Thermal signal data processing method

Abstract

The invention discloses a thermal signal data processing method, which relates to the technical field of thermal signal processing, and is characterized in that a pulse thermal imaging sequence is segmented into a plurality of subsequences through an incremental multiplication relation segmentation method, and the length of each subsequence is gradually increased to ensure that each subsequence can effectively capture the complete change of a thermal signal. The automatic segmentation mode not only avoids the problem of information loss possibly caused by a fixed-length sub-sequence in a traditional method, but also can dynamically adjust the length of each sub-sequence according to the change of the thermal signal, thereby improving the flexibility and efficiency of data processing. A moving average window method is adopted, time sequence noise can be effectively removed, and main information of thermal signals is reserved. Time dimension averaging is carried out on multiple frames of data through the sliding window, random noise and high-frequency jitter generated in the collection process of the thermal imager and transient interference signals in the heat conduction process in the image can be effectively restrained, and therefore the signal-to-noise ratio of a heat map sequence is increased.

Description

Technical Field

[0001] This invention relates to the field of thermal signal processing technology, specifically a thermal signal data processing method. Background Technology

[0002] Thermal imaging technology is widely used in various fields, especially in non-destructive testing, intelligent monitoring, and industrial automation. As a non-contact detection method based on temperature changes, thermal imaging can capture real-time temperature changes on the surface of an object, and is widely used in equipment maintenance, quality inspection, and building safety monitoring. Particularly in the field of pulsed thermal imaging, it excites the surface of an object with thermal pulses and detects temperature fluctuations during the thermal response process, effectively reflecting internal defects or inhomogeneities in materials, and is applied to tasks such as structural inspection and defect identification. Specifically, pulsed thermal imaging technology, in the acquisition and processing of three-dimensional thermal image sequences, can provide accurate detection results for problems such as internal defects and structural changes in objects.

[0003] Pulse thermal imaging is a non-destructive testing technique that uses extremely short flash pulses applied to the surface of a material. The energy of the flash pulse is absorbed by the material surface and converted into heat, which is then conducted within the material. Furthermore, an infrared camera can record temperature changes on the material surface, and thermal anomalies generated on the surface can reveal the presence of internal structural abnormalities such as porosity, delamination, and cracks. However, the thermal signals acquired by infrared cameras are often interfered with by various noise sources, the most common of which include thermal radiation from the external environment, instability of the infrared camera detector, and non-uniformity of the material surface emissivity. Moreover, thermal anomalies caused by internal defects are superimposed on non-uniform heating. As the defect depth increases, the thermal signal decays exponentially, and the effect of lateral thermal diffusion becomes increasingly significant. This greatly limits the detection depth and spatial resolution of pulse thermal imaging; therefore, it is necessary to develop advanced thermal signal processing methods to effectively detect or characterize potential defects in materials.

[0004] Existing thermal signal processing methods, especially in the post-processing of pulsed thermal imaging sequences, face several challenges and shortcomings. Traditional methods typically process single-pixel sequences one by one. This inefficient approach requires extensive manual input and parameter selection and is susceptible to interference from redundant data, leading to inaccurate detection results. Particularly when dealing with complex 3D thermal image sequences, traditional methods are slow and cannot effectively improve data processing efficiency. Furthermore, existing technologies rely heavily on prior information, such as the need to obtain information about the detection sample and reference region beforehand. This not only increases operational complexity but also raises the technical barrier for users, limiting the widespread application of this technology. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a thermal signal data processing method that solves the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a thermal signal data processing method, comprising the following steps:

[0007] S1. Collect complete thermal imaging data using an infrared thermal imager and perform noise reduction processing using a moving average window method to obtain a pulsed thermal imaging sequence.

[0008] S2. By using the incremental multiplication relationship segmentation method, the pulse thermal imaging sequence is segmented to obtain several sub-sequences;

[0009] S3. By using the dynamic compression reconstruction DCR function to process each subsequence, noise in redundant data is eliminated, the signal contrast of defects is improved, and the reconstructed subsequence is obtained.

[0010] S4. Combine the reconstructed subsequences into a reconstructed heatmap, and then combine them in sequence into a complete three-dimensional heatmap sequence.

[0011] S5. Based on the three-dimensional heat map sequence, defect identification is performed using the signal-to-noise ratio (SNR) index, and the signal in the defect area is compared with the signal in the non-defect area.

[0012] S6. Optimize and output the 3D heatmap sequence, including adjusting the contrast and brightness of the 3D heatmap sequence.

[0013] Preferably, S1 includes S11 and S12;

[0014] S11. Collect complete thermal imaging data under pulsed thermal excitation using an infrared thermal imager. The acquisition process includes multiple thermal imaging frames, and each frame records the instantaneous state of the material surface temperature change.

[0015] Based on the thermophysical properties of the material (such as thermal conductivity, specific heat capacity, density, etc.), geometric thickness, and the characteristics of the thermal excitation source (such as pulse width, energy density, pulse frequency, etc.), the acquisition process is optimized to ensure that the data has high quality and accuracy.

[0016] Specifically, select an appropriate sampling frequency based on the thermal response time of the material. For example, thicker materials require a longer sampling time to capture the thermal response of deep defects. Select excitation source parameters based on the thermal diffusion characteristics of the material and the depth of the defects, including pulse width (typically 5 milliseconds), energy density (e.g., 1000 J / m²), and excitation frequency (20 Hz).

[0017] S12. Denoise the acquired thermal imaging data to obtain a pulsed thermal imaging sequence;

[0018] Denoising is achieved by using a moving average window method: First, the size of the moving window is set to W time frames. For each pixel (x, y), a sliding time window of length W is constructed in the time series with that pixel as the center.

[0019] Then, the average temperature value of the pixel over W consecutive frames is calculated within each time window and replaced with the original temperature in the current time frame.

[0020] The moving window slides along the time axis, gradually updating the temporal response of each pixel throughout the entire thermal imaging sequence;

[0021] This method smooths out local abrupt changes and high-frequency disturbance fluctuations, while preserving the thermal diffusion trend and overall temperature evolution process of the material after stimulation.

[0022] Preferably, S2 includes S21 and S22;

[0023] S21. The pulsed thermal imaging sequence is divided into several sub-sequences by incremental multiplication relationship segmentation method. Each sub-sequence covers complete thermal signal information and the length of the sub-sequence is gradually increased, starting from 1 to 4 frames, gradually expanding to 1 to 8 frames and 1 to 12 frames, and finally reaching 1 to 4k frames.

[0024] The incremental multiplication relation segmentation method increases the number of frames in each subsequence incrementally, enabling each subsequence to effectively capture the changing trend of the thermal signal, while avoiding the information loss problem that may be caused by fixed-length subsequences in traditional methods.

[0025] The formula for splitting incremental multiplication relationships is as follows:

[0026] nk = 4 × k; where nk represents the length of the k-th subsequence, which gradually increases as the subsequence number k increases.

[0027] The incremental multiplication relationship segmentation formula ensures that the length of each subsequence increases continuously, covering more comprehensive thermal signal information and providing accurate data support for subsequent thermal signal reconstruction and defect analysis.

[0028] By using the incremental multiplication relationship segmentation method, the length of the subsequence is gradually expanded from one frame, ensuring the comprehensiveness of the data. In practical applications, the segmentation length is gradually increased according to the thermal response characteristics of thermal imaging images to ensure a comprehensive reflection of the material's thermal behavior. This incremental segmentation method overcomes the limitations of fixed-length subsequences in traditional methods, enabling each subsequence to dynamically adapt to changes in the thermal signal.

[0029] Preferably, S3 includes S31;

[0030] S31. Each subsequence is compressed and denoised by using the dynamic compression reconstruction DCR function;

[0031] Compression: The dynamic compression reconstruction DCR function removes redundant information from the subsequence data through a compression algorithm, retaining the data portion with thermal signal characteristics;

[0032] Denoising: Redundant noise, timing noise, and spatial noise are reduced by using the DCR function;

[0033] The noise reduction formula is as follows: ;

[0034] In the formula, zT(x, y, t) represents the denoised temperature value at position (x, y) in frame t, nT(x, y, t+wk) is the normalized temperature value at position (x, y) in frame t+wk, wk represents the time index offset within the window, and W represents the size of the moving average window.

[0035] Preferably, in step S32, based on the denoised subsequence data, the contrast is enhanced and the thermal signal is reconstructed using the DCR reconstruction function;

[0036] The Dynamic Compression Reconstruction (DCR) function reconstructs each compressed and denoised subsequence, restoring or enhancing the thermal signal characteristics of the defect region. The reconstruction function adjusts the amplitude of temperature changes in the subsequence by mathematically modeling the thermal signal, reducing interference from surrounding irrelevant noise.

[0037] The formula for the DCR reconstruction function is as follows:

[0038] ;

[0039] In the formula, T(n) represents the original pulse heatmap of the nth frame, F(k) represents the DCR heatmap of the kth frame, and k represents the frame number;

[0040] Contrast Enhancement: The Dynamic Compression Reconstruction (DCR) function enhances the contrast of the defect area, making the difference between the defect signal and the background noise more obvious and improving the accuracy of defect identification. During the contrast enhancement process, the temperature change of the reconstructed signal is amplified, especially in the defect area, where the peak of the defect signal is more obvious.

[0041] The formula for enhancing contrast is as follows:

[0042] ;

[0043] In the formula, C represents the contrast value, which measures the temperature difference between the defect area and the reference area; Tde represents the average temperature of the defect area; Tref represents the average temperature of the reference area; and σref represents the standard deviation of the reference area.

[0044] Preferably, S4 includes S41 and S42;

[0045] S41. The reconstructed sub-sequences processed by reconstruction are sequentially stitched into a complete heat map to obtain a three-dimensional heat map sequence; this sub-sequence image has had redundant noise removed and defect signals enhanced in the previous processing steps;

[0046] Based on the order of the reconstructed subsequences in the time series, they are pieced together one by one to form a complete heatmap sequence; the splicing order is based on the subsequence number.

[0047] The splicing formula is as follows: Htotal = [H1, H2, ..., Hk];

[0048] In the formula, Htotal represents the complete 3D heat map sequence after stitching, and H1, H2, ..., Hk represent the 1st to the kth reconstructed subsequence images, respectively;

[0049] S42. After stitching all the reconstructed subsequences into a complete three-dimensional heat map sequence, perform heat map analysis to identify the location and characteristics of defects. This analysis identifies possible defect areas by calculating indicators such as temperature change and contrast in each region, and further analyzes the defects.

[0050] Thermal image analysis identifies defective regions in the image by analyzing the stitched complete 3D thermal image sequence; by calculating the contrast of the defective regions, the location and severity of the defects are confirmed.

[0051] The contrast of the defect area is obtained as follows: First, the difference between the average temperature of the defect area and the average temperature of the reference area is calculated; then, the noise level is measured by the temperature standard deviation of the reference area. The larger the standard deviation, the stronger the noise and the lower the signal discrimination; finally, the difference in temperature between the defect areas is divided by the standard deviation of the reference area to obtain the contrast of the defect area.

[0052] The location and characteristics of defects are extracted by analyzing the stitched heat map, such as the depth, area, and thermal response characteristics of the defects.

[0053] The formula for defect feature extraction is as follows:

[0054] ;

[0055] In the formula, Adef represents the area of ​​the defective region. The indicator function indicates that if the contrast Cde of the defect area exceeds the preset contrast threshold Tthe, then the position (x, y) is part of the defect area.

[0056] Preferably, S5 includes S51 and S52;

[0057] S51. Based on the three-dimensional thermal image sequence, the defect area and the reference area are located. The defect area is selected by stitching thermal images. The area with large temperature change shows a different thermal response than the normal area.

[0058] The reference region is selected as a noise baseline by choosing a defect-free area; the temperature variation in the reference region is small, representing background noise; the signal-to-noise ratio (SNR) of the defective region is calculated by re-comparing the defective region with the reference region and using the temperature difference and the standard deviation of the reference region.

[0059] ;

[0060] In the formula, Tde represents the average temperature of the defect area, Tref represents the average temperature of the reference area, Nref represents the number of pixels in the reference area, and Trefi is the temperature value of the i-th pixel in the reference area.

[0061] Preferably, the signal of the defect is further extracted based on the signal-to-noise ratio (SNR) of the defect area, and its features are identified;

[0062] By comparing the signals from the defect area with those from the reference area, relevant information about the defect, such as its size, depth, and shape, can be extracted.

[0063] The steps for extracting defect signals are as follows: compare the acquired signal-to-noise ratio (SNR) with the preset SNR threshold (Tsnr) to extract the defect signals;

[0064] When the signal-to-noise ratio (SNR) is less than or equal to the signal-to-noise ratio threshold (Tsnr), it is considered a normal signal.

[0065] When the signal-to-noise ratio (SNR) is greater than the SNR threshold (Tsnr), it is considered a defect signal, and the region to which the defect signal belongs is marked as a defect region.

[0066] By calculating the signal-to-noise ratio (SNR) between the defective region and the reference region, defects and normal areas in a thermal image can be effectively distinguished. A high SNR value indicates that the defect signal is more significant relative to noise, thus making defect detection more accurate. During defect signal extraction and recognition, a contrast threshold can be used to clearly extract specific features of the defect, such as its size, location, and intensity.

[0067] Preferably, S6 includes S61 and S62;

[0068] S61. By adjusting the contrast and brightness in the three-dimensional heat map sequence, the visibility of the defect area is optimized, the defect features are highlighted, and the adjusted three-dimensional heat map sequence is obtained. The goal of contrast adjustment is to improve the significance of the defect signal relative to the background noise, while brightness adjustment ensures that the overall display effect of the heat map is clearer, which is convenient for subsequent processing and analysis.

[0069] Contrast adjustment highlights defect signals by increasing the temperature difference between defective and normal areas;

[0070] The contrast adjustment formula is as follows:

[0071] ;

[0072] In the formula, Cadj represents the adjusted contrast, c1 represents the amplification factor of the contrast adjustment, and c2 represents the brightness offset.

[0073] Brightness adjustment enhances the signal in the defect area by adjusting the overall brightness of the three-dimensional heat map sequence;

[0074] The brightness adjustment formula is as follows: Tadj = zT + γ;

[0075] In the formula, Tadj represents the adjusted image temperature, zT represents the denoising temperature value, and γ represents the brightness adjustment amount;

[0076] S62. Convert the adjusted 3D heatmap sequence into a visualization image and report;

[0077] The visualization process converts the adjusted 3D heatmap sequence into visualization and report formats and outputs them. In the image format, users can directly view the optimized heatmap data; in the report format, users can obtain detailed analysis results and data support.

[0078] By generating a report containing heatmaps and relevant analytical data, the report includes information on the location, size, and contrast of defective areas.

[0079] This invention provides a method for processing thermal signal data, which has the following beneficial effects:

[0080] (1) The present invention greatly simplifies the operation process by eliminating the need for users to perform tedious manual input and parameter selection. Traditional methods usually require complex parameter adjustments and preset reference areas. The automated processing method of the present invention reduces human intervention and lowers the technical requirements for operators, thereby making the use of thermal imaging technology more convenient and widespread.

[0081] Traditional methods often require processing each pixel sequence individually, resulting in slow processing speed and low efficiency. By using incremental multiplication relationship segmentation and dynamic compression reconstruction DCR function, this invention can directly process the entire thermal imaging sequence as a whole, eliminating redundant data, improving the contrast of defect signals, and significantly improving data processing efficiency and accuracy. This invention effectively eliminates noise in redundant data through the dynamic compression reconstruction DCR method, enhancing the signal contrast in defect areas. By applying the signal-to-noise ratio (SNR) metric, defect areas can be clearly distinguished from background noise, making defect detection more accurate and reducing noise interference with the results.

[0082] (2) By using the incremental multiplication relationship segmentation method, the pulsed thermal imaging sequence is divided into several subsequences, and the length of each subsequence gradually increases to ensure that each subsequence can effectively capture the complete changes in the thermal signal. This automated segmentation method not only avoids the information loss problem that may be caused by fixed-length subsequences in traditional methods, but also dynamically adjusts the length of each subsequence according to the changes in the thermal signal, thereby improving the flexibility and efficiency of data processing.

[0083] By verifying the temperature difference ratio ΔT of each expanded subsequence, it can be determined whether the number of frames in the subsequence is sufficient, ensuring that each subsequence can fully reflect the changes in the thermal signal. If the number of frames in the subsequence is insufficient, the system will automatically expand it to ensure the integrity and validity of the data. This method avoids unnecessary redundant data interference and improves the accuracy of subsequent defect detection and signal analysis.

[0084] (3) By sequentially stitching together the reconstructed subsequences, a complete 3D heatmap sequence can be generated efficiently. The stitching operation is performed according to the order of the subsequences in the time series, ensuring the temporal continuity and spatial consistency of the heatmap sequence. Compared with the inefficient mode of processing single pixel sequences one by one in the traditional method, this invention improves data processing efficiency by stitching together the whole, avoids repeated calculations, reduces the computational burden, and improves the speed and accuracy of data processing.

[0085] The stitched sequence of complete 3D thermal images, when analyzed, can accurately identify the location and characteristics of defective regions. Contrast calculations effectively distinguish defective and normal regions, and further extract the depth, area, and thermal response characteristics of the defects.

[0086] (4) By averaging the data over time using a sliding window, random noise, high-frequency jitter, and transient interference signals generated during thermal imaging can be effectively suppressed, thereby improving the signal-to-noise ratio of the thermal image sequence and providing a clearer data foundation for subsequent defect signal extraction and discrimination. The denoising process weakens the impact of high-frequency outliers such as local excitation fluctuations and occasional hot spots on the stability of the overall thermal imaging sequence by averaging the temperature time series, and enhances the robustness of subsequent feature extraction and classification algorithms in multi-frame overlay analysis.

[0087] The moving average method not only improves the clarity of static heatmaps, but also lays the data foundation for constructing multi-time temperature evolution curves, making the time response characteristics of defects (such as response delay, location of maximum temperature rise, and diffusion range) clearer, which helps to achieve quantitative assessment of deep defects based on thermal dynamic behavior. Attached Figure Description

[0088] Figure 1 This is a schematic diagram of the steps of a thermal signal data processing method according to the present invention;

[0089] Figure 2 This is a schematic diagram illustrating the principle of a pulsed thermal imaging sequence in a thermal signal data processing method according to the present invention.

[0090] Figure 3 This is a schematic diagram showing the temperature changes between the defect center region and the non-defect region of the present invention. Detailed Implementation

[0091] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0092] Example 1

[0093] This invention provides a method for processing thermal signal data. Please refer to [link / reference]. Figures 1 to 3 This includes the following steps:

[0094] S1. Collect complete thermal imaging data using an infrared thermal imager and perform noise reduction processing using a moving average window method to obtain a pulsed thermal imaging sequence.

[0095] S2. By using the incremental multiplication relationship segmentation method, the pulse thermal imaging sequence is segmented to obtain several sub-sequences;

[0096] S3. By using the dynamic compression reconstruction DCR function to process each subsequence, noise in redundant data is eliminated, the signal contrast of defects is improved, and the reconstructed subsequence is obtained.

[0097] S4. Combine the reconstructed subsequences into a reconstructed heatmap, and then combine them in sequence into a complete three-dimensional heatmap sequence.

[0098] S5. Based on the three-dimensional heat map sequence, defect identification is performed using the signal-to-noise ratio (SNR) index, and the signal in the defect area is compared with the signal in the non-defect area.

[0099] S6. Optimize and output the 3D heatmap sequence, including adjusting the contrast and brightness of the 3D heatmap sequence.

[0100] In this embodiment, the present invention greatly simplifies the operation process by eliminating the need for tedious manual input and parameter selection by the user. Traditional methods typically require complex parameter adjustments and preset reference areas. The automated processing method of the present invention reduces human intervention and lowers the technical skill requirements for operators, thereby making the use of thermal imaging technology more convenient and widespread.

[0101] Traditional methods often require processing each pixel sequence individually, resulting in slow processing speed and low efficiency. By using incremental multiplication relationship segmentation and dynamic compression reconstruction DCR function, this invention can directly process the entire thermal imaging sequence as a whole, eliminating redundant data, improving the contrast of defect signals, and significantly improving data processing efficiency and accuracy. This invention effectively eliminates noise in redundant data through the dynamic compression reconstruction DCR method, enhancing the signal contrast in defect areas. By applying the signal-to-noise ratio (SNR) metric, defect and non-defect areas can be clearly distinguished, making defect detection more accurate and reducing noise interference with the results.

[0102] By adjusting the contrast and brightness of the 3D thermal image sequence, this invention optimizes the visibility and clarity of the thermal images, making defect areas more prominent and easier to identify. This optimization step not only improves the visual effect of the thermal images but also helps subsequent analysts extract and identify defect features more efficiently. Existing thermal imaging detection methods typically require prior information about the test sample or reference area, while this invention does not rely on this information and can complete the processing of the entire thermal imaging sequence without additional prior data. This feature makes the method more adaptable and applicable to a wider range of scenarios, especially enabling accurate defect detection even in the absence of sufficient prior knowledge.

[0103] Example 2

[0104] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically: S1 includes S11 and S12;

[0105] S11. Collect complete thermal imaging data under pulsed thermal excitation using an infrared thermal imager. The acquisition process includes multiple thermal imaging frames, and each frame records the instantaneous state of the material surface temperature change.

[0106] Specifically, select an appropriate sampling frequency based on the material's thermal response time; select excitation source parameters, including pulse width, energy density, and excitation frequency, based on the material's thermal diffusion characteristics and defect depth.

[0107] S12. Denoise the acquired thermal imaging data to obtain a pulsed thermal imaging sequence;

[0108] Denoising is achieved by using a moving average window method: First, the size of the moving window is set to W time frames. For each pixel (x, y), a sliding time window of length W is constructed in the time series with that pixel as the center.

[0109] Then, the average temperature value of the pixel over W consecutive frames is calculated within each time window and replaced with the original temperature in the current time frame.

[0110] The moving window slides along the time axis, gradually updating the temporal response of each pixel throughout the entire thermal imaging sequence;

[0111] This method smooths out local abrupt changes and high-frequency disturbance fluctuations, while preserving the thermal diffusion trend and overall temperature evolution process of the material after stimulation.

[0112] In this embodiment, by selecting appropriate sampling frequencies and excitation source parameters, it is ensured that the instantaneous state reflecting the temperature change of the material surface can be accurately acquired under pulsed thermal excitation. Based on factors such as the material's thermal response time, thermal diffusion characteristics, and defect depth, appropriate pulse widths, energy densities, and excitation frequencies are selected, enabling the thermal imaging sequence to more comprehensively and accurately reflect thermal response information. This ensures that the acquired thermal imaging data is more representative and reflects defect signals more clearly.

[0113] By optimizing the sampling frequency and excitation source parameters based on material properties and defect characteristics, thermal imaging data can be more comprehensively and accurately reflected in thermal response, especially when facing complex thermal diffusion and deep defects, providing more reliable data support.

[0114] This embodiment constructs a sliding time window for each pixel and combines it with the time response characteristics of the material after stimulation to locally smooth temperature fluctuations in different frames. This effectively preserves the true evolution trend of the material during thermal diffusion and reduces thermal image distortion caused by abnormal fluctuations in a single frame. Compared with frequency domain filtering, the moving average window method maintains the main trend of thermal diffusion while avoiding edge effects or frequency aliasing caused by frequency domain truncation. It is particularly suitable for the continuous interpretation of surface and shallow defects, improving the coherence and stability of the defect identification process.

[0115] By averaging multiple frames of data over time using a sliding window, random noise, high-frequency jitter, and transient interference signals generated during thermal image acquisition can be effectively suppressed, thereby improving the signal-to-noise ratio of the thermal image sequence and providing a clearer data foundation for subsequent defect signal extraction and discrimination. The denoising process, through moving average of the temperature time series, weakens the impact of high-frequency outliers such as local excitation fluctuations and occasional hot spots on the overall stability of the thermal imaging sequence, enhancing the robustness of subsequent feature extraction and classification algorithms in multi-frame overlay analysis.

[0116] The moving average method not only improves the clarity of static heatmaps, but also lays the data foundation for constructing multi-time temperature evolution curves, making the time response characteristics of defects (such as response delay, location of maximum temperature rise, and diffusion range) clearer, which helps to achieve quantitative assessment of deep defects based on thermal dynamic behavior.

[0117] Example 3

[0118] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 2 and Figure 3 Specifically: S2 includes S21;

[0119] S21. The pulsed thermal imaging sequence is divided into several sub-sequences by incremental multiplication relationship segmentation method. Each sub-sequence covers complete thermal signal information and the length of the sub-sequence is gradually increased, starting from 1 to 4 frames, gradually expanding to 1 to 8 frames and 1 to 12 frames, and finally reaching 1 to 4k frames.

[0120] The incremental multiplication relation segmentation method increases the number of frames in each subsequence incrementally, enabling each subsequence to effectively capture the changing trend of the thermal signal, while avoiding the information loss problem that may be caused by fixed-length subsequences in traditional methods.

[0121] The formula for splitting incremental multiplication relationships is as follows:

[0122] nk = 4 × k; where nk represents the length of the k-th subsequence, which gradually increases as the subsequence number k increases.

[0123] S3 includes S31 and S32;

[0124] S31. Each subsequence is compressed and denoised by using the dynamic compression reconstruction DCR function;

[0125] Compression: The dynamic compression reconstruction DCR function removes redundant information from the subsequence data through a compression algorithm, retaining the data portion with thermal signal characteristics;

[0126] Denoising: Redundant noise, timing noise, and spatial noise are reduced by using the DCR function.

[0127] The noise reduction formula is as follows: ;

[0128] In the formula, zT(x, y, t) represents the denoised temperature value at position (x, y) in frame t, nT(x, y, t+wk) is the normalized temperature value at position (x, y) in frame t+wk, wk represents the time index offset within the window, and W represents the size of the moving average window.

[0129] S32. Based on the denoised subsequence data, the contrast is enhanced and the thermal signal is reconstructed using the DCR reconstruction function;

[0130] The Dynamic Compression Reconstruction (DCR) function reconstructs each compressed and denoised subsequence, restoring or enhancing the thermal signal characteristics of the defect region. The reconstruction function adjusts the amplitude of temperature changes in the subsequence by mathematically modeling the thermal signal, reducing interference from surrounding irrelevant noise.

[0131] The formula for the DCR reconstruction function is as follows:

[0132] ;

[0133] In the formula, T(n) represents the original pulse heatmap of the nth frame, F(k) represents the DCR heatmap of the kth frame, and k represents the frame number;

[0134] Contrast Enhancement: The Dynamic Compression Reconstruction (DCR) function enhances the contrast of the defect area, making the difference between the defect signal and the background noise more obvious and improving the accuracy of defect identification.

[0135] The formula for enhancing contrast is as follows:

[0136] ;

[0137] In the formula, C represents the contrast value, Tde represents the average temperature of the defect area, Tref represents the average temperature of the reference area, and σref represents the standard deviation of the reference area.

[0138] In this embodiment, the pulsed thermal imaging sequence is divided into several subsequences using an incremental multiplication relationship segmentation method, with the length of each subsequence gradually increasing to ensure that each subsequence can effectively capture the complete changes in the thermal signal. This automated segmentation method not only avoids the information loss problem that may be caused by fixed-length subsequences in traditional methods, but also dynamically adjusts the length of each subsequence according to the changes in the thermal signal, thereby improving the flexibility and efficiency of data processing.

[0139] By verifying the temperature difference ratio ΔT of each expanded subsequence, it can be determined whether the number of frames in the subsequence is sufficient, ensuring that each subsequence can fully reflect the changes in the thermal signal. If the number of frames in the subsequence is insufficient, the system will automatically expand it to ensure the integrity and validity of the data. This method avoids unnecessary redundant data interference and improves the accuracy of subsequent defect detection and signal analysis.

[0140] The Dynamic Compression Reconstruction (DCR) function is used to denoise and compress each subsequence, effectively removing interference from external noise, equipment instability, and non-uniform thermal response. Furthermore, the DCR function retains key information of the thermal signal by compressing redundant data, reducing unnecessary data volume and improving data processing efficiency. In addition to compressing and denoising the data, the DCR function also enhances the contrast of defect areas, making the difference between the defect signal and background noise more apparent. This contrast enhancement technique improves the accuracy of defect identification, making defect areas more prominent and facilitating subsequent analysis and processing. This function is particularly suitable for complex or noisy thermal imaging data, improving the accuracy and reliability of defect detection.

[0141] Specific examples: refer to Figure 3 ;

[0142] To preliminarily verify the feasibility of the DCR method, we used the DCR method to post-process the normalized temperature signals of the defect center region and non-defect regions, and obtained the reconstructed signal, such as... Figure 3 As shown in b. Subsequently, we plotted the normalized temperature contrast curve and the reconstructed signal contrast curve, as shown in Figures 1-2. Figure 3 c and Figure 3 As shown in d;

[0143] according to Figure 3 As can be seen from b, the reconstructed thermal signal curve corresponding to the defect has a peak point, while no such peak is observed in the curve of the intact region. Therefore, when extracting the proposed thermal features (peaks), it is not necessary to consider the intact signal. According to... Figure 3 As can be seen from d, the contrast of the reconstructed thermal signal corresponding to the defect has been reversed, with positive values ​​becoming negative. We can identify defects based on whether the contrast of the reconstructed thermal signal has been reversed (the intact area needs to be used as a reference).

[0144] By comparison Figure 3 c and Figure 3 As can be seen from point d, the contrast curve of the reconstructed thermal signal is smoother than the temperature contrast curve, indicating that the temporal noise in the temperature-time series signal is effectively suppressed after processing by the DCR method. Furthermore, for the reconstructed thermal signal, the maximum contrast of the defect occurs at frame 65 of the image, meaning that only 260 frames of the original thermal signal are needed to reconstruct the maximum contrast.

[0145] However, for the original thermal signal, the maximum contrast of the defect appears in frame 445. The above analysis indicates that the maximum contrast of the defect appears earlier in the reconstructed thermal image, thus effectively reducing the blurring of defect edges.

[0146] Example 4

[0147] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 Specifically: S4 includes S41 and S42;

[0148] S41. Sequentially stitch the reconstructed sub-sequences of the reconstruction process into a complete heat map to obtain a three-dimensional heat map sequence;

[0149] Based on the order of the reconstructed subsequences in the time series, they are pieced together one by one to form a complete heatmap sequence; the splicing order is based on the subsequence number.

[0150] The splicing formula is as follows: Htotal = [H1, H2, ..., Hk];

[0151] In the formula, Htotal represents the complete 3D heat map sequence after stitching, and H1, H2, ..., Hk represent the 1st to the kth reconstructed subsequence images, respectively;

[0152] S42. After splicing all the reconstructed subsequences into a complete three-dimensional heat map sequence, perform heat map analysis to identify the location and characteristics of the defects;

[0153] Thermal image analysis identifies defective regions in the image by analyzing the stitched complete 3D thermal image sequence; by calculating the contrast of the defective regions, the location and severity of the defects are confirmed.

[0154] The contrast of the defect area is obtained as follows: First, the difference between the average temperature of the defect area and the average temperature of the reference area is calculated; then, the noise level is measured by the temperature standard deviation of the reference area. The larger the standard deviation, the stronger the noise and the lower the signal discrimination; finally, the difference in temperature between the defect areas is divided by the standard deviation of the reference area to obtain the contrast of the defect area.

[0155] The location and characteristics of defects are extracted by analyzing the stitched heat map, such as the depth, area, and thermal response characteristics of the defects.

[0156] The formula for defect feature extraction is as follows:

[0157] ;

[0158] In the formula, Adef represents the area of ​​the defective region. The indicator function indicates that if the contrast Cde of the defect area exceeds the preset contrast threshold Tthe, then the position (x, y) is part of the defect area.

[0159] S5 includes S51 and S52;

[0160] S51. Based on the three-dimensional thermal image sequence, the defect area and the reference area are located. The defect area is selected by stitching thermal images. The area with large temperature change shows a different thermal response than the normal area.

[0161] The reference region is selected as a noise baseline by selecting a defect-free area; the signal-to-noise ratio (SNR) of the defective region is calculated by re-comparing the defective region with the reference region and using the temperature difference and the standard deviation of the reference region.

[0162] ;

[0163] In the formula, Tde represents the average temperature of the defect area, Tref represents the average temperature of the reference area, Nref represents the number of pixels in the reference area, and Trefi is the temperature value of the i-th pixel in the reference area.

[0164] In this embodiment, a complete 3D heatmap sequence can be efficiently generated by sequentially stitching together the reconstructed subsequences. The stitching operation is performed according to the order of the subsequences in the time series, ensuring the temporal continuity and spatial consistency of the heatmap sequence. Compared with the inefficient traditional method of processing single-pixel sequences one by one, this invention improves data processing efficiency through overall stitching, avoids redundant calculations, reduces computational burden, and improves the speed and accuracy of data processing.

[0165] The stitched sequence of complete 3D thermal images, when analyzed, accurately identifies the location and characteristics of defective regions. Contrast calculation effectively distinguishes defective and normal regions, further extracting the depth, area, and thermal response characteristics of defects. Compared to traditional methods, this invention, through automated analysis and calculation, achieves more precise defect identification, rapidly locating defects and assessing their severity, significantly improving the efficiency and reliability of defect detection. By calculating the signal-to-noise ratio (SNR) between the defective and reference regions, this invention effectively improves the accuracy of defect signal identification. The SNR calculation considers not only the temperature difference between the defective and reference regions but also measures the noise level based on the standard deviation of the reference region, ensuring a more distinct difference between the defective signal and background noise. In this way, the detection system can stably identify defects even in noisy environments, improving the system's robustness and accuracy.

[0166] This embodiment reduces the need for manual operation by automating 3D thermal image stitching, defect analysis, and signal-to-noise ratio calculation. Users do not need to manually select reference areas or perform complex parameter settings, making the operation simpler and more intuitive. This feature makes the method have a low technical threshold, suitable for widespread application in actual production, reducing operational difficulty and human error, and improving the automation and efficiency of detection.

[0167] Example 5

[0168] This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 Specifically: S52, based on the obtained signal-to-noise ratio (SNR) of the defect area, further extract the defect signal and identify its features;

[0169] By comparing the signals from the defect area with those from the reference area, relevant information about the defect, such as its size, depth, and shape, can be extracted.

[0170] The steps for extracting defect signals are as follows: compare the acquired signal-to-noise ratio (SNR) with the preset SNR threshold (Tsnr) to extract the defect signals;

[0171] When the signal-to-noise ratio (SNR) is less than or equal to the signal-to-noise ratio threshold (Tsnr), it is considered a normal signal.

[0172] When the signal-to-noise ratio (SNR) is greater than the SNR threshold (Tsnr), it is considered a defect signal, and the region to which the defect signal belongs is marked as a defect region.

[0173] S6 includes S61 and S62;

[0174] S61. By adjusting the contrast and brightness in the three-dimensional heat map sequence, the visibility of the defect area is optimized, the defect features are highlighted, and the adjusted three-dimensional heat map sequence is obtained.

[0175] Contrast adjustment highlights defect signals by increasing the temperature difference between defective and normal areas;

[0176] The contrast adjustment formula is as follows:

[0177] ;

[0178] In the formula, Cadj represents the adjusted contrast, c1 represents the amplification factor of the contrast adjustment, and c2 represents the brightness offset.

[0179] Brightness adjustment enhances the signal in the defect area by adjusting the overall brightness of the three-dimensional heat map sequence;

[0180] The brightness adjustment formula is as follows: Tadj = zT + γ;

[0181] In the formula, Tadj represents the adjusted image temperature, zT represents the denoising temperature value, and γ represents the brightness adjustment amount;

[0182] S62. Convert the adjusted 3D heatmap sequence into a visualization image and report;

[0183] The visualization process converts the adjusted 3D heatmap sequence into visualization and report formats and outputs them. In the image format, users can directly view the optimized heatmap data; in the report format, users can obtain detailed analysis results and data support.

[0184] The report generates a heatmap and related analysis data, including information on the location, size, and contrast of the defective area.

[0185] In this embodiment, an automated signal extraction method based on a signal-to-noise ratio (SNR) threshold effectively distinguishes between defect signals and normal signals. When the SNR exceeds a preset threshold (Tsnr), the system automatically identifies the defect signal and marks the corresponding region as a defect area. This automated signal extraction not only improves the accuracy of defect identification but also avoids potential manual intervention and human error in traditional methods. By extracting the defect signal, the system can accurately identify the size, depth, and shape of the defect, providing more precise data support for subsequent analysis.

[0186] This embodiment significantly improves the visibility of defect areas by adjusting the contrast and brightness of the 3D heatmap sequence. Contrast adjustment enhances the temperature difference between defect and normal areas, making defect signals more prominent and facilitating subsequent analysis and decision-making. Brightness adjustment ensures that the signal in the defect area is sufficiently brightened, allowing users to clearly observe defects in the heatmap. Especially in low-contrast or noisy conditions, the optimized image effect is more obvious, enhancing the visual appeal of the image.

[0187] By adjusting the contrast and brightness of the 3D heatmap sequence, not only was the visual effect of the heatmap optimized, but the clarity of defect signals was also enhanced, making them easier to identify. The optimized heatmap not only helps to visually display defects but also improves the accuracy of defect analysis. Especially in complex backgrounds or situations with little temperature variation, the optimized heatmap provides a clearer display of defect areas, helping analysts better locate defects and assess their severity.

[0188] Based on the optimized 3D heatmap sequence, this embodiment generates a visualized image containing defect information and a detailed analysis report. The report not only includes the optimized heatmap but also provides information such as the location, size, and contrast of the defect area. This detailed report output format allows users to fully understand the defect situation and provides data support for subsequent decision-making and processing. Furthermore, the dual output of image and report formats ensures that the needs of different user groups are met, providing convenient support for both those who need to quickly view heatmap data and technical personnel who require in-depth analysis.

[0189] The entire process, through automated defect identification, signal extraction, image optimization, and report generation, greatly simplifies the operation. Users only need to focus on the final output of the image and report, without the need for complex parameter settings or manual intervention. This simplification not only reduces operational difficulty but also enhances the user experience, making the application of thermal imaging technology more convenient, especially suitable for industrial automation and other scenarios requiring rapid and efficient processing of large amounts of thermal image data.

[0190] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for processing thermal signal data, characterized in that: Includes the following steps: S1. Collect complete thermal imaging data using an infrared thermal imager and perform noise reduction processing using a moving average window method to obtain a pulsed thermal imaging sequence. S2. By using the incremental multiplication relationship segmentation method, the pulse thermal imaging sequence is segmented to obtain several sub-sequences; S3. By using the dynamic compression reconstruction DCR function to process each subsequence, noise in redundant data is eliminated, the signal contrast of defects is improved, and the reconstructed subsequence is obtained. S4. Combine the reconstructed subsequences into a reconstructed heatmap, and then combine them in sequence into a complete three-dimensional heatmap sequence. S5. Based on the three-dimensional heat map sequence, defect identification is performed using the signal-to-noise ratio (SNR) index, and the signal in the defect area is compared with the signal in the non-defect area. S6. Optimize and output the 3D heatmap sequence, including adjusting the contrast and brightness of the 3D heatmap sequence.

2. The thermal signal data processing method according to claim 1, characterized in that: S1 includes S11 and S12; S11. Collect complete thermal imaging data under pulsed thermal excitation using an infrared thermal imager. The acquisition process includes multiple thermal imaging frames, and each frame records the instantaneous state of the material surface temperature change. Specifically, select an appropriate sampling frequency based on the material's thermal response time; select excitation source parameters, including pulse width, energy density, and excitation frequency, based on the material's thermal diffusion characteristics and defect depth. S12. Denoise the acquired thermal imaging data to obtain a pulsed thermal imaging sequence; Denoising is achieved by using a moving average window method: First, the size of the moving window is set to W time frames. For each pixel (x, y), a sliding time window of length W is constructed in the time series with that pixel as the center. Then, the average temperature value of the pixel over W consecutive frames is calculated within each time window and replaced with the original temperature in the current time frame. The moving window slides along the time axis, gradually updating the temporal response of each pixel throughout the entire thermal imaging sequence; This method smooths out local abrupt changes and high-frequency disturbance fluctuations, while preserving the thermal diffusion trend and overall temperature evolution process of the material after stimulation.

3. The thermal signal data processing method according to claim 2, characterized in that: S2 includes S21; S21. The pulsed thermal imaging sequence is divided into several sub-sequences by incremental multiplication relationship segmentation method. Each sub-sequence covers complete thermal signal information and the length of the sub-sequence is gradually increased, starting from 1 to 4 frames, gradually expanding to 1 to 8 frames and 1 to 12 frames, and finally reaching 1 to 4k frames. The incremental multiplication relation segmentation method increases the number of frames in each subsequence incrementally, enabling each subsequence to effectively capture the changing trend of the thermal signal, while avoiding the information loss problem that may be caused by fixed-length subsequences in traditional methods. The formula for splitting incremental multiplication relationships is as follows: nk = 4 × k; where nk represents the length of the k-th subsequence, which gradually increases as the subsequence number k increases.

4. The thermal signal data processing method according to claim 3, characterized in that: S3 includes S31 and S32; S31. Each subsequence is compressed and denoised by using the dynamic compression reconstruction DCR function; Compression: The dynamic compression reconstruction DCR function removes redundant information from the subsequence data through a compression algorithm, retaining the data portion with thermal signal characteristics; Denoising: Redundant noise, timing noise, and spatial noise are reduced by using the DCR function; The noise reduction formula is as follows: ; In the formula, zT(x, y, t) represents the denoised temperature value at position (x, y) in frame t, nT(x, y, t+wk) is the normalized temperature value at position (x, y) in frame t+wk, wk represents the time index offset within the window, and W represents the size of the moving average window.

5. A thermal signal data processing method according to claim 4, characterized in that: S32. Based on the denoised subsequence data, the contrast is enhanced and the thermal signal is reconstructed using the DCR reconstruction function; The Dynamic Compression Reconstruction (DCR) function reconstructs each compressed and denoised subsequence, restoring or enhancing the thermal signal characteristics of the defect region. The reconstruction function adjusts the amplitude of temperature changes in the subsequence by mathematically modeling the thermal signal, reducing interference from surrounding irrelevant noise. The formula for the DCR reconstruction function is as follows: ； In the formula, T(n) represents the original pulse heatmap of the nth frame, F(k) represents the DCR heatmap of the kth frame, and k represents the frame number; Contrast Enhancement: The Dynamic Compression Reconstruction (DCR) function enhances the contrast of the defect area, making the difference between the defect signal and the background noise more obvious and improving the accuracy of defect identification. The formula for enhancing contrast is as follows: ； In the formula, C represents the contrast value, Tde represents the average temperature of the defect area, Tref represents the average temperature of the reference area, and σref represents the standard deviation of the reference area.

6. The thermal signal data processing method according to claim 1, characterized in that: S4 includes S41 and S42; S41. Sequentially stitch the reconstructed sub-sequences of the reconstruction process into a complete heat map to obtain a three-dimensional heat map sequence; Based on the order of the reconstructed subsequences in the time series, they are pieced together one by one to form a complete heatmap sequence; the splicing order is based on the subsequence number. The splicing formula is as follows: Htotal = [H1, H2, ..., Hk]; In the formula, Htotal represents the complete 3D heat map sequence after stitching, and H1, H2, ..., Hk represent the 1st to the kth reconstructed subsequence images, respectively; S42. After splicing all the reconstructed subsequences into a complete three-dimensional heat map sequence, perform heat map analysis to identify the location and characteristics of the defects; Thermal image analysis identifies defective regions in the image by analyzing the stitched complete 3D thermal image sequence; by calculating the contrast of the defective regions, the location and severity of the defects are confirmed. The contrast of the defect area is obtained as follows: First, the difference between the average temperature of the defect area and the average temperature of the reference area is calculated; then, the noise level is measured by the temperature standard deviation of the reference area. The larger the standard deviation, the stronger the noise and the lower the signal discrimination; finally, the difference in temperature between the defect areas is divided by the standard deviation of the reference area to obtain the contrast of the defect area. The location and characteristics of defects are extracted by analyzing the stitched heat map, such as the depth, area, and thermal response characteristics of the defects. The formula for defect feature extraction is as follows: ； In the formula, Adef represents the area of ​​the defective region. The indicator function indicates that if the contrast Cde of the defect area exceeds the preset contrast threshold Tthe, then the position (x, y) is part of the defect area.

7. A thermal signal data processing method according to claim 6, characterized in that: S5 includes S51 and S52; S51. Based on the three-dimensional thermal image sequence, the defect area and the reference area are located. The defect area is selected by stitching thermal images. The area with large temperature change shows a different thermal response than the normal area. The reference region is selected as a noise baseline by selecting a defect-free area; the signal-to-noise ratio (SNR) of the defective region is calculated by re-comparing the defective region with the reference region and using the temperature difference and the standard deviation of the reference region. ； In the formula, Tde represents the average temperature of the defect area, Tref represents the average temperature of the reference area, Nref represents the number of pixels in the reference area, and Trefi is the temperature value of the i-th pixel in the reference area.

8. A thermal signal data processing method according to claim 7, characterized in that: S52. Based on the obtained signal-to-noise ratio (SNR) of the defect area, further extract the signal of the defect and identify its features; By comparing the signals from the defect area with those from the reference area, relevant information about the defect, such as its size, depth, and shape, can be extracted. The steps for extracting defect signals are as follows: compare the acquired signal-to-noise ratio (SNR) with the preset SNR threshold (Tsnr) to extract the defect signals; When the signal-to-noise ratio (SNR) is less than or equal to the signal-to-noise ratio threshold (Tsnr), it is considered a normal signal. When the signal-to-noise ratio (SNR) is greater than the SNR threshold (Tsnr), it is considered a defect signal, and the region to which the defect signal belongs is marked as a defect region.

9. A thermal signal data processing method according to claim 8, characterized in that: S6 includes S61 and S62; S61. By adjusting the contrast and brightness in the three-dimensional heat map sequence, the visibility of the defect area is optimized, the defect features are highlighted, and the adjusted three-dimensional heat map sequence is obtained. Contrast adjustment highlights defect signals by increasing the temperature difference between defective and normal areas; The contrast adjustment formula is as follows: ； In the formula, Cadj represents the adjusted contrast, c1 represents the amplification factor of the contrast adjustment, and c2 represents the brightness offset. Brightness adjustment enhances the signal in the defect area by adjusting the overall brightness of the three-dimensional heat map sequence; The brightness adjustment formula is as follows: Tadj = zT + γ; In the formula, Tadj represents the adjusted image temperature, zT represents the denoising temperature value, and γ represents the brightness adjustment amount; S62. Convert the adjusted 3D heatmap sequence into a visualization image and report; The visualization process converts the adjusted 3D heatmap sequence into visualization and report formats and outputs them. In the image format, users can directly view the optimized heatmap data; in the report format, users can obtain detailed analysis results and data support. The report generates a heatmap and related analysis data, including information on the location, size, and contrast of the defective area.

Images