CN102779332A - Nonlinear-fitting infrared non-uniform correction method based on time-domain Kalman filtering correction - Google Patents

技术领域 technical field

本发明属于红外图像预处理技术领域，涉及一种对红外图像进行非均匀校正的方法。The invention belongs to the technical field of infrared image preprocessing, and relates to a method for non-uniform correction of infrared images.

背景技术 Background technique

红外图像非均匀校正是图像处理领域里的一个重要研究方向。红外图像非均匀校正的主要目的是利用图像处理手段，解决由于材料缺陷、电路的稳定性以及集成工艺水平的限制，导致的探测器输出响应不一致的问题，使校正后的图像便于后续处理。目前，红外图像非均匀校正的主要方法可分为两类：基于参考源的温度定标校正法和基于场景的自适应校正法。这两类方法的基本原理和技术特点如下：Non-uniform correction of infrared images is an important research direction in the field of image processing. The main purpose of infrared image non-uniformity correction is to use image processing methods to solve the problem of inconsistent detector output response due to material defects, circuit stability and integration process level limitations, so that the corrected image is convenient for subsequent processing. At present, the main methods of infrared image non-uniformity correction can be divided into two categories: temperature calibration correction method based on reference source and adaptive correction method based on scene. The basic principles and technical characteristics of these two types of methods are as follows:

一、基于参考源的温度定标校正方法1. Temperature calibration and correction method based on reference source

基于参考源的温度定标校正方法主要包括单点温度定标法，两点温度定标法和多点温度定标法等。这类方法的设计思想是：利用参考辐射源给红外焦平面阵列提供均匀辐照度，对每个探测器单元的响应输出进行测量，由此计算得出个探测器单元的校正参数。The temperature calibration and correction methods based on the reference source mainly include single-point temperature calibration method, two-point temperature calibration method and multi-point temperature calibration method. The design idea of this type of method is: use a reference radiation source to provide uniform irradiance to the infrared focal plane array, measure the response output of each detector unit, and calculate the correction parameters of each detector unit.

二、基于场景的自适应校正法2. Scene-based adaptive correction method

基于场景的自适应校正方法主要包括时域高通滤波法、神经网络校正法、常量统计法、线性滤波校正法和场景匹配法等。这类算法的基本原理是计算增益系数和偏移量的数据不是取自参考辐射源，而是全部或部分来自于场景的估计。The scene-based adaptive correction methods mainly include time-domain high-pass filter method, neural network correction method, constant statistics method, linear filter correction method and scene matching method, etc. The basic principle of this type of algorithm is that the data for calculating the gain coefficient and offset are not taken from the reference radiation source, but all or part of it comes from the estimation of the scene.

非均匀性是红外焦平面探测器的固有属性，在理想情况下，红外焦平面面阵探测器受均匀入射辐射时，其各个像元的信号输出值应该完全一致；但实际上在制作器件的半导体材料不均匀（杂质浓度、晶体缺陷、内部结构的不均匀性等）、器件工作状态、生产工艺过程以及外界输入等的综合影响下，其输出幅度并不相同，这就是所谓的红外图像非均匀性（Non-Uniformity，NU）。对于单点扫描方式的探测器来说不存在非均匀性问题，线阵扫描方式的探测器中的非均匀性存在于线阵方向，而红外焦平面探测器的非均匀性存在于整个焦平面上，愈是大规模的器件，非均匀性问题就愈突出。这种非均匀性会导致系统的温度分辨率下降，使目标图像的质量受到严重影响，从而限制了其在高灵敏度检测方面的应用。Non-uniformity is an inherent property of infrared focal plane detectors. Ideally, when infrared focal plane area detectors are subjected to uniform incident radiation, the signal output values of each pixel should be completely consistent; Under the comprehensive influence of semiconductor material inhomogeneity (impurity concentration, crystal defect, internal structure inhomogeneity, etc.), device working state, production process and external input, the output amplitude is not the same, which is the so-called infrared image non-uniformity. Uniformity (Non-Uniformity, NU). There is no non-uniformity problem for the single-point scanning detector, the non-uniformity of the linear scanning detector exists in the line array direction, and the non-uniformity of the infrared focal plane detector exists in the entire focal plane In general, the larger the scale of the device, the more prominent the problem of non-uniformity. This non-uniformity will lead to a decrease in the temperature resolution of the system, seriously affecting the quality of the target image, thus limiting its application in high-sensitivity detection.

现有的图像校正方法在解决红外图像非均匀性问题时存在不足，主要表现在以下几个方面：The existing image correction methods have deficiencies in solving the problem of infrared image non-uniformity, mainly in the following aspects:

基于参考源的温度定标校正方法：校正参数是固定不变的，而实际上，随着器件工作温度和环境温度的变化，器件的工作状态会发生变化，如果还采用原先计算出的校正参数进行校正，就会使校正效果变差；基于场景的自适应校正法：便于处理存在运动目标的红外图像，而且计算量大，实时处理需要先进的多处理器机构。Temperature calibration correction method based on reference source: the correction parameters are fixed, but in fact, with the change of device operating temperature and ambient temperature, the working state of the device will change, if the original calculated correction parameters are still used Correction will make the correction effect worse; scene-based adaptive correction method: it is convenient to process infrared images with moving targets, and has a large amount of calculation, and real-time processing requires advanced multi-processor mechanisms.

综上所述，针对红外图像非均匀性问题设计适于工程实现的、适用图像范围广的非均匀校正方法是一个急需解决的工程技术问题。目前尚未发现有关这个问题的公开研究资料。To sum up, it is an urgent engineering technical problem to design a non-uniformity correction method suitable for engineering realization and applicable to a wide range of images for the problem of infrared image non-uniformity. No published research on this issue has been found.

发明内容 Contents of the invention

本发明的目的是提供一种红外图像非均匀校正方法，解决红外焦平面探测器输出响应不一致所导致的图像降质问题。The purpose of the present invention is to provide a non-uniform infrared image correction method to solve the problem of image degradation caused by inconsistent output responses of infrared focal plane detectors.

技术方案是首先假设红外焦平面探测器各阵列单元的响应变化曲线在时间上是连续的，采用高阶多参数非线性多项式可对其进行拟合和描述；然后采集四个不同温度点各阵列单元的响应输出值，利用非线性方程求解方法确定各阵列单元响应表达式中的多个参数；第三步为了解决探测器响应随时间漂移问题，利用时域Kalman滤波对响应表达式中各参数进行时漂修正，得到探测器阵列单元响应曲线的解析表示；最后利用该解析表示计算任意时间任意温度条件下探测器各阵列单元响应输出，从而实现对探测器响应不一致的校正，解决红外图像非均匀性问题。The technical solution is to first assume that the response curve of each array unit of the infrared focal plane detector is continuous in time, and it can be fitted and described by using a high-order multi-parameter nonlinear polynomial; For the response output value of the unit, use the nonlinear equation solving method to determine multiple parameters in the response expression of each array unit; in the third step, in order to solve the problem of detector response drifting with time, use the time domain Kalman filter to analyze each parameter in the response expression Time drift correction is carried out to obtain the analytical expression of the response curve of the detector array unit; finally, the analytical expression is used to calculate the response output of each array unit of the detector at any time and at any temperature, so as to realize the correction of the inconsistency of the detector response and solve the problem of infrared image irregularities. uniformity problem.

本发明的技术方案包括以下步骤：Technical scheme of the present invention comprises the following steps:

第一步：参数估计Step 1: Parameter Estimation

在红外探测器常规工作温度10℃到70℃之间，任意选择未经校正的四个不同温度点T₁,T₂,T₃,T₄的红外图像，f_i,j(T₁),f_i,j(T₂),f_i,j(T₃),f_i,j(T₄)分别为这四幅红外图像像素点的灰度值，其中下标i,j表示像素点行和列，将其分别代入公式一，求解探测器响应方程：The infrared images of four different temperature points T ₁ , T ₂ , T ₃ , and T ₄ are arbitrarily selected between the normal operating temperature of the infrared detector between 10°C and 70°C, f _i,j (T ₁ ), f _i,j (T ₂ ), f _i,j (T ₃ ), and f _i,j (T ₄ ) are the gray values of the pixels in the four infrared images respectively, where the subscripts i,j represent the pixel row and Columns, respectively substituting them into formula 1 to solve the detector response equation:

F ‾ ( T 1 ) = A i , j · ( f i , j ( T 1 ) ) 3 + B i , j · ( f i , j ( T 1 ) ) 2 + C i , j · ( f i , j ( T 1 ) ) + D ij F ‾ ( T 2 ) = A i , j · ( f i , j ( T 2 ) ) 3 + B i , j · ( f i , j ( T 2 ) ) 2 + C i , j · ( f i , j ( T 2 ) ) + D ij F ‾ ( T 3 ) = A i , j · ( f i , j ( T 3 ) ) 3 + B i , j · ( f i , j ( T 3 ) ) 2 + C i , j · ( f i , j ( T 3 ) ) + D ij F ‾ ( T 4 ) = A i , j · ( f i , j ( T 4 ) ) 3 + B i , j · ( f i , j ( T 4 ) ) 2 + C i , j ( f i , j ( T 4 ) ) + D ij （公式一） f ‾ ( T 1 ) = A i , j · ( f i , j ( T 1 ) ) 3 + B i , j &Center Dot; ( f i , j ( T 1 ) ) 2 + C i , j &Center Dot; ( f i , j ( T 1 ) ) + D. ij f ‾ ( T 2 ) = A i , j · ( f i , j ( T 2 ) ) 3 + B i , j &Center Dot; ( f i , j ( T 2 ) ) 2 + C i , j &Center Dot; ( f i , j ( T 2 ) ) + D. ij f ‾ ( T 3 ) = A i , j &Center Dot; ( f i , j ( T 3 ) ) 3 + B i , j · ( f i , j ( T 3 ) ) 2 + C i , j · ( f i , j ( T 3 ) ) + D. ij f ‾ ( T 4 ) = A i , j &Center Dot; ( f i , j ( T 4 ) ) 3 + B i , j &Center Dot; ( f i , j ( T 4 ) ) 2 + C i , j ( f i , j ( T 4 ) ) + D. ij (Formula 1)

其中，

第二步：实现校正Step Two: Implement Correction

将任意温度点T下需校正的红外图像像素点的灰度值g_i,j(T)代入公式二，得到校正后红外图像像素点的灰度值

g ^ i , j ( T ) = A ^ i , j · ( g i , j ( T ) ) 3 + · B ^ i , j ( g i , j ( T ) ) 2 + C ^ i , j · ( g i , j ( T ) ) + D ^ ij （公式二） g ^ i , j ( T ) = A ^ i , j &Center Dot; ( g i , j ( T ) ) 3 + &Center Dot; B ^ i , j ( g i , j ( T ) ) 2 + C ^ i , j · ( g i , j ( T ) ) + D. ^ ij (Formula 2)

第三步：时域Kalman滤波修正Step 3: Time-domain Kalman filter correction

校正后的红外探测器在工作一段时间后，会出现焦平面阵列输出响应随工作时长产生波动的现象，即发生非均匀性时漂，时漂的出现严重影响探测器的工作性能，因此需要对校正后红外图像像素点的灰度值

第1步：依据图像NU(T)（Non-uniformity，非均匀度）值确定是否进行时漂处理，并确定α_k和β_k两个时漂因子值，NU(T)值计算方法如公式三所示：Step 1: Determine whether to perform time drift processing according to the NU(T) (Non-uniformity) value of the image, and determine the two time drift factor values of α _k and β _k . The calculation method of NU(T) value is as follows: Three shown:

NU ( T ) = Σ i , j | g ^ i , j ( T ) - G ‾ ( T ) | N G ‾ ( T ) （公式三） NU ( T ) = Σ i , j | g ^ i , j ( T ) - G ‾ ( T ) | N G ‾ ( T ) (Formula 3)

其中：

当NU(T)≤3.0‰时不需要进行时漂处理，则技术方案结束；当NU(T)＞3.0‰，取α_k=α∈[0.99,1]，β_k=β∈[0.99,1]，并进行时漂处理；当NU(T)＞4.0‰，则令α_k=α∈[0.9,0.99]，β_k=β∈[0.9,0.99]，并进行时漂处理；如果时漂更大，即当NU(T)＞4.5‰，令α_k=α∈[0.8,0.9]，β_k=β∈[0.8,0.9]，并进行时漂处理。When NU(T)≤3.0‰, there is no need for time drift treatment, the technical solution ends; when NU(T)＞3.0‰, take α _k =α∈[0.99,1], β _k =β∈[0.99, 1], and perform time drift processing; when NU(T)＞4.0‰, then set α _k =α∈[0.9,0.99], β _k =β∈[0.9,0.99], and perform time drift processing; if The drift is larger, that is, when NU(T)>4.5‰, set α _k = α∈[0.8,0.9], β _k =β∈[0.8,0.9], and perform time drift processing.

时漂处理是指下述第2步至第4步：Time drift treatment refers to the following steps 2 to 4:

第2步：建立Kalman滤波的状态方程和观测方程：Step 2: Establish the state equation and observation equation of the Kalman filter:

状态方程：X_i,j(k+1)＝Φ_kX_i,j(k)+M_k+W_k    （公式四）Equation of state: X _i,j (k+1)＝Φ _k X _i,j (k)+M _k +W _k (Formula 4)

观测方程：Y_i,j(k)＝H_kX_i,j(k)+V_k    （公式五）Observation equation: Y _i,j (k)=H _k X _i,j (k)+V _k (Formula 5)

其中，k表示当前时刻，k+1表示下一时刻，k＝0,1,2,…，并且k＝0对应于得到需要进行时漂处理的红外图像的时刻T；状态向量X_i,j(k)定义为

Q k = ( 1 - α k 2 ) σ α 0 2 0 0 ( 1 - β k 2 ) σ β 0 2 , R k = Iσ vk 2 . (公式六) Q k = ( 1 - α k 2 ) σ α 0 2 0 0 ( 1 - β k 2 ) σ β 0 2 , R k = Iσ vk 2 . (Formula 6)

根据实践经验取 σ α 0 2 ∈ [ 0.2,0.25 ] , σ β 0 2 ∈ [ 0.05,0.1 ] , σ vk 2 ∈ [ 0.05,0.1 ] . based on practical experience σ α 0 2 ∈ [ 0.2,0.25 ] , σ β 0 2 ∈ [ 0.05,0.1 ] , σ vk 2 ∈ [ 0.05,0.1 ] .

第3步：将α_k和β_k的取值代入Kalman滤波的状态方程和观测方程，迭代得到最终的滤波估计值

第4步：实现时漂修正Step 4: Implement Time Drift Correction

将

g ^ ^ i , j ( T ) = ( 1 - α k ) · A ^ ^ i , j ( k ) · G ‾ ( T ) + ( 1 - β k ) · B ^ ^ i , j ( k ) (公式七) g ^ ^ i , j ( T ) = ( 1 - α k ) &Center Dot; A ^ ^ i , j ( k ) &Center Dot; G ‾ ( T ) + ( 1 - β k ) · B ^ ^ i , j ( k ) (Formula 7)

采用本发明可以取得以下技术效果：Adopt the present invention can obtain following technical effect:

本发明能够稳定可靠地实现红外图像非均匀性校正，明显改善由于探测器各像元响应的不一致所导致的图像非均匀性噪声，提高系统温度分辨率、提升图像预处理质量，为后续的图像检测、识别、跟踪提供良好数据源。本发明所提出的红外图像非均匀校正方法，具有以下比较明显的特点和优势：The invention can stably and reliably realize infrared image non-uniformity correction, significantly improve the image non-uniformity noise caused by the inconsistency of the response of each pixel of the detector, improve the temperature resolution of the system, improve the quality of image preprocessing, and improve the quality of subsequent images. Detection, identification, and tracking provide good data sources. The infrared image non-uniform correction method proposed by the present invention has the following obvious characteristics and advantages:

1.根据本发明的实验结果以及和其它常用红外图像非均匀校正方法的性能对比结果可见：本发明能够有效地克服非均匀噪声的干扰，具有简单易行、准确率高、鲁棒性强、易于FPGA（Field Programmable Gate Array，现场可编程门阵列）；1. According to the experimental results of the present invention and the performance comparison results with other common infrared image non-uniform correction methods, it can be seen that the present invention can effectively overcome the interference of non-uniform noise, and has the advantages of simplicity, high accuracy, strong robustness, Easy to FPGA (Field Programmable Gate Array, Field Programmable Gate Array);

2.依据本发明第一步和第二步，可很好地表征红外探测器输出响应随温度的非线性变化规律，解决了传统校正方法难于处理接近温度饱和区域的红外图像校正问题；2. According to the first step and the second step of the present invention, the nonlinear change rule of the infrared detector output response with temperature can be well characterized, and the traditional correction method is difficult to deal with the infrared image correction problem near the temperature saturation region;

3.依据本发明第三步“时域Kalman滤波修正”，可动态完成探测器响应时漂修正，解决了校正后的红外探测器在工作一段时间后，阵列输出响应随工作时长产生波动的问题，极大提高了校正方法的场景适应性。3. According to the third step "time-domain Kalman filter correction" of the present invention, the time-drift correction of the detector response can be dynamically completed, which solves the problem that the array output response fluctuates with the working time after the corrected infrared detector works for a period of time , which greatly improves the scene adaptability of the correction method.

附图说明 Description of drawings

图1是本发明的原理流程图；Fig. 1 is a principle flow chart of the present invention;

图2是20℃温度点时利用本发明进行的仿真实验结果；Fig. 2 is the simulation experiment result that utilizes the present invention to carry out when 20 ℃ of temperature points;

图3是50℃温度点时利用本发明进行的仿真实验结果；Fig. 3 is the simulation experiment result that utilizes the present invention to carry out when 50 ℃ of temperature points;

图4是利用本发明和两点法进行对比实验的结果；Fig. 4 is the result that utilizes the present invention and two-point method to carry out comparative experiment;

图5是利用本发明和稳态卡尔曼滤波法进行对比实验的结果。Fig. 5 is the result of a comparative experiment using the present invention and the steady-state Kalman filtering method.

具体实施方式 Detailed ways

下面结合附图对本发明做进一步说明。The present invention will be further described below in conjunction with the accompanying drawings.

图2至图5进行实验的结果，横坐标均为图像帧号，纵坐标为不同的图像非均匀性描述参数。实验中利用本发明的第一步进行参数估计时，选择四个不同温度点T₁,T₂,T₃,T₄分别是15℃、25℃、40℃、55℃。对温度点T下需校正的红外图像进行处理，温度T的取值分别是20℃、50℃和40℃。The results of experiments in Fig. 2 to Fig. 5, the abscissa is the image frame number, and the ordinate is different image non-uniformity description parameters. In the experiment, when using the first step of the present invention to estimate parameters, four different temperature points T ₁ , T ₂ , T ₃ , and T ₄ were selected to be 15°C, 25°C, 40°C, and 55°C, respectively. The infrared images to be corrected at the temperature point T are processed, and the values of the temperature T are 20°C, 50°C and 40°C respectively.

图2（a）为20℃温度点红外图像序列灰度均方差变化曲线图，左图为校正前结果，右图为校正后结果；图2（b）为20℃温度点红外图像序列非均匀度变化曲线图，左图为校正前结果，右图为校正后结果；从实验结果可以看出，本发明对20℃温度点红外图像的非均匀性具有明显的校正效果，图像灰度均方差由校正前的171左右降低到校正后的1.4左右，图像非均匀度由校正前的9.5‰左右，降低到校正后的0.14‰左右，结果均证明了本发明的有效性。Figure 2(a) is the change curve of gray mean square error of infrared image sequence at 20°C temperature point, the left figure is the result before correction, and the right figure is the result after correction; Figure 2(b) is the non-uniform infrared image sequence at 20°C temperature point Temperature change curve, the left picture is the result before correction, and the right picture is the result after correction; from the experimental results, it can be seen that the present invention has a significant correction effect on the non-uniformity of the infrared image at the temperature point of 20 °C, and the mean square error of the image gray level It is reduced from about 171 before correction to about 1.4 after correction, and the image non-uniformity is reduced from about 9.5‰ before correction to about 0.14‰ after correction. The results all prove the effectiveness of the present invention.

图3（a）为50℃温度点红外图像序列灰度均方差变化曲线图，左图为校正前结果，右图为校正后结果；图3（b）为50℃温度点红外图像序列非均匀度变化曲线图，左图为校正前结果，右图为校正后结果；从实验结果可以看出，本发明对50℃温度点红外图像的非均匀性具有明显的校正效果，图像灰度均方差由校正前的384-385降低到校正后的1.3-2.0，图像非均匀度由校正前的13.08‰-13.10‰，降低到校正后的0.6‰-0.10‰，结果均证明了本发明的有效性。Figure 3(a) is the change curve of the gray mean square error of the infrared image sequence at 50°C, the left figure is the result before correction, and the right figure is the result after correction; Figure 3(b) is the non-uniform infrared image sequence at 50°C Temperature change curve, the left picture is the result before correction, and the right picture is the result after correction; it can be seen from the experimental results that the present invention has obvious correction effect on the non-uniformity of the infrared image at the temperature point of 50 ℃, and the mean square error of the image gray level Reduced from 384-385 before correction to 1.3-2.0 after correction, image non-uniformity from 13.08‰-13.10‰ before correction to 0.6‰-0.10‰ after correction, the results all prove the effectiveness of the present invention .

图4为针对50℃温度点红外图像序列，本发明与两点定标校正法的校正效果对比图，本发明对应曲线用“----”表示，两点定标校正法对应曲线用“——”表示；图4（a）为灰度均方差变化曲线图，左图为校正前对比结果，右图为校正后对比结果；图4（b）为非均匀度变化曲线图，左图为校正前对比结果，右图为校正后对比结果；从实验结果可以看出，与在工程实践过程中广泛用到的两点定标校正法比较，本发明的校正效果明显优于两点定标校正法，图像灰度均方差和非均匀度两个图像非均匀性描述参数明显小于两点定标法相应的参数指标，特别是在接近温度饱和的情况下，由于本发明采用了非线性拟合方法实现对探测器响应参数的估计，避免了线性拟合方法带来的失真问题，保证了校正效果的鲁棒性。Fig. 4 is a comparison chart of the correction effect between the present invention and the two-point calibration correction method for the infrared image sequence at a temperature point of 50°C. The corresponding curve of the present invention is represented by "----", and the corresponding curve of the two-point calibration correction method is represented by " ——" indicates; Figure 4(a) is the change curve of gray mean square error, the left picture is the comparison result before correction, and the right picture is the comparison result after correction; Figure 4(b) is the non-uniformity change curve, the left picture It is the comparison result before correction, and the right figure is the comparison result after correction; it can be seen from the experimental results that compared with the two-point calibration correction method widely used in engineering practice, the correction effect of the present invention is obviously better than that of the two-point calibration method. Calibration correction method, the two image non-uniformity description parameters of image gray mean square error and non-uniformity are obviously smaller than the corresponding parameter indexes of the two-point calibration method, especially when the temperature is close to saturation, because the present invention adopts the non-linear The fitting method realizes the estimation of the detector response parameters, avoids the distortion problem caused by the linear fitting method, and ensures the robustness of the correction effect.

图5为针对40℃温度点红外图像序列，本发明与稳态卡尔曼滤波校正方法的校正效果对比图，本发明对应曲线用“----”表示，稳态卡尔曼滤波校正方法对应曲线用“——”表示；图5（a）为灰度均方差变化曲线图，左图为校正前对比结果，右图为校正后对比结果；图5（b）为非均匀度变化曲线图，左图为校正前对比结果，右图为校正后对比结果；从实验结果可以看出，与典型的基于场景的红外非均匀校正方法—稳态卡尔曼滤波校正方法相比，利用图像灰度均方差和非均匀度两个评价指标进行非均匀性校正效果评价，由于本发明充分利用了实时成像特征数据，因此在校正性能明显优于稳态卡尔曼滤波法。Figure 5 is a comparison diagram of the correction effect of the present invention and the steady-state Kalman filter correction method for the infrared image sequence at a temperature point of 40°C. The corresponding curve of the present invention is represented by "----", and the corresponding curve of the steady-state Kalman filter correction method Indicated by "—"; Figure 5(a) is the change curve of gray mean square error, the left picture is the comparison result before correction, and the right picture is the comparison result after correction; Figure 5(b) is the non-uniformity change curve, The left picture is the comparison result before correction, and the right picture is the comparison result after correction; from the experimental results, it can be seen that compared with the typical scene-based infrared non-uniformity correction method—steady-state Kalman filter correction method, using image gray level uniformity Two evaluation indexes of variance and non-uniformity are used to evaluate the effect of non-uniformity correction. Since the present invention makes full use of real-time imaging characteristic data, the correction performance is obviously better than that of the steady-state Kalman filter method.

1. A nonlinear fitting infrared non-uniform correction method based on time domain Kalman filtering correction is characterized by comprising the following steps:

the first step is as follows: parameter estimation:

optionally, four different uncorrected temperature points T are selected between 10 ℃ and 70 DEG C₁,T₂,T₃,T₄Infrared image of (f)_i,j(T₁),f_i,j(T₂),f_i,j(T₃),f_i,j(T₄) The gray values of the four infrared image pixel points are respectivelyAnd the middle subscript i, j represents the row and column of the pixel points, and the pixel points are respectively substituted into a formula I to solve a detector response equation:

The second step is that: and (3) realizing correction:

gray value g of infrared image pixel point needing to be corrected at any temperature point T_i,j(T) is substituted into a formula II to obtain the gray value of the corrected infrared image pixel point

The third step: and (3) time domain Kalman filtering modification:

step 1: calculating the image nu (t):

when NU (T) is less than or equal to 3.0 per mill, time bleaching treatment is not needed, and the technical scheme is ended; when NU (T) is more than 3.0 ‰, taking alpha _k=α∈[0.99,1]，β_k=β∈[0.99,1]And performing time bleaching treatment; when NU (T) is greater than 4.0 ‰, alpha is caused_k=α∈[0.9,0.99]，β_k=β∈[0.9,0.99]And performing time bleaching treatment; if the time drift is larger, namely NU (T) is more than 4.5 per thousand, let alpha_k=α∈[0.8,0.9]，β_k=β∈[0.8,0.9]And performing time bleaching treatment;

the time drift treatment refers to the following steps 2 to 4:

step 2: establishing a Kalman filtering state equation and an observation equation:

the state equation is as follows: x_i,j(k+1)＝Φ_kX_i,j(k)+M_k+W_k(formula four)

The observation equation: y is_i,j(k)＝H_kX_i,j(k)+V_k(formula five)

and 4, step 4: and (3) realizing time drift correction:

Will be provided withSubstituting into formula seven, pairPerforming time drift correction to obtain the gray value of the infrared image pixel point subjected to time drift processing