A single-frame corrosion boundary detection method and system based on a logarithmic first derivative of a pulsed infrared thermal imaging signal

Abstract

This invention discloses a single-frame corrosion boundary detection method and system based on the first logarithmic derivative of pulsed infrared thermal imaging signals. The method includes the following steps: applying pulsed thermal excitation to the surface of the component under test and simultaneously acquiring an infrared temperature sequence; performing a logarithmic transformation on the infrared temperature sequence to obtain a logarithmic thermal response sequence; calculating the first derivative of the logarithmic thermal response sequence along the time direction to obtain a derivative field; using the region to be detected as the statistical object, automatically determining the single-frame observation time t based on the spatial distribution and intensity of the negative derivative response in the derivative field; and extracting the single-frame derivative image corresponding to the single-frame observation time t to extract the boundary of the corrosion thinning region. The beneficial effects of this invention are: unified timing ensures stability; logarithmic derivative suppresses emissivity and heating unevenness interference; automatic frame selection avoids subjective bias and achieves rapid detection; zero-crossing boundaries have physical meaning and do not require manual thresholding; low computational cost and strong robustness.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, and in particular to a single-frame corrosion boundary detection method and system based on the first-order logarithmic derivative of pulsed infrared thermal imaging signals. Background Technology

[0002] Metal structures are susceptible to corrosion during long-term service due to the combined effects of various factors, including humid and hot environments, salt spray, industrial pollutants, and mechanical loads. Corrosion leads to localized wall thinning, decreased cross-sectional load-bearing capacity, intensified stress concentration, and reduced service reliability. In severe cases, it can even cause structural failure and safety accidents. Therefore, rapid, accurate, and reliable detection of corrosion defects in metal structures is of great significance for engineering equipment condition assessment, life prediction, and safe maintenance.

[0003] Currently, commonly used non-destructive testing methods in metal corrosion detection mainly include visual inspection, ultrasonic testing, eddy current testing, and conventional infrared thermography. Pulsed infrared thermography, as an active thermal non-destructive testing technology, has advantages such as non-contact operation, fast detection speed, ability to achieve large-area imaging, and suitability for rapid on-site screening. Under pulsed thermal excitation, the corroded thinned area and the intact area will exhibit different transient thermal response characteristics due to differences in local thickness, heat capacity, and heat diffusion path. Therefore, identifying and extracting the boundaries of corroded areas based on thermal response differences has a good physical basis. In existing research, methods such as temperature amplitude analysis, thermal contrast enhancement, thermal signal reconstruction, second derivative processing, frequency domain phase analysis, and threshold segmentation have been widely used to enhance defect visibility and improve defect identification.

[0004] However, existing methods still have certain shortcomings in accurately extracting corrosion boundaries. First, corroded surfaces are usually accompanied by rust products, rough textures, and emissivity variations, making results based on temperature amplitude or thermal contrast easily affected by surface conditions and environmental noise. Second, non-uniform heating, background temperature fluctuations, and imaging noise weaken the thermal response differences between corroded and intact areas, leading to discontinuous boundary contours or positioning shifts. Third, some methods rely on empirical thresholds, frequency domain transformations, complex reconstructions, or multi-frame fusion processing, making parameter selection and calculation processes complex and unsuitable for rapid on-site deployment and online processing. Furthermore, traditional evaluation indicators focus more on whether defect areas are visible or whether signals are enhanced, rather than directly guaranteeing good consistency between the extracted contour and the actual corrosion boundary. Therefore, it is necessary to further develop a corrosion boundary extraction method with clear physical meaning, fewer parameter dependencies, a simple calculation process, and suitability for rapid single-frame processing to improve the practicality and reliability of pulsed infrared thermography in the corrosion detection of metal structures. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to solve the problem of the difficulty in stable, accurate and rapid identification of corrosion boundaries during the pulse infrared thermography detection of metal structure corrosion.

[0006] The above-mentioned technical problems are solved by the following technical solution: This invention proposes a single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals, which includes the following steps: A pulsed thermal excitation was applied to the surface of the component under test, and an infrared temperature sequence was acquired simultaneously. Perform a logarithmic transformation on the infrared temperature sequence to obtain a logarithmic thermal response sequence; The first derivative of the logarithmic thermal response sequence is calculated along the time direction to obtain the derivative field; Using the region to be detected as the statistical object, the observation time of a single frame is automatically determined based on the spatial distribution and intensity of the negative derivative response in the derivative field; Extract the single-frame derivative image corresponding to the single-frame observation time, and extract the boundary of the corrosion thinning region based on the zero-crossing feature between the negative derivative band and the near-zero response region in the single-frame derivative image.

[0007] In a preferred embodiment of the single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: the step of automatically determining the single-frame observation time includes: For each post-pulse cooling candidate frame, the noise threshold is adaptively estimated based on the frame derivative field, and a set of negative responses is defined. Calculate the negative derivative response intensity across the entire detection region; The candidate frame that maximizes the negative derivative response intensity is selected as the single-frame observation time.

[0008] In a preferred embodiment of the single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: a robust statistical method is used to adaptively determine the derivative noise threshold required for extracting the corrosion boundary, specifically as follows: Calculate the median of the derivative field within the detection area, and calculate the standard deviation based on the median absolute deviation of the median; The derivative noise threshold is determined based on the standard deviation, and the derivative noise threshold is a set multiple of the standard deviation.

[0009] In a preferred embodiment of the single-frame erosion boundary detection method based on the first-order logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: after extracting candidate boundary points based on zero-crossing features, false boundaries are also eliminated based on connectivity length, area, or closure constraints. Among them, the minimum area constraint and the minimum length constraint are determined based on the spatial resolution and the minimum detectable corrosion diameter, or based on the proportion of the detection area and the number of pixels in the derivative image.

[0010] In a preferred embodiment of the single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: the infrared temperature sequence is logarithmically transformed by superimposing the temperature rise signal to avoid small positive numbers with unstable values ​​and then taking the natural logarithm. The small positive number is adaptively determined based on the maximum value among the standard deviation of the infrared thermal imager's temperature noise and the set multiple of the percentile value of the temperature rise image of the detection area.

[0011] In a preferred embodiment of the single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: the timing is controlled uniformly by data synchronization and storage unit, and the thermal excitation start time is defined as time zero. Temperature sequences with timestamps were continuously acquired during the pre-pulse reference segment, the pulse heating segment, and the post-pulse cooling segment.

[0012] A single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals is also proposed, characterized by comprising: The pulsed thermal excitation unit is used to apply pulsed thermal excitation to the surface of the component under test. Infrared thermal imaging acquisition unit, used to synchronously acquire infrared temperature sequences; A data synchronization and storage unit is connected to the pulse thermal excitation unit and the infrared thermal imaging acquisition unit, and is used to unify the trigger timing and store the infrared temperature sequence. The processor unit, which is connected to the storage unit in sync with the data, includes: The logarithmic first derivative calculation module is used to perform logarithmic transformation on the infrared temperature sequence and calculate the first derivative field. The single-frame time automatic selection module is used to automatically determine the single-frame observation time with the most significant difference in corrosion thermal response based on the spatial distribution and intensity of the negative derivative response in the derivative field, taking the entire area to be detected as the statistical object. The zero-crossing boundary extraction module is used to extract zero-crossing features from the single-frame derivative image at a single-frame observation time to obtain the erosion boundary.

[0013] In a preferred embodiment of the single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: the single-frame time automatic selection module calculates the average value of the negative derivative response intensity exceeding the adaptive noise threshold within the entire detection area, and selects the candidate frame with the largest average value of the negative derivative response intensity as the single-frame observation time.

[0014] In a preferred embodiment of the single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals described in this invention: the zero-crossing boundary extraction module adopts a robust statistical method, adaptively calculates the standard deviation based on the absolute deviation of the median of the derivative field, and determines the derivative noise threshold accordingly.

[0015] In a preferred embodiment of the single-frame corrosion boundary detection system based on the first-order logarithmic derivative of the pulsed infrared thermal imaging signal described in this invention: the data synchronization and storage unit, as a unified timing control core, sends a trigger signal to the pulsed thermal excitation unit and a synchronous acquisition signal to the infrared thermal imaging acquisition unit at the same time, defines the thermal excitation start time as the time zero point, and sends the infrared temperature sequence with timestamp to the processor unit.

[0016] The beneficial effects of this invention are as follows: Data synchronization and unified timing of the storage unit ensure detection stability; the logarithmic first derivative calculation module reduces dependence on absolute temperature amplitude and suppresses interference from emissivity and uneven heating; the automatic single-frame time selection module avoids the subjectivity of manual frame selection and achieves rapid detection; the zero-crossing boundary extraction module gives the boundary a physical meaning of thermal diffusion, eliminating the need for manual thresholding. Overall, it has low computational complexity and strong robustness. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention. Wherein: Figure 1 A system block diagram of a single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals is shown. Figure 2 A simulation model of corrosion of Q235 steel plate is shown; Figure 3 The simulated mesh generation for Q235 steel plate is shown; Figure 4 The corrosion numbering / defect arrangement of the Q235 steel plate is shown; Figure 5 The first-order logarithmic derivative plots are shown for different excitation durations and acquisition times; Figure 6 The transient temperature evolution curves of corrosion defects with different depths and diameters are shown; Figure 7 A physical-driven detection framework for corrosion thinning based on pulsed infrared thermal imaging is shown. Figure 8 The optimal single-frame selection result based on the two-stage criterion is shown; Figure 9 The pretreated Q235 steel plate is shown; Figure 10 The image shows a Q235 steel plate after a salt spray test; Figure 11 An infrared detection device is shown; Figure 12 It shows that the minimum response is reached at approximately 0.24 s under a 1 s heating condition. Figure 13 The minimum response is shown to be reached at approximately 1.64 s under a 2 s heating condition; Figure 14 The minimum response is shown to be reached at approximately 3.52 s under a 3 s heating condition; Figure 15 The single-frame logarithmic derivative imaging and line profile analysis of the salt spray corrosion zone under 1s heating are shown. Figure 16 The single-frame logarithmic derivative imaging and line profile analysis of the salt spray corrosion zone under 2s heating are shown. Figure 17 The single-frame logarithmic derivative imaging and line profile analysis of the salt spray corrosion zone under 3s heating are shown. Figure 18 The image shows a Q235 steel plate simulating natural corrosion. Figure 19 The identification of severely corroded regions is shown based on the logarithmic derivative response; Figure 20 The results show that the minimum response is reached at approximately 0.24 s under a 1 s heating condition on the naturally corroded sample; Figure 21 The results show that the minimum response is reached at approximately 1.48 s under a 2 s heating condition on the naturally corroded sample; Figure 22 The results show that the minimum response is reached at approximately 3.52 s under a 3 s heating condition on the naturally corroded sample; Figure 23 The single-frame logarithmic derivative imaging and contour analysis of the reference corrosion region at 0.24s are shown; Figure 24 The single-frame logarithmic derivative imaging and contour analysis of the reference corrosion region at 1.48 s are shown; Figure 25 The single-frame logarithmic derivative imaging and contour analysis of the 3.52s reference corrosion region are shown. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0019] The terminology used in this invention is that which is currently widely used in the art in consideration of the function of the invention; however, these terms may vary according to the intent of those skilled in the art, precedent, or new technology in the art. Furthermore, specific terms may be chosen by the applicant, and in such cases, their detailed meanings will be described in the detailed description of the invention. Therefore, the terms used in this specification should not be construed as simple names, but rather based on their meanings and the overall description of the invention.

[0020] Reference Figure 1 This embodiment provides a single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals. The method utilizes the physical law of temperature decay over time on the surface of a metal component after pulsed thermal excitation to transform the reduction in local heat capacity, effective thickness, and shortened heat diffusion path caused by corrosion thinning into characteristic responses in the derivative domain, thereby achieving rapid and accurate extraction of the corrosion region boundary. Specifically, it includes the following steps: A pulsed thermal excitation is applied to the surface of the component under test, and infrared temperature sequences are acquired simultaneously. In this embodiment, the component under test is a metal component surface. Under the action of pulsed thermal excitation, heat is conducted inside the metal component. Different regions of the component surface will exhibit different transient thermal response characteristics due to differences in local thickness, heat capacity, and heat diffusion path. At the same time, infrared temperature sequences of the component surface before heating, during heating, and during cooling are acquired simultaneously using an infrared thermal imaging device. The infrared temperature sequences contain temperature distribution information in both spatial and temporal dimensions. The duration is temperature sequence ,in These are spatial pixel coordinates. For the first Frame sampling time, ,1,..., This provides the raw data foundation for subsequent transient thermal response analysis; After acquiring the infrared temperature sequence, it is first subjected to basic preprocessing. Preprocessing may include bad pixel correction, region of interest truncation, timestamp correction, background temperature subtraction, temperature rise normalization, and lightweight spatial filtering. To avoid excessive smoothing in subsequent derivative calculations, spatial filtering can be performed using 3×3 median filtering or Gaussian filtering, and the filter kernel size is preferably no more than 1 / 3 of the corrosion boundary width. The infrared temperature sequence is subjected to a logarithmic transformation to obtain a logarithmic thermal response sequence. The logarithmic transformation includes the natural logarithm, common logarithms, and their mathematically equivalent transformation forms. The acquired infrared temperature sequence is transformed from the original temperature-time domain to the logarithmic domain. The transient thermal response of each pixel is then subjected to a logarithmic transformation using the following logarithmic transformation formula: ,in The initial temperature before the application of the heat pulse. To avoid unstable small positive values ​​when the temperature rise is close to zero, a logarithmic thermal response sequence is obtained. Logarithmic transformation effectively reduces the impact of differences in heat input intensity, surface radiation, and absolute temperature amplitude bias on the detection results. This allows the transformed signal to better represent the transient decay characteristics dominated by thermal diffusion, laying the foundation for subsequent extraction of features insensitive to surface conditions. The temperature rise signal is constructed as follows: Then perform a logarithmic transformation, that is , Press Adaptive determination, where The standard deviation of temperature noise for infrared thermal imagers. To detect the 95th percentile value of the temperature rise image in the detection area, in conventional infrared thermal imaging detection, The empirical value range is 0.01K to 0.05K. The main component enters the logarithmic temperature rise term. Its function is to prevent numerical instability when the temperature rises close to zero; for the time derivative term, its local effects can be mitigated by... Approximate characterization, when the temperature rise involved in the calculation hour, The relative impact is usually less than 5%; when At this time, the impact is usually less than 1%. In actual processing, this patent prioritizes selecting frames after heating and in the early cooling phase where the temperature rise signal is sufficiently higher than the noise level, and avoids the tail frames where the temperature rise is close to zero. The impact on the accuracy of the final derivative calculation is small and within an acceptable range. The first derivative of the logarithmic thermal response sequence is calculated along the time direction to obtain the derivative field. The following calculation formula is used: / · In discrete infrared sequences, time difference can be performed based on the logarithmic temperature rise sequence of adjacent sampling times. When the sampling interval is approximately uniform, it is preferable to use central difference to calculate the inner frame, and forward / backward difference to calculate the first and last frames. The corrosion-thinned region has a smaller effective thickness, lower local thermal mass, and shorter diffusion path, so it releases heat faster in the early cooling stage, forming a negative derivative band in the derivative plot. In contrast, the intact region of the metal component has a relatively stable thermal diffusion behavior, and the derivative value changes slowly, exhibiting a near-zero response in the derivative field. This difference in the derivative domain provides a physical driving feature for the identification of corrosion regions. Using the region to be detected as the statistical object, the observation time of a single frame is automatically determined based on the spatial distribution and intensity of the negative derivative response in the derivative field. During pulsed infrared thermal imaging, the thermal diffusion state varies at different acquisition times, and the degree of difference in transient thermal response between the corroded area and the intact area also changes accordingly. To avoid the subjectivity and uncertainty brought about by manually selecting frame images, the entire area to be detected is used as the statistical object. Instead of relying on the pre-known corrosion location, the moment with the strongest difference in corrosion-induced cooling and the most significant negative derivative response is automatically found in the time series based on the spatial distribution range and intensity characteristics of the negative derivative response in the derivative field. This moment is then determined as the single-frame observation time. This automatic selection mechanism ensures that the image at the selected moment is in a state where the thermal diffusion characteristics are most favorable for defect identification. Extract the single-frame derivative image corresponding to the single-frame observation time t*, and extract the boundary of the erosion thinning region based on the zero-crossing feature between the negative derivative band and the near-zero response region in the single-frame derivative image. This process is performed after determining the optimal single-frame observation time. Subsequently, the single-frame derivative image corresponding to that moment is extracted. In this single-frame derivative image, the erosion-thinned region is represented by a negative derivative band, while the intact region is represented by a near-zero response region. The boundary of the erosion-thinned region physically corresponds to the transition position where the thermal diffusion behavior changes. In the derivative field, it is represented by the sign change when transitioning from the negative derivative band to the near-zero response region. Based on the zero-crossing feature between the negative derivative band and the near-zero response region in the single-frame derivative image, the boundary contour of the erosion-thinned region can be accurately located and extracted. The boundary extraction criterion has a clear physical meaning of thermal diffusion and does not simply rely on experience, thereby improving the stability and accuracy of erosion boundary extraction.

[0021] Preferably, the observation time for a single frame is automatically determined. The steps include: For each post-pulse cooling candidate frame Noise threshold is adaptively estimated based on frame derivative field. And define the set of negative responses. Due to environmental disturbances and instrument-specific factors, noise inevitably exists in the derivative field. By adaptively estimating the noise threshold, it is possible to effectively distinguish the true corrosion negative derivative response from background noise fluctuations, thereby accurately defining the set of effective negative response pixels belonging to corrosion features. Calculate the negative derivative response intensity over the entire detection region.

[0022] in The area to be detected. The first derivative, For pixels in The first derivative at time t, As the noise threshold, using The function processes the data, retaining only those exceeding the noise threshold. The effective negative derivative response portion, with responses below the noise threshold set to zero, is obtained through coefficients. The total response over the entire detection region is normalized, and the negative derivative response intensity is calculated accordingly. It can comprehensively reflect the spatial coverage of the negative derivative response and the magnitude of the response in the current candidate frame; The candidate frame that maximizes the negative derivative response intensity is selected as the single-frame observation time. The system searches for the moment with the strongest negative derivative response, the most complete spatial coverage, and exceeding the noise level across the entire detection area. Therefore, it does not require prior knowledge of the corrosion location. To further improve stability, it can also be used in... Among the candidate frames that reach more than 95% of the maximum value, the frame with the strongest regional average negative response or the most stable connected negative response profile is selected as the [frame name / frame name]. In the consistency verification of theory, finite element method, and experiment, known geometric corrosion zones or manually marked corrosion zones are only used to evaluate IoU, Dice, and boundary positioning errors, and do not participate in the automatic selection of t* during engineering inspection. This avoids a dead loop caused by relying on unknown corrosion zones, and improves the response intensity in the negative derivative. At the peak moment, the difference in transient thermal diffusion between the corroded and thinned regions and the intact substrate regions is most significant, and the corrosion-induced cooling characteristics reach their peak in both spatial distribution and response intensity. Therefore, this moment is automatically determined as the single-frame observation moment. This ensures that the erosion boundary extracted from the single frame image is the clearest and most stable, effectively avoiding boundary blurring or positioning offset caused by observing too early or too late.

[0023] Furthermore, a robust statistical method is used to adaptively determine the derivative noise threshold required for extracting the corrosion boundary. When extracting zero-crossing features, in order to accurately distinguish between the real corrosion thermal response and background noise, it is necessary to reasonably set the derivative noise threshold. Since the corrosion region appears as a significant negative derivative band in a single-frame derivative image, if conventional mean and standard deviation calculation methods are used, these significant negative outliers will severely increase the standard deviation, leading to an overly high threshold and thus causing missed detection of corrosion boundaries. Therefore, this method uses a robust statistical method to adaptively determine the threshold, specifically: In the detection area Median of the internally computed derivative field , The median is insensitive to extreme outliers and can more accurately represent the central location of the near-zero response in the intact matrix region, and the standard deviation can be calculated. ,in For a single-frame derivative image, the formula utilizes the median absolute deviation to estimate the dispersion of the derivative field. The coefficient 1.4826 is a correction factor that ensures the absolute deviation estimate aligns with the standard deviation estimate under a normal distribution. The median absolute deviation only considers the typical magnitude of deviation from the median; large negative deviations caused by corrosion regions do not significantly affect it, thus accurately reflecting the fluctuation level of background noise. This median absolute deviation estimate is insensitive to a small amount of negative corrosion response, avoiding the corrosion region from raising the standard deviation. Furthermore, an iterative approach can be used to estimate the approximately complete background region: first, using the entire region... Calculate the initial And obtain the initial negative response region; then the negative response region and its neighborhood are separated from... Exclude from the middle and recalculate in the remaining area. The process is iterated 1 to 2 times to obtain the result. If the corrosion area is large, the noise level can be estimated using the pre-pulse reference frame or the post-cooling stable frame. Alternatively, the threshold can be limited to 1% to 5% of the 95th percentile of the absolute value of the derivative field to avoid introducing false boundaries due to an excessively low threshold or missing true boundaries due to an excessively high threshold. In a discrete pixel matrix, if adjacent pixels... and satisfy and ,or and Then determine and There are candidate zero-crossing boundary points; the positions of the candidate points can be determined by linear interpolation, with an interpolation ratio of... To reduce false boundaries, candidate zero-crossings must also satisfy connectivity length, area, or closure constraints. If the spatial resolution is... The minimum detectable corrosion diameter is ,but Desirable , Desirable If no prior dimensions are available, Desirable (pixels) or 0.05% to 0.5% × , Desirable or For those who do not meet the requirements , Isolated contours with closed constraints can be removed, and morphological closure, opening, and smoothing operations can be used. The effective zero-crossing boundary of the logarithmic first derivative field, the corresponding formula for this criterion. The transition point between the negative derivative band of the corrosion zone and the near-zero response of the intact zone is the corrosion boundary. In quantitative verification, the predicted corrosion region enclosed by the zero-crossing profile is defined as... The reference corrosion area is ,in The evaluation metrics used, derived from salt spray corrosion of a fixed shape and manually labeled reference corrosion masks, include cross-union ratio (CUP). , and boundary positioning error The validation results show that the zero-crossing boundary extraction method was applied to six groups of corrosion samples and was effective. , The results in the table reflect the overall performance of multiple samples rather than a single sample, as the boundary errors are averaged. In discrete image implementation, zero-crossing can be determined by the change in the sign of the derivatives of adjacent pixels or adjacent sampling points. This is combined with the continuity of the negative derivative band, the connectivity of the boundary contour, and the physical stability of the single-frame derivative map emphasized in this technical solution to eliminate isolated jumps caused by local noise. It should be noted that the reference erosion area... This is for implementation examples and performance verification only. In practical engineering applications, the algorithm only uses the temperature sequence obtained by the infrared thermal imaging acquisition unit, the detection area R, and the adaptive threshold. Complete single-frame selection and zero-crossing boundary extraction Determine the derivative noise threshold based on the standard deviation. ,in The threshold is a constant between 2 and 3 (including endpoint values). Setting the threshold to 2 to 3 times the noise standard deviation can effectively suppress background noise interference and eliminate false zero cross points caused by noise, while ensuring that the true corrosion negative derivative response is completely preserved, thereby achieving reliable extraction of corrosion boundaries.

[0024] Preferably, in step S5, after extracting candidate boundary points based on zero-crossing features, false boundaries are also eliminated based on connectivity length, area, or closure constraints. After extracting candidate boundary points based on zero-crossing features, due to factors such as infrared imaging noise, local surface state disturbances, or minor non-defect thermal anomalies, the extracted candidate boundaries may contain some isolated, discrete, or discontinuous false boundaries. Real corrosion defects usually have a certain geometric scale and continuity. Their corresponding boundary contours are spatially represented as curves with a relatively long connected length, a certain coverage area, and tend to be closed, while false boundaries often appear as scattered short line segments or tiny areas.

[0025] To ensure the accuracy and reliability of erosion boundary extraction, these false boundaries are further eliminated based on connectivity length, area, or closure constraints. By introducing connectivity length, area, or closure constraints, the true boundaries and false boundaries can be effectively distinguished. Among them, minimum area constraint and minimum length constraint The determination can be based on spatial resolution and the minimum detectable corrosion diameter, or on the proportion of the detection area and the number of pixels in the derivative image. The first method is based on physical detection scale, that is, based on spatial resolution and minimum detectable corrosion diameter. and This approach directly links the constraints to the spatial resolution of the testing instrument and the minimum physical size of the actual corrosion to be inspected, giving the rejection criteria a clear physical meaning. The second method is based on image statistical features, specifically on the proportion of the detection region and the number of pixels in the derivative image. and This approach starts from the statistical regularities of the entire image, and sets a constraint threshold proportionally based on the size of the detection region and the total number of pixels in the image, enabling it to adapt to detection scenarios of different sizes and resolutions. By setting and filtering the above constraints, those that do not meet the minimum area requirement can be eliminated. Minimum length Isolated contours that do not meet the closure requirement are discarded as false boundaries, thereby significantly reducing the false detection rate and extracting the true and continuous corrosion defect boundaries.

[0026] Preferably, in step S2, the formula used for the logarithmic transformation is: , The temperature rise signal represents the temperature rise of the metal component surface after pulsed thermal excitation relative to the initial state before thermal excitation. It directly reflects the transient response characteristics of thermal excitation conduction and diffusion within the component. To avoid small positive values ​​with unstable values, in actual thermal diffusion, the surface temperature rise signal gradually decays and approaches zero as cooling time progresses. If logarithmic calculations are performed directly, even small temperature fluctuations can easily cause drastic changes in the logarithmic value, severely affecting the accuracy and stability of subsequent derivative calculations. Therefore, a method is introduced... Crucial; The according to Adaptive determination, where The standard deviation of temperature noise in an infrared thermal imager reflects the temperature measurement noise level of the infrared acquisition equipment itself. The 95th percentile value of the temperature rise image in the detection area represents the effective distribution level of the overall temperature rise and the intensity of heat input within the detection area.

[0027] By taking and The maximum of the two is used as This ensured The value is always greater than three standard deviations of the instrument's underlying noise to effectively suppress numerical singularities caused by noise fluctuations, while ensuring that ε is proportional to the overall temperature rise level of the region. This adaptive determination method avoids the loss of smoothing protection due to ε being relatively small under strong heat input scenarios, and also avoids the masking of the true temperature rise decay characteristics due to ε being relatively large under weak heat input scenarios. This ensures that the logarithmic transformation has excellent numerical stability throughout the entire time series, laying the foundation for subsequent accurate calculation of the first logarithmic derivative and reliable extraction of corrosion boundaries.

[0028] Preferably, in step S1, the thermal excitation start time is defined by unifying the timing control through data synchronization and storage units. The core of pulsed infrared thermal imaging detection lies in capturing and analyzing the transient thermal response process of a material surface after thermal excitation. A strict time correspondence must exist between thermal excitation and temperature acquisition. By controlling the timing through data synchronization and storage units, it is possible to ensure that the acquisition action of the infrared thermal imaging device and the application of the pulsed thermal excitation are precisely synchronized on the time axis, avoiding the disconnect between the temperature sequence and the actual thermal diffusion process due to timing misalignment. The start time of thermal excitation is uniformly defined as... This establishes an absolute time reference origin for the entire transient thermal response sequence, which facilitates subsequent logarithmic transformation and time derivative calculation based on a unified time coordinate, and accurately characterizes the temperature decay law during the cooling stage. Time-stamped temperature sequences were continuously acquired during the pre-pulse reference segment, pulse heating segment, and post-pulse cooling segment. The pre-pulse reference segment was used to obtain the initial temperature distribution and background noise level of the component surface, providing a benchmark for subsequent temperature rise signal calculation. The pulse heating segment recorded the surface temperature rise response during heat flow input. The post-pulse cooling segment contained complete transient information on heat diffusion within the component and heat exchange with the external environment, a crucial stage for identifying differences in heat diffusion caused by corrosion thinning. Continuous acquisition of time-stamped temperature sequences ensured the integrity of the time-domain signal, enabling each pixel to obtain a continuous and accurately time-calibrated temperature-time evolution curve. And the sampling frequency is not less than 20Hz. The sampling frequency should be no less than 20Hz, preferably between 20Hz and 200Hz. Because the temperature change of the metal structure is extremely rapid in the early cooling stage after pulsed thermal excitation, the difference in transient thermal response between the corroded and intact regions often reaches its peak within a very short time window. Therefore, a sufficiently high temporal resolution must be used to capture this transient characteristic. Setting the sampling frequency to no less than 20Hz ensures that sufficiently dense data points are collected in the time dimension. This avoids truncation errors, feature blurring, or omission of the optimal transient moment due to excessively long sampling intervals when calculating the first derivative along the time direction, ensuring the accuracy of the derivative field calculation and the reliability of boundary extraction. The total acquisition time is... The total number of collected frames meets the requirements. .

[0029] Experiment 1, finite element simulation verification, details are as follows: Figure 2 The finite element model shown is a Q235 steel plate with dimensions of 100mm × 50mm × 5mm. Figure 4 As shown, nine flat-bottomed corrosion / wall thinning defects are arranged near the center of the steel plate. The defect depths are 1 mm, 3 mm, and 4 mm, and the defect diameters are 6 mm, 8 mm, and 10 mm, forming a 3×3 depth-diameter combination. Each defect reduces the local plate thickness from the nominal thickness. =5mm reduced to Implementation, in which This represents the corrosion depth.

[0030] like Figure 4 As shown, to analyze the steep temperature gradient near the corrosion boundary, this embodiment employs a structured-unstructured hybrid mesh with localized refinement at the pit bottom, pit sidewalls, upper observation surface, and the transition region between the thinned and intact areas. The final mesh comprises 206,253 domain elements, 22,180 boundary elements, and 1,360 edge elements. Regarding boundary conditions, this embodiment applies a 2-second rectangular heat flux pulse to the upper surface to simulate experimental flash thermal excitation, while simultaneously applying natural convection and radiation to the exposed surface. The ambient temperature is fixed at 293.15 K. To capture the rapid thermal transients generated by the pulsed heating, the transient thermal analysis governing equations satisfy... ,in A rectangular heat flux pulse is used to capture the rapid thermal transients generated by pulse heating, employing a specific method during the initial 0-2s phase. The minimum time step is set to s, and then gradually adjusted to 0.05~0.1s.

[0031] After the simulation is completed, the temperature sequence of the observation surface is extracted. The temperature rise signal is constructed, logarithmic transformation is performed, and the first derivative is calculated according to the method steps of this invention, such as... Figure 6As shown, the transient temperature evolution curves of corrosion defects with different depths and diameters are displayed. It can be seen that there is a significant difference in the cooling rate between the corrosion-thinned region and the intact region. Figure 5 As shown, in the first-order logarithmic derivative plots under different excitation durations and acquisition times, the eroded region exhibits a distinct negative derivative band, such as... Figure 7 As shown, the physical-driven detection framework for corrosion thinning based on pulsed infrared thermography exhibits a high degree of consistency between the zero-crossing position of the outer edge of the negative derivative band and the preset corrosion geometric boundary. Figure 8 As shown, the optimal single-frame selection result based on the two-stage criterion further verifies the rationality of automatically selecting time t*.

[0032] Experiment 2 involves the detection of salt spray corrosion samples. This embodiment uses salt spray corrosion samples to verify the single-frame detection strategy based on the first logarithm of the first derivative. Figure 9 and Figure 10 The images shown are of Q235 steel plates after pretreatment and after salt spray testing, respectively. Figure 11 The infrared detection device is shown. The experiment uses pulse heating conditions consistent with the finite element simulation, and sets three pulse durations of 1s, 2s and 3s. The infrared camera records the complete heating-cooling process, and performs logarithmic transformation on the temperature sequence of each pixel and calculates the first derivative.

[0033] Figures 12-14 The curves showing the change of the first derivative of the spatial mean logarithm with time under thermal excitation conditions of 1s, 2s, and 3s are presented respectively. It can be seen from the figure that the curve reaches its minimum value at approximately 0.24s after 1s heating, approximately 1.64s after 2s heating, and approximately 3.52s after 3s heating, which are the automatically selected single-frame observation times t*.

[0034] Figures 15-17 Single-frame logarithmic derivative imaging and line profile analysis of the salt spray corrosion zone under heating for 1s, 2s, and 3s respectively are presented. The colored areas in the figure represent the magnitude of the first logarithmic derivative, the negative bands correspond to the corrosion thinning response, and the contour lines represent the effective zero-crossing boundaries. The profile curves clearly show the negative valleys and zero-crossing points on both sides of the selected scan line. The experimental results show that different pulse durations will change the response amplitude and the optimal observation time, but the zero-crossing points can stably correspond to the corrosion boundary. The boundary indices are mainly controlled by the transient thermal diffusion difference between the thinned area and the intact area.

[0035] Experiment 3 involved the testing of naturally corroded samples. In this embodiment, the naturally corroded samples were tested under the same pulse heating and infrared acquisition parameters as the salt spray samples. Figure 18 The image shows a Q235 steel plate simulating natural corrosion. Natural corrosion has stronger surface roughness, rust products, differences in emissivity distribution, and irregular geometric morphology. Therefore, brightness changes and reflection-related disturbances are likely to occur in the original temperature field.

[0036] Figures 20-22 The spatially averaged logarithmic first derivative response curves of the naturally corroded samples under pulse heating conditions of 1s, 2s, and 3s are presented. The minimum points of the curves represent the automatically selected optimal observation time t*, which reaches the representative minimum response at approximately 0.24s, 1.48s, and 3.52s, respectively. Figure 19 The results of identifying severely corroded areas based on logarithmic derivative response are presented. Areas with stronger negative responses correspond to severely corroded areas with more pronounced thermal diffusion disturbances.

[0037] Figures 23-25 The single-frame logarithmic derivative imaging and contour analysis of the reference corrosion region at 0.24s, 1.48s, and 3.52s are shown respectively. The color fill in the figure represents the single-frame logarithmic first derivative response, the closed / near-closed contour lines represent the effective zero-crossing boundaries, and the line profile shows the negative valleys and boundary points. The detection results show that the single-frame logarithmic first derivative map can effectively suppress low-frequency background and multiplicative noise. The derivative profile shows a three-segment structure: the plateau of the intact region, the negative valley of the corrosion region, and the edge transition region. For the irregular surface texture and rust products caused by natural corrosion, the zero-crossing contour highlights the boundary formed by the change of thermal diffusion behavior, rather than a simple visible light color edge.

[0038] Specifically, this embodiment uses salt spray corrosion samples to verify a single-frame detection strategy based on the first logarithmic derivative. The experiment adopts pulse heating conditions consistent with finite element simulation, and sets three pulse durations of 1s, 2s, and 3s. An infrared camera records the complete heating-cooling process, and each pixel forms a temperature-time series. Subsequently, a logarithmic transformation is performed on the temperature series of each pixel, and the first derivative is calculated. It should be noted that the purpose of preset pulse duration is not to optimize all possible excitation parameters, but to verify whether the zero-crossing profile of the first logarithmic derivative can serve as an effective corrosion boundary index under given pulse heating conditions. The experimental results show that different pulse durations will change the response amplitude and the optimal observation time, but the boundary index is mainly controlled by the transient thermal diffusion difference between the thinned area and the intact area. In the salt spray corrosion experiment, the representative minimum response is reached at approximately 0.24s after 1s heating, approximately 1.64s after 2s heating, and approximately 3.52s after 3s heating. Figures 15-17 The line profiles all show negative valleys in the corrosion zone and zero intersection points on both sides, indicating that the zero intersection points can stably correspond to the corrosion boundary.

[0039] Experiment 3 was conducted on naturally corroded samples. The main basis for this experiment is as follows: In this embodiment, the naturally corroded samples were subjected to pulsed infrared thermal imaging under the same pulsed heating and infrared acquisition parameters as the salt spray samples. Natural corrosion has stronger surface roughness, rust products, emissivity distribution differences and irregular geometric morphology. Therefore, brightness changes, halo effects and reflection-related disturbances are likely to occur in the original temperature field. The detection results show that the single-frame logarithmic first derivative plot can suppress these low-frequency background and multiplicative noise. The derivative profile still shows a three-segment structure: the plateau of the intact region, the negative valley of the corrosion region and the edge transition region. The representative minimum response times of the naturally corroded samples under 1s, 2s and 3s pulses are approximately 0.24s, 1.48s and 3.52s, respectively. Figures 23-25 The results show that the zero-crossing points have a high spatial correspondence with the visually observed corrosion boundaries.

[0040] To evaluate the performance of the method of the present invention, the above detection results were quantitatively analyzed. Table 1 shows the corrosion boundary localization performance at different observation times. The results show that at the automatically selected optimal single frame time, the cross-over ratio and Dice coefficient are maintained at a high level, and the boundary error is extremely small.

[0041] Table 1 Corrosion boundary localization performance at different observation times

[0042] Table 2 compares the method of this patent with representative infrared thermal imaging processing methods such as thermal contrast imaging, thresholded thermal signal reconstruction, and frequency domain phase analysis. As shown in Table 2, the cross-union ratio of this patent method reaches 0.701723, the Dice coefficient reaches 0.824721, and the boundary error is only 1.453456 px. Compared with the thermal contrast imaging method, the boundary error is reduced by approximately 81.31%, fully demonstrating the significant superiority of the method of this invention in accurately extracting corrosion boundaries.

[0043] Table 2 Comparison with representative infrared thermal imaging processing methods Note in Table 2: The results of this patented method in Table 2 are the optimal single-frame detection results automatically selected by step S5 under the 2s pulse heating condition, i.e., IoU=0.701723, Dice=0.824721, and boundary error Eb=1.453456px. Table 2 compares these results with representative methods such as thermal contrast imaging, thresholded thermal signal reconstruction, and frequency domain phase analysis.

[0044] Reference Figure 1 As an optional embodiment, a single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals is also provided. This system specifically includes: The pulsed thermal excitation unit is used to apply pulsed thermal excitation to the surface of the metal component under test. During the detection, the unit provides transient, high-intensity heat input, causing thermal diffusion inside the component. It utilizes the difference in thermophysical properties between the corrosion-thinned area and the intact area to excite different transient thermal response characteristics, providing a physical basis for subsequent infrared detection. The infrared thermal imaging acquisition unit is used to synchronously acquire infrared temperature sequences including the pre-pulse, during-pulse, and post-pulse cooling stages. It continuously records the temperature distribution changes of the metal component surface before thermal excitation, during heating, and during cooling in a non-contact manner, and obtains complete transient thermal response data R including time and space dimensions. The data synchronization and storage unit is connected to the pulse thermal excitation unit and the infrared thermal image acquisition unit. It is used to unify the trigger timing and store the infrared temperature sequence, ensuring that the application of thermal excitation and the acquisition of temperature images strictly correspond on the time axis, avoiding analysis errors caused by timing misalignment, and storing the large amount of acquired temperature sequence data in an orderly manner for the processor unit to call. The processor unit, synchronized with the data and connected to the storage unit, is used to execute core algorithm processing, including: The logarithmic first derivative calculation module is used to perform logarithmic transformation on the infrared temperature sequence and calculate the first derivative field. By performing logarithmic transformation on the temperature sequence, the dependence on the absolute temperature amplitude is reduced, and the interference caused by surface emissivity differences and uneven heating is suppressed. The first derivative field calculated along the time direction transforms the cooling rate difference caused by corrosion thinning into negative response characteristics in the derivative domain. The single-frame time automatic selection module is used to take the entire area to be detected as the statistical object and, based on the spatial distribution and intensity of the negative derivative response in the derivative field, automatically determine the single-frame observation time t* with the most significant difference in corrosion thermal response. This avoids the subjectivity and uncertainty caused by manual selection of image frames based on experience, and ensures the automation and optimization of the detection process. The zero-crossing boundary extraction module is used to extract zero-crossing features from the single-frame derivative image at time t* to obtain the erosion boundary. In the selected single-frame derivative image, the zero-crossing feature is used to directly extract the erosion boundary contour by utilizing the sign change between the negative derivative band of the erosion region and the near-zero response of the complete region, so that the extracted boundary has a clear physical meaning of thermal diffusion. The results display / output module is used to output the corrosion boundary detection results. The module will visualize the extracted corrosion boundary location, contour and other information, or overlay it on the original infrared thermal image or visible light image for output, providing an intuitive and reliable basis for engineers to assess the condition and predict the life.

[0045] Preferably, during the cooling phase after pulsed thermal excitation, the thermal diffusion state varies at different times, and the difference in derivatives between the corroded region and the intact region also changes over time. To automatically lock the optimal detection time without relying on prior corrosion locations, the single-frame time automatic selection module is based on the formula... Calculate the negative derivative response intensity and select the candidate frame with the highest intensity as the single-frame observation time. , The entire area to be inspected. The total number of pixels in the detection area; Spatial coordinates are pixels in The first derivative value at time t; t is the noise threshold adaptively estimated based on the derivative field of the current frame, which is calculated by accumulating and averaging the effective negative responses of all pixels in the entire detection area. The value comprehensively reflects the At time t, the spatial coverage and intensity amplitude of the negative derivative response in the derivative field, The maximum value is precisely when the transient cooling difference between the etched thinning zone and the intact matrix zone is most significant and the negative derivative feature is most strongly expressed, thus ensuring that the single frame image automatically selected by the system is most conducive to subsequent etch boundary extraction.

[0046] Furthermore, the zero-crossing boundary extraction module employs a robust statistical method, adaptively calculating the standard deviation based on the median absolute deviation of the derivative field. Based on this, the derivative noise threshold is determined. Because the eroded and thinned regions exhibit strong negative derivative bands in a single-frame derivative image, these large-area or high-intensity anomalous responses are statistically significant outliers. If conventional mean and standard deviation calculation methods are used to assess noise levels, these erosion outliers will severely increase the overall standard deviation, leading to an overestimation of the estimated background noise level, and consequently causing the set derivative noise threshold to be lowered. If the value is too large, the actual corrosion boundary will be misjudged as noise and missed. The zero-crossing boundary extraction module employs robust statistical methods to process derivative field data. This module assesses the data dispersion based on the median absolute deviation of the derivative field. Since the median and median absolute deviation are highly resistant to extreme outliers, they can accurately and stably reflect the near-zero response fluctuations of a large, intact matrix region within the derivative field. The standard deviation is adaptively calculated based on the median absolute deviation of the derivative field. The zero-crossing boundary extraction module can obtain a true estimate of the background noise that is not affected by the abnormal corrosion response. Based on this result, the module determines the derivative noise threshold η, ensuring that the threshold can effectively filter out false zero-crossing points caused by background noise and completely preserve the true negative derivative response characteristics of corrosion. This ensures that the corrosion boundaries extracted by the system under various complex detection conditions have high accuracy and robustness.

[0047] Preferably, the data synchronization and storage unit, as a unified timing control core, sends a trigger signal to the pulse thermal excitation unit and a synchronization acquisition signal to the infrared thermal imaging acquisition unit simultaneously, defining the thermal excitation start time as... The infrared temperature sequence with timestamps is sent to the processor unit. The core of pulse infrared thermal imaging detection is to capture and analyze the transient thermal response process of the component surface after thermal excitation. This requires that the application of thermal excitation and the acquisition of temperature data must maintain a strict correspondence on the time axis. In this system, the data synchronization and storage unit acts as the unified timing control core of the entire detection system, responsible for coordinating the action rhythm of each unit. At the start of detection, the data synchronization and storage unit sends a trigger signal to the pulse thermal excitation unit to initiate thermal excitation. Simultaneously, this unit also sends a synchronization acquisition signal to the infrared thermal imaging acquisition unit, ensuring that the application of thermal excitation and the recording of the infrared image begin at the same instant. Through this synchronous triggering mechanism, the system precisely defines the start time of thermal excitation as time zero (i.e.,...). This establishes an absolute time reference origin for the entire transient thermal response process, avoiding timing misalignment or time drift caused by independent operation between devices. During the acquisition process, the data synchronization and storage unit binds the temperature data acquired by the infrared thermal imaging acquisition unit with time information, forming a timestamped infrared temperature sequence. Upon completion of acquisition or when preset conditions are met, the data synchronization and storage unit sends these timestamped infrared temperature sequences to the processor unit. The introduction of timestamps ensures that each temperature data point has accurate time attributes, enabling the processor unit to analyze the evolution of thermal diffusion behavior over time based on a unified and precise time coordinate during subsequent processing. This provides reliable temporal data support for accurately extracting corrosion boundaries.

[0048] Finally, it should be noted that the methods and devices described in detail above are merely embodiments, and those skilled in the art can modify these embodiments in different ways as long as they do not depart from the scope of the present invention.

Claims

1. A single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals, characterized in that: Includes the following steps, A pulsed thermal excitation was applied to the surface of the component under test, and an infrared temperature sequence was acquired simultaneously. Perform a logarithmic transformation on the infrared temperature sequence to obtain a logarithmic thermal response sequence; The first derivative of the logarithmic thermal response sequence is calculated along the time direction to obtain the derivative field; Using the region to be detected as the statistical object, the observation time of a single frame is automatically determined based on the spatial distribution and intensity of the negative derivative response in the derivative field; Extract the single-frame derivative image corresponding to the single-frame observation time, and extract the boundary of the corrosion thinning region based on the zero-crossing feature between the negative derivative band and the near-zero response region in the single-frame derivative image.

2. The single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 1, characterized in that: The steps for automatically determining the observation time of a single frame include: For each post-pulse cooling candidate frame, the noise threshold is adaptively estimated based on the frame derivative field, and a set of negative responses is defined. Calculate the negative derivative response intensity across the entire detection region; The candidate frame that maximizes the negative derivative response intensity is selected as the single-frame observation time.

3. The single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 2, characterized in that: A robust statistical method is used to adaptively determine the derivative noise threshold required for extracting corrosion boundaries, specifically: Calculate the median of the derivative field within the detection area, and calculate the standard deviation based on the median absolute deviation of the median; The derivative noise threshold is determined based on the standard deviation, and the derivative noise threshold is a set multiple of the standard deviation.

4. The single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to any one of claims 1 to 3, characterized in that: After extracting candidate boundary points based on zero-crossing features, false boundaries are also eliminated based on connectivity length, area, or closure constraints. Among them, the minimum area constraint and the minimum length constraint are determined based on the spatial resolution and the minimum detectable corrosion diameter, or based on the proportion of the detection area and the number of pixels in the derivative image.

5. The single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 1, characterized in that: The infrared temperature sequence is logarithmically transformed by superimposing a preset small positive number on the temperature rise signal and then taking the natural logarithm. The small positive number is adaptively determined based on the maximum value among the standard deviation of the infrared thermal imager's temperature noise and the set multiple of the percentile value of the temperature rise image of the detection area.

6. The single-frame corrosion boundary detection method based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 1, characterized in that: By unifying the timing control through data synchronization and storage units, the start time of thermal excitation is defined as time zero. Temperature sequences with timestamps were continuously acquired during the pre-pulse reference segment, the pulse heating segment, and the post-pulse cooling segment.

7. A single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals, characterized in that, include: The pulsed thermal excitation unit is used to apply pulsed thermal excitation to the surface of the component under test. Infrared thermal imaging acquisition unit, used to synchronously acquire infrared temperature sequences; A data synchronization and storage unit is connected to the pulse thermal excitation unit and the infrared thermal imaging acquisition unit, and is used to unify the trigger timing and store the infrared temperature sequence. The processor unit, which is connected to the storage unit in sync with the data, includes: The logarithmic first derivative calculation module is used to perform logarithmic transformation on the infrared temperature sequence and calculate the first derivative field. The single-frame time automatic selection module is used to automatically determine the single-frame observation time with the most significant difference in corrosion thermal response based on the spatial distribution and intensity of the negative derivative response in the derivative field, taking the entire area to be detected as the statistical object. The zero-crossing boundary extraction module is used to extract zero-crossing features from the single-frame derivative image at a single-frame observation time to obtain the erosion boundary.

8. The single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 7, characterized in that: The single-frame time automatic selection module calculates the average value of the negative derivative response intensity exceeding the adaptive noise threshold within the entire detection area, and selects the candidate frame with the largest average value of the negative derivative response intensity as the single-frame observation time.

9. The single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 7 or 8, characterized in that: The zero-crossing boundary extraction module employs a robust statistical method, adaptively calculating the standard deviation based on the median absolute deviation of the derivative field, and determining the derivative noise threshold accordingly.

10. The single-frame corrosion boundary detection system based on the first logarithmic derivative of pulsed infrared thermal imaging signals according to claim 7, characterized in that: The data synchronization and storage unit, as the unified timing control core, sends a trigger signal to the pulse thermal excitation unit and a synchronous acquisition signal to the infrared thermal imaging acquisition unit at the same time. It defines the start time of thermal excitation as the zero point and sends the infrared temperature sequence with timestamps to the processor unit.

Images