gain. In the linear region, the gain is essentially constant, and the rate of change of the first derivative is close to 0; in the nonlinear region, the gain changes rapidly, and the rate of change of the first derivative is relatively large. The inflection point corresponds to the critical point where the first derivative (gain) begins to change significantly from stability (linear region), or the critical point where it returns to stability from significant change, i.e., the extreme point of the rate of change of the first derivative.

[0090] In one embodiment, the inflection point of the response curve for each integral channel can be determined based on the extreme points of the second derivative. The second derivative describes the rate of change of the first derivative, that is, the speed of change of the slope, or the "acceleration" of the curve. In the linear region, the second derivative is close to 0 (the slope remains unchanged); in the nonlinear region, the second derivative is positive (the slope increases) or negative (the slope decreases). The inflection point is precisely the transition point where the second derivative increases significantly from 0 (or returns to 0 from a significant value), corresponding to the maximum or minimum value of the second derivative.

[0091] In one embodiment, the inflection point of the response curve for each integration channel can be determined based on the change in the curve fitting residual. Specifically, when fitting the curve piecewise using a linear model, the magnitude of the residual reflects the degree to which that segment deviates from linearity. When transitioning from the linear region to the nonlinear region, if a linear model is still used for fitting, the residual will suddenly increase. The inflection point is the critical position where the residual begins to increase significantly.

[0092] The methods described above essentially use mathematical tools to quantitatively describe the curvature characteristics of the curve and its changing position, thereby accurately locating the boundary between the linear and nonlinear regions, i.e., the inflection point.

[0093] The second derivative method will be used as an example for explanation.

[0094] For a continuous function Its second derivative This represents the rate of change of the curve's slope, i.e., the degree and direction of the curve's curvature. when When the slope increases, the curve bends upward. when At that time, the curve bends downward (the slope decreases). when When the slope of the curve changes at zero, it may be in the linear region or at an inflection point.

[0095] For S-shaped response curves, such as Figure 7 As shown: In low irradiance areas: the curve gradually becomes steeper from a flat (small slope) state, with the slope increasing. And as the bending intensifies, Gradually increase.

[0096] In the intermediate linear region: the slope is basically constant. .

[0097] In areas of high irradiance: the curve gradually flattens out from a steep slope, and the slope decreases. And as the bending intensifies, The absolute value gradually increases.

[0098] Suppose that for a certain integration channel, its position in... Average response value sequence at each temperature , .

[0099] Step 1: Calculate the first difference (approximate first derivative) First-order difference Reflects the temperature range The average rate of change of the internal response, i.e., the average gain of that interval.

[0100] Step 2: Calculate the second difference (approximate second derivative) Second-order difference This reflects the rate of change of gain, i.e., the curvature of the curve: A positive value indicates that the gain is increasing (the curve bends upward); A negative value indicates that the gain is decreasing (the curve bends downwards); The larger the absolute value, the more severe the bending.

[0101] Step 3: Locate the inflection point Find the lower inflection point (start of the linear region): Observe the second-order difference sequence in the low-temperature region (the first half of the sequence). .

[0102] The lower inflection point corresponds to the position where the second-order difference starts to decrease from a positive value and approaches 0, that is, the critical point where the curve transitions from "bending upward" (low-end nonlinearity) to "straight line" (linear region).

[0103] In practical discrete data, the temperature (or response value) corresponding to the maximum value point in the positive region of the second-order difference sequence is usually taken as the lower inflection point. This point marks the location where the gain increases the fastest, after which the gain begins to stabilize and the curve enters the linear region.

[0104] Find the upper inflection point (end of the linear region): In the high-temperature region (the latter half of the sequence), observe the second-order difference sequence. .

[0105] The upper inflection point corresponds to the position where the second-order difference starts to rise from the maximum absolute value of the negative value and approaches 0, that is, the critical point where the curve transitions from "straight line" (linear region) to "downward bending" (high-end nonlinearity).

[0106] In practical discrete data, the temperature (or response value) corresponding to the minimum point (i.e., the negative value with the largest absolute value) in the negative region of the second-order difference sequence is usually taken as the upper inflection point. This point marks the location where the gain decreases the fastest, after which the curve begins to enter the saturation region.

[0107] Step 4: Determine the response value corresponding to the inflection point The temperatures corresponding to the found lower and upper inflection points and Mapping back onto the response value curve yields the corresponding response value. and These two response values ​​are the lower inflection point and the upper inflection point of the integral channel.

[0108] In one embodiment, the inflection points of the response curve include a first inflection point where the response curve changes from a first nonlinear region to a linear region, and a second inflection point where the response curve changes from a linear region to a second nonlinear region. Based on the characteristics of the response curves of each integration channel, multiple continuous response value intervals are calibrated for the infrared detector, including: The first dividing point of the infrared detector's response value is determined based on the minimum value at the first inflection point of the response curve of each integration channel. The second dividing point of the infrared detector's response value is determined based on the minimum value among the second inflection points of the response curves of each integration channel. The response value range of the infrared detector is divided based on the first and second dividing points.

[0109] It should be understood that the minimum value among the first inflection points of the response curves of each integral channel refers to the minimum value among the first inflection points of the response curves of all integral channels. Similarly, the minimum value among the second inflection points of the response curves of each integral channel refers to the minimum value among the second inflection points of the response curves of all integral channels.

[0110] like Figure 7As shown, in low-irradiance regions (such as low-temperature sky backgrounds), the detector's response curve typically exhibits a slow change (small slope) and a certain degree of curvature. As irradiance increases, the curve gradually transitions to a linear region with a relatively constant slope. The first inflection point corresponds to the position where the second derivative drops from a large positive value to near 0 (i.e., one of the points where the curvature changes most drastically), or the critical point where the first derivative (gain) tends to stabilize from a rapid change.

[0111] In high-irradiance regions (such as high-temperature targets), the response curve gradually deviates from a straight line, entering a saturation trend, the slope begins to decrease, and the curve bends downwards. For example... Figure 7 As shown, the second inflection point corresponds to the position where the second derivative drops from 0 to the minimum negative value (i.e., the point where the curvature changes most drastically in the opposite direction), or the critical point where the first derivative (gain) starts to decrease significantly from stability.

[0112] By determining the first and second inflection points, the detector's entire response range is divided into three intervals with clear physical meaning: First nonlinear region (response value < first inflection point): low irradiance scene (such as sky, cold background); Linear region (response value between the first inflection point and the second inflection point): main working scenario; The second nonlinear region (response value > second inflection point): high irradiance scenarios (such as high temperature targets).

[0113] For N integration channels, the first inflection point of each channel was first obtained. (That is, the starting point of the linear region of each channel). These first inflection point values ​​are sorted, and the minimum value is taken as the first dividing point of the entire infrared detector response value. : Due to hardware differences, the starting point of the linear region varies among different integration channels. Some channels may have an earlier starting point (smaller first inflection point), while others may have a later starting point (larger first inflection point). To ensure that all channels are correctly classified outside the linear region in low-irradiance scenarios (i.e., to ensure that the response value range covers the channel whose linear region starts earliest), the minimum value among all first inflection points needs to be taken as a unified dividing point.

[0114] Observe carefully Figure 7 By analyzing the response curves and combining them with practical applications, it can be observed that low irradiance often corresponds to the sky in a scene and has a large curvature. Therefore, the first inflection point is taken as the first dividing point. The minimum value ensures the correction effect for low-irradiance scenes such as the sky. Furthermore, it ensures that after this dividing point, all channels have entered the linear region, thus providing a prerequisite for applying a unified correction model within the linear region.

[0115] Similarly, for N integration channels, first obtain the second inflection point of each channel. (That is, the end point of the linear region of each channel). The minimum value among them is taken as the second dividing point of the entire infrared detector response value. : Similar to the first inflection point, the endpoints of the linear regions differ across channels. Some channels may enter saturation earlier (smaller second inflection point), while others may enter later (larger second inflection point). To ensure that all channels remain in the linear region before reaching the second dividing point in high-irradiance scenarios, the minimum value among all second inflection points must be taken as the unified dividing point. Careful observation is needed. Figure 7 By analyzing the response curves and combining them with practical applications, it can be observed that ultra-high temperature targets in high-irradiance scenarios rarely appear over large areas and have significant curvature. Therefore, the second inflection point is taken as the second division point. The minimum value of this value ensures the correction effect in most application scenarios. Furthermore, it ensures that all channels remain in the linear region before this dividing point, thus guaranteeing the effectiveness of the linear correction model.

[0116] After determining the first dividing point Second dividing point Then, the entire response value range of the infrared detector can be divided into the following three consecutive response value intervals: First response value interval (low irradiance region / nonlinear region): Response value Second response value interval (middle linear region): response value Third response value range (high irradiance region / nonlinear region): Response value Taking an infrared detector with a four-line integrating circuit as an example, the inflection points of the response curves of each integrating channel, and the division points of a set of response value intervals determined based on the inflection points of the response curves of each integrating channel, are as follows: Figure 8 As shown, by establishing independent response model curves for each integrator circuit, and using the principle of minimizing the response value while considering the morphological characteristics of the multi-channel response curves, the problem is simplified within each interval to normalizing all pixels to the same response curve, ultimately achieving non-uniformity correction based on response value segmentation. This method ensures that all channels can be correctly corrected within the linear region. Although the non-linear regions at both ends are slightly wider, the multi-interval segmentation effectively suppresses horizontal stripes. Furthermore, it is simple to implement in engineering, requiring only the storage of a set of response value interval boundaries.

[0117] The ripple noise introduced by multi-row integrator circuits is fundamentally caused by the inconsistent response characteristics (including the linear region range) of different integration channels. In this embodiment, by taking the minimum value of the first inflection point of each integration channel at the first dividing point, it can be ensured that all channels are correctly classified into the nonlinear region before the start of the linear region, preventing misjudgment due to different channel start points; by taking the minimum value of the second inflection point of each integration channel at the second dividing point, it can be ensured that all channels are still in the linear region before the end of the linear region, preventing correction failure due to different channel end points.

[0118] The three response value intervals provided clear interval boundaries for subsequent steps (calibrating the correction coefficients of each pixel in different intervals).

[0119] In one embodiment, based on the characteristics of the response curves of each integration channel of the infrared detector, multiple continuous response value intervals are defined for each integration channel of the infrared detector, including: determining the boundary of the response value interval for each integration channel based on the inflection point of the response curve of each integration channel, thereby dividing multiple response value intervals for each integration channel.

[0120] In this embodiment, firstly, for each integration channel... ( By analyzing the mathematical characteristics of its response curve (such as the extreme points of the second derivative), two key inflection points of this integral channel were identified: First turning point The point where the integral channel response curve transitions from the low-irradiance nonlinear region to the linear region is the starting point of the linear region.

[0121] Second turning point The integral channel response curve is at the boundary point from the linear region to the high-irradiance nonlinear region, which is the end point of the linear region.

[0122] These two inflection points are calculated entirely based on the response data of the integration channel itself, reflecting the inherent hardware characteristics and response features of the integration channel.

[0123] For each points channel Using its two inflection points as boundaries, the response value range of this integral channel is divided into three consecutive response value intervals: First response value interval (low irradiance nonlinear region): Response value Second response value interval (middle linear region): response value Third response value interval (high irradiance nonlinear region): Response value Therefore, each integration channel has an independent and continuous range of response values ​​that fits its own response curve characteristics. For N integration channels, a total of N different sets of response value range boundaries are formed.

[0124] During the real-time correction phase, for each pixel, its response value range is determined by the following steps: 1. Determine the integration channel to which the pixel belongs based on the pixel's row address. .

[0125] 2. Retrieve the integration channel from storage. Corresponding interval boundary and .

[0126] 3. The original response value of the pixel Compare with the boundary of the integral channel: like If so, it is determined to be the first interval; like If so, it is determined to be the second interval; like If it is, then it is determined to be the third interval.

[0127] 4. Based on the judgment result, call the pre-calibrated correction coefficient of the pixel in the current response value range of its own integration channel for correction.

[0128] This method independently divides the interval for each integration channel, ensuring that the boundary of the response value interval is determined entirely by the inflection point of that integration channel itself, rather than by taking the minimum value after combining all channels. Therefore, the linear region of each integration channel is fully preserved. This personalized division can more accurately compensate for the response deviation of each channel, theoretically achieving higher correction accuracy.

[0129] In one embodiment, the correction factor includes a gain factor and a bias factor.

[0130] like Figure 9 As shown, the correction coefficients for each pixel are calibrated within different response value ranges, including: Step 902: For each of the response value intervals, select at least two radiation conditions corresponding to the response value interval, and calculate the global average value of all pixel response values ​​under each selected radiation condition as the expected pixel response value under that radiation condition.

[0131] Among the predefined intervals of response values, for the interval for which the coefficients need to be calibrated, select at least two radiation condition points corresponding to the interval from the multiple radiation conditions collected during the calibration phase.

[0132] The radiation condition point corresponding to this response value interval refers to the point where, under these radiation conditions, the pixel's response value falls within or near the interval's boundary. For example, for the intermediate linear interval... The response value can be selected to be close to The low temperature / irradiation point and response value are close to The high temperature / irradiation point; if the linearity within the interval is poor or higher accuracy is desired, multiple radiation conditions can also be selected.

[0133] For each selected radiation condition point, calculate the global average of the response values ​​of all pixels (regardless of the integration channel) at that radiation condition point: in, The total number of pixels, For pixels At radiation condition point The original response value. This average value. That is, the radiation condition point. The expected response value of the pixel.

[0134] The desired pixel response value represents the ideal output level of all pixels in the detector under given irradiance conditions. Using the global average as the correction target ensures consistent overall brightness in the corrected image while preserving the relative relationships between pixels. By employing a global average rather than an intra-channel average, a unified correction benchmark can be provided for all pixels, reducing non-uniformity caused by differences in desired values ​​between channels.

[0135] Step 904: Based on the actual response values ​​of each pixel under at least two radiation conditions and the corresponding expected response values ​​of the pixel, calculate the gain coefficient and bias coefficient of each pixel within the response value range to obtain the correction coefficient of each pixel within the response value range.

[0136] In one embodiment, a two-point method can be used, that is, two radiation condition points within each response value interval are selected, and the gain coefficient and bias coefficient are solved according to the following formula.

[0137] Solve the system of equations: in, Let i be the gain coefficient of pixel i in the response value interval j. is the bias coefficient of pixel i in the response value interval j.

[0138] In one embodiment, the least squares method can be used, and multiple radiation condition points can be selected for each response value interval to construct an error function. By minimizing the error, the gain coefficient and bias coefficient can be solved to obtain the correction coefficient with the best overall fit.

[0139] In this embodiment, for uncooled infrared detectors employing multi-row integrating circuits, the response curves of the integrating channels are obtained by independently modeling pixels sharing the same integrating circuit. The response value ranges are then divided based on the combined characteristics of the response curves from multiple integrating channels, and the correction coefficients are calculated after simplifying the model to a linear form within the response values. This method effectively improves the image non-uniformity correction effect and reduces the computational complexity of real-time correction, making it crucial for high frame rate applications.

[0140] In one embodiment, the correction coefficients for each pixel are calibrated within different response value ranges in its respective integration channel, such as... Figure 10 As shown, it includes: Step 1002: For each response value interval of each integration channel, select at least two radiation conditions corresponding to the response value interval. Based on each selected radiation condition, calculate the average value of the response values ​​of all pixels in the integration channel as the channel's expected response value under that radiation condition.

[0141] The implementation process of step 1002 is similar to that of step 902. The difference is that step 902 uses the global average of all pixels within the response value interval as the expected value, aiming to provide a unified correction benchmark for all pixels. Step 1002, on the other hand, uses the average of all pixels within the response value interval of a channel as the expected value, aiming to preserve the hardware characteristics of each integration channel and retain difference information for subsequent channel normalization.

[0142] Step 1004: Based on the actual response values ​​of each pixel under at least two radiation conditions and the corresponding channel expected response values ​​in different response value ranges of its respective integration channel, calculate the original gain coefficient and original bias coefficient of each pixel in different response value ranges of its respective integration channel.

[0143] The implementation process of step 1004 is similar to that of step 904. The difference is that step 904 calculates the universal correction coefficient for the entire detector based on the globally unified expected response value, without distinguishing the response characteristics of different integration channels; while step 1004 uses the interval expected response value within each integration channel as a benchmark to calculate the original gain and bias coefficients corresponding to each pixel channel by channel and interval by interval, which fully preserves the hardware response differences between different integration channels and provides channel-level correction parameters for subsequent channel-to-channel normalization processing and targeted elimination of periodic row non-uniformity caused by multi-row integration architecture.

[0144] Step 1006: Normalize and fit the response curves of each integral channel to obtain normalized curves, and calculate the correction coefficients of each integral channel based on the response curves and normalized curves of the integral channels respectively.

[0145] The normalization curve can be the average of the response curves of all integration channels. By normalizing the response curves of each channel to the same curve, and comparing the response curves of the integration channels with the normalized curve, the response curves of all channels are unified onto the same reference curve, reducing or eliminating the response differences between the integration channels. Furthermore, based on the normalized curve and the response curves of each integration channel, the correction coefficients for each integration channel are calculated, thereby reducing the periodic response deviation caused by hardware differences in the integration channels, and thus reducing or eliminating ripple noise.

[0146] In one embodiment, the correction coefficients for each integration channel can be calculated using either the two-point method or the least squares method, based on the response curves and normalized curves of the integration channels, respectively. In one embodiment, the correction coefficients include channel gain correction coefficients and channel bias correction coefficients.

[0147] The implementation process of step 1006 is similar to that of steps 904 and 1004. Both use the two-point method or the least squares method to fit and obtain the linear correction coefficients. The difference is that steps 904 and 1004 use pixels as the processing unit and solve the pixel-level correction coefficients based on the actual response value of a single pixel and the expected response value of the corresponding interval; while step 1006 uses the integral channel as the processing unit and solves the channel-level correction coefficients based on the average response curve of each integral channel and the global normalized reference curve. The aim is to align the response characteristics of different integral channels to a unified reference curve and specifically eliminate the periodic horizontal ripple noise caused by the multi-row integral architecture.

[0148] Step 1008: Determine the correction coefficient of each pixel in different response value intervals within its respective integration channel based on the original gain coefficient, original bias coefficient, and correction coefficient of the integration channel for each pixel in different response value intervals within its respective integration channel.

[0149] In one embodiment, the calculation formula is as follows: in, For pixels The original response value, For pixels Current response value range The original gain coefficient, For pixels Current response value range The original bias coefficient, For pixels The corresponding points channel Channel gain correction factor, For pixels The corresponding points channel Channel offset correction coefficient. This is the corrected output value.

[0150] Unlike the previous embodiment, this embodiment achieves two-stage correction by calibrating the correction coefficients of each pixel within the response value range of its respective integration channel. First, using data from each integration channel, calibration points are independently selected for each channel, and the expected value is calculated, ensuring that the first-stage coefficient calibration is entirely based on the response data of that integration channel itself. Unlike the globally uniform interval method which uses an average across the entire detector, this method uses an intra-channel average as the expected value. The expected value of each channel reflects the overall response level of that integration channel, fully preserving the hardware characteristics of each channel. The goal of the first-stage correction is to align the pixel response within each channel with the expected level of that channel, rather than aligning it with the average across the entire detector. By establishing a correlation between the actual response value of the pixel and the expected value of the channel itself, the solved original gain coefficient and original bias coefficient can accurately correct the pixel's response within that interval to the expected level of its respective channel.

[0151] Building upon this, a second-level correction, namely channel normalization correction, is applied to the average response curves of each integration channel. This involves normalizing and fitting the average response curves of each channel to a global average reference curve, and then calculating the gain correction coefficient and bias correction coefficient for each channel. This second-level correction unifies the average response curves of different channels onto the same reference curve, macroscopically reducing or eliminating periodic response deviations caused by hardware differences in the integration channels.

[0152] Finally, by combining the first-level original correction coefficients and the second-level channel correction coefficients, the final correction coefficients for each pixel within the response value range of its respective integration channel are obtained. This two-level correction structure preserves the hardware characteristics of each channel while eliminating response differences between channels, thereby achieving high pixel consistency within a channel while completely eliminating the periodic horizontal ripple noise introduced by the multi-row integration circuit.

[0153] In one embodiment, the process of constructing multiple consecutive response value intervals for the infrared detector and calibrating the correction coefficients corresponding to each pixel in different response value intervals is as follows: Figure 11 As shown, it includes the following steps: Step 1102: Obtain the response value sequence of different integration channels of the infrared detector under multiple radiation conditions.

[0154] Step 1104: Generate the response curves for each integration channel based on the response value sequence of each integration channel.

[0155] Specifically, the response value data of the infrared detector under different target blackbody radiation conditions are obtained; according to the integration circuit architecture of the infrared detector, the response value data is split according to the integration channel to which the pixel belongs, and the response value data of each integration channel is obtained; based on the blackbody radiation conditions of each target, the average value of the response value of all pixels in each integration channel is calculated to obtain the response value sequence of each integration channel under multiple radiation conditions.

[0156] Step 1106: Identify the inflection points of the response curves for each integration channel.

[0157] Specifically, by analyzing the mathematical characteristics of the response curve, the inflection points of the linear and nonlinear regions in the response curve of each integral channel are identified. The mathematical characteristics include the curvature change of the response curve, the rate of change of the first derivative, the extreme points of the second derivative, or the change of the curve fitting residual.

[0158] Step 1108: Based on the characteristics of the response curves of each integration channel, calibrate multiple continuous response value ranges for the infrared detector.

[0159] The inflection points of the response curve include the first inflection point when the response curve changes from the first nonlinear region to the linear region, and the second inflection point when the response curve changes from the linear region to the second nonlinear region.

[0160] Specifically, the first dividing point of the response value of the infrared detector is determined based on the minimum value of the first inflection point of the response curve of each integration channel; The second dividing point of the infrared detector's response value is determined based on the minimum value among the second inflection points of the response curves of each integration channel. The response value range of the infrared detector is divided based on the first and second dividing points.

[0161] Step 1110: For each response value interval, select at least two radiation conditions corresponding to the response value interval, calculate the global average value of all pixel response values ​​under each selected radiation condition, and use it as the expected response value of the pixel under that radiation condition.

[0162] Step 1112: Based on the actual response values ​​of each pixel under at least two radiation conditions and the corresponding expected response values ​​of the pixel, calculate the gain coefficient and bias coefficient of each pixel within the response value range to obtain the correction coefficient of each pixel within the response value range.

[0163] Based on the calibrated uniform response value range and the calibration coefficients of each pixel within the response value range, the application process is as follows: Figure 2As shown, by calibrating the correction coefficients for each pixel within a globally unified response value range, a correction coefficient table is generated. This table is then written into the storage module of the imaging system's non-uniformity correction system and loaded into the dynamic cache. When the imaging system receives an input signal, the non-uniformity correction module automatically matches the correction coefficients for the corresponding response value range based on the original response value of each pixel, thereby correcting the output signal of each pixel in real time.

[0164] The correction method for the infrared detector described in this application can eliminate the non-uniformity of line response caused by multi-line integrating circuits, thereby improving image quality. A comparison of similar scenes using an infrared detector with a four-line integrating circuit is provided. Figure 12 and Figure 13 As shown. Figure 12 For real-world footage using traditional, simple two-point correction, Figure 13 The images shown are actual photographs using the correction method described in this application. Comparison demonstrates that the infrared detector correction method of this application can effectively improve image quality.

[0165] In another aspect, this application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described infrared detector calibration method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0166] In another aspect, this application also provides an infrared imaging device, which includes an infrared detector, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the various processes of the above-described infrared detector calibration method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0167] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0168] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0169] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A calibration method for an infrared detector, characterized in that, The method includes: Obtain the raw response values ​​of each pixel of the infrared detector; Based on the original response value and the boundaries of multiple response value intervals of the infrared detector pre-calibrated, the current response value interval of each pixel is determined, wherein multiple response value intervals are pre-calibrated based on the characteristics of the response curves of each integration channel of the infrared detector. Based on the pre-calibrated correction coefficients corresponding to each pixel in different response value intervals, obtain the correction coefficients corresponding to the current response value interval of each pixel; The original response value of each pixel is corrected according to the correction coefficient, and the corrected response value of each pixel is output.

2. The calibration method for an infrared detector according to claim 1, characterized in that, The methods for pre-calibrating the response value range and the correction coefficient include: The response value data of the infrared detector under multiple radiation conditions are obtained, and the response curves of each integration channel of the infrared detector are generated based on the response value data. Identify the characteristics of the response curves of each integration channel; Based on the characteristics of the response curves of each integration channel, multiple consecutive response value ranges are calibrated for the infrared detector; The correction coefficients for each pixel are calibrated within different response value ranges and stored.

3. The calibration method for an infrared detector according to claim 1, characterized in that, The multiple response value ranges of the infrared detector include multiple response value ranges corresponding to each integration channel of the infrared detector. The step of determining the current response value interval of each pixel based on the original response value and the pre-defined boundaries of multiple response value intervals of the infrared detector includes: Obtain the row address of the cell and determine the integration channel to which the cell belongs; Based on the original response value of the pixel and the boundaries of multiple response value intervals of the integration channel to which the pixel belongs, the current response value interval of each pixel is determined.

4. The calibration method for an infrared detector according to claim 3, characterized in that, The methods for pre-calibrating the response value range and the correction coefficient include: The response value data of the infrared detector under multiple radiation conditions are obtained, and the response curves of each integration channel of the infrared detector are generated based on the response value data. Identify the characteristics of the response curves of each integration channel; Based on the characteristics of the response curves of each integration channel of the infrared detector, multiple continuous response value ranges are calibrated for each integration channel of the infrared detector. The correction coefficients for each pixel within the specified response value range of its respective integral channel are calibrated and stored.

5. The calibration method for the infrared detector according to claim 2 or 4, characterized in that, The step of acquiring the response value data of the infrared detector under multiple radiation conditions and generating the response curves of each integration channel of the infrared detector based on the response value data includes: Obtain the response value sequence of different integration channels of the infrared detector under multiple radiation conditions; Based on the response value sequence of each integration channel, the response curve of each integration channel is generated.

6. The calibration method for an infrared detector according to claim 5, characterized in that, The step of obtaining the response value sequence of different integration channels of the infrared detector under multiple radiation conditions includes: Acquire the response data of the infrared detector under different target blackbody radiation conditions; Based on the integration circuit architecture of the infrared detector, the response value data is split according to the integration channel to which the pixel belongs, to obtain the response value data of each integration channel. Based on the blackbody radiation conditions of each target, the average response value of all pixels in each integration channel is calculated to obtain the response value sequence of each integration channel under multiple radiation conditions.

7. The calibration method for an infrared detector according to claim 2 or 4, characterized in that, The characteristics of the response curve include the inflection point of the response curve; The features for identifying the response curves of each integration channel include: By analyzing the mathematical characteristics of the response curve, the inflection points of the linear and nonlinear regions in the response curve of each integral channel are identified. The mathematical characteristics include the curvature change of the response curve, the rate of change of the first derivative, the extreme points of the second derivative, or the change of the curve fitting residual.

8. The calibration method for an infrared detector according to claim 2, characterized in that, The characteristics of the response curve include the inflection points of the response curve; the inflection points of the response curve include the first inflection point where the response curve changes from the first nonlinear region to the linear region, and the second inflection point where the response curve changes from the linear region to the second nonlinear region. The process of calibrating multiple continuous response value ranges for the infrared detector based on the characteristics of the response curves of each integration channel includes: The first dividing point of the response value of the infrared detector is determined based on the minimum value among the first inflection points of the response curves of each integration channel. The second dividing point of the response value of the infrared detector is determined based on the minimum value of the second inflection point of the response curve of each integration channel; The response value range of the infrared detector is divided according to the first dividing point and the second dividing point.

9. The calibration method for an infrared detector according to claim 4, characterized in that, The characteristics of the response curve include the inflection point of the response curve; Based on the characteristics of the response curves of each integration channel of the infrared detector, multiple continuous response value intervals are calibrated for each integration channel of the infrared detector, including: Based on the inflection points of the response curves of each integration channel, the boundaries of the response value intervals for each integration channel are determined, thus dividing each integration channel into multiple response value intervals.

10. The calibration method for an infrared detector according to claim 2 or 8, characterized in that, The correction coefficients include gain coefficients and bias coefficients; The calibration of each pixel within different response value ranges includes: For each of the response value intervals, at least two radiation conditions corresponding to the response value intervals are selected, and the global average value of all pixel response values ​​under each selected radiation condition is calculated as the expected pixel response value under that radiation condition. Based on the actual response values ​​of each pixel under at least two radiation conditions and the corresponding expected response values ​​of the pixel, the gain coefficient and bias coefficient of each pixel within the response value range are calculated to obtain the correction coefficient of each pixel within the response value range.

11. The calibration method for an infrared detector according to claim 4 or 9, characterized in that, The correction coefficients include gain coefficients and bias coefficients; The correction coefficients for each pixel within the response value range of its respective integral channel include: For each response value range of each integration channel, at least two radiation conditions corresponding to the response value range are selected. Based on each selected radiation condition, the average value of the response values ​​of all pixels in the integration channel is calculated as the channel's expected response value under that radiation condition. Based on the actual response values ​​of each pixel under at least two radiation conditions and the corresponding expected response values ​​of the channel in different response value intervals of its respective integration channel, calculate the original gain coefficient and original bias coefficient of each pixel in different response value intervals of its respective integration channel. The response curves of each integral channel are normalized and fitted to obtain normalized curves. The correction coefficients of each integral channel are calculated based on the response curves of the integral channels and the normalized curves. Based on the original gain coefficient, the original bias coefficient, and the correction coefficient of the integration channel for each pixel in different response value intervals within its respective integration channel, the correction coefficient for each pixel in different response value intervals within its respective integration channel is determined.

12. An infrared imaging device, characterized in that, The device includes an infrared detector, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the calibration method for the infrared detector as described in any one of claims 1 to 11.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the calibration method for the infrared detector as described in any one of claims 1 to 11.

Images