A tank truck leakage detection method based on thermal imaging and acoustic imaging fusion

Abstract

The present invention discloses a tank truck leakage detection method based on the fusion of thermal imaging and acoustic wave imaging, which relates to the fields of industrial safety monitoring, nondestructive testing technology, and image and signal processing technology, and includes the following steps: using a microphone array acquisition system to collect the acoustic wave signal of the tank truck; collecting the visible light image of the tank truck at the same position and angle; fusing the acoustic wave signal with the visible light image to generate an acoustic wave image through a data processing algorithm, wherein the acoustic wave abnormal area is marked; using an infrared thermal imager to take an infrared image of the tank truck at the same position and angle; and geometrically correcting, registering, and feature-level fusion the infrared image and the acoustic wave image. The present invention uses the infrared thermal imager to perform rapid and large-scale temperature anomaly detection, and provides accurate leakage positioning through acoustic wave imaging. Finally, the infrared image and the acoustic wave image are feature-level fused after geometric correction and image registration, thereby locating the tank leakage position.

Description

Technical Field

[0001] The invention relates to the fields of industrial safety monitoring, nondestructive testing technology and image and signal processing technology, and specifically to a tank truck leakage detection method based on the fusion of thermal imaging and acoustic wave imaging. Background Art

[0002] Dangerous chemicals such as gasoline, toluene, sulfuric acid, liquid ammonia, etc. are indispensable raw materials for the petrochemical industry. A tank truck is a normal pressure metal tank truck used to transport dangerous chemicals. As the main means of transporting flammable, explosive and toxic liquefied gases, the safety of tank trucks will directly affect the transportation safety of the entire railway or road. Due to defects in the manufacturing process, stress concentration and long-term fatigue loads, cracks will appear on the inner wall of the tank, causing chemical leakage. Therefore, it is particularly important to detect defects and leaks in the tank body of the tank truck.

[0003] At present, the commonly used detection technologies are mainly divided into two types: using thermal imaging technology to detect the difference in thermal radiation on the surface of an object. In the field of tank truck leak detection, this technology can effectively identify temperature changes caused by chemical or fuel leaks. The liquid or gas caused by the leak will form an abnormal temperature area on the surface of the tank, and the thermal imager can capture these anomalies and convert them into a visual thermal image. However, the main limitation of thermal imaging technology is that it may be affected by ambient temperature fluctuations, and in some cases it is impossible to accurately locate the specific location of the leak source, especially when the leak area is small or the temperature of the leaking material is close to the ambient temperature; using highly sensitive acoustic wave sensors to capture sound waves (especially ultrasonic waves) caused by pressure or liquid leakage. Acoustic wave sensors are deployed at key locations of tank trucks, and by analyzing the intensity and propagation path of the acoustic wave signal, the leak point can be accurately located. However, this technology may encounter problems with background noise interference in noisy environments, and its effectiveness may be limited for tank trucks made of non-metallic materials. Summary of the invention

[0004] In response to the above problems, the present invention provides a tank truck leakage detection method based on the fusion of thermal imaging and acoustic wave imaging. An infrared thermal imager is used to perform rapid and large-scale temperature anomaly detection, and acoustic wave imaging is used to provide accurate leakage positioning. Finally, the infrared image and the acoustic wave image are feature-level fused after geometric correction and image registration, thereby locating the tank leakage position.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] 1. A tank truck leakage detection method based on the fusion of thermal imaging and acoustic imaging, comprising the following steps:

[0007] Step 1: Use a microphone array acquisition system to collect the sound wave signal of the tanker truck;

[0008] Step 2: Collect visible light images of the tanker truck at the same position and angle;

[0009] Step 3: The acoustic wave signal is fused with the visible light image through a data processing algorithm to generate an acoustic wave image, in which the abnormal acoustic wave area is marked;

[0010] Step 4: Use an infrared thermal imager to take an infrared image of the tanker truck at the same position and angle;

[0011] Step 5: Perform geometric correction, registration and feature-level fusion of the infrared image and the acoustic wave image.

[0012] The beneficial effect of the above scheme is: it provides a method for detecting tanker leakage by comprehensively utilizing thermal imaging and acoustic imaging technology and integrating these two imaging methods, thereby enhancing the accuracy and reliability of detection.

[0013] The specific working principle adopted is: first, the acoustic wave signal of the tank truck is collected using a microphone array, and then the visible light image of the tank truck is collected at the same position and angle. Through a specific data processing algorithm, the acoustic wave signal is fused with the visible light image to generate an acoustic wave image, and the acoustic wave abnormal area is marked in it. Next, an infrared image of the tank truck is taken with an infrared thermal imager, and the infrared image and the acoustic wave image are geometrically corrected, registered, and feature-level fused to achieve accurate detection of leaks.

[0014] 2. In the preferred implementation, the microphone array acquisition system includes a 64-channel microphone board, a USB acquisition board and a 1080P camera.

[0015] The beneficial effects of the above implementation are: by adopting a microphone array acquisition system including a 64-channel microphone board, a USB acquisition board and a 1080P camera, the high density and high definition of the sound wave signal acquisition are ensured, providing high-quality raw data for subsequent data processing and analysis. The microphone array acquisition system simultaneously acquires the sound wave signal through the 64-channel microphone board, the USB acquisition board is responsible for digital processing of the acquired signal, and the 1080P camera synchronously acquires the visible light image of the tanker truck, ensuring the consistency of the sound wave signal and the visible light image in time and space.

[0016] 3. In a preferred implementation, the microphone data is processed by the FPGA, converted into PCM format, and transmitted to the PC via USB or Gigabit Ethernet.

[0017] The beneficial effects of the above implementation are: using FPGA to process microphone data and convert it into PCM format, the data transmission efficiency and processing speed are improved, while ensuring the quality and integrity of the data.

[0018] The specific working principle is: the analog sound wave signal collected by the microphone is first converted into digital signal by FPGA, and then formatted into PCM data. Then, the PCM data is transmitted to the PC via USB or Gigabit network for further processing and analysis.

[0019] 4. In a preferred implementation, feature-level fusion uses the following method: the intensity of the fused image is obtained by weighted summation of the infrared image intensity, the acoustic wave image intensity and the correlation between the two, and the correlation is obtained by calculating the correlation function of the two images.

[0020] The beneficial effect of the above implementation is: through the feature-level fusion technology, the advantages of infrared images and acoustic wave images are combined to improve the comprehensive performance of leakage detection.

[0021] The specific working principle adopted is: feature-level fusion adopts a weighted summation method, in which the intensity of the fused image is obtained by weighted summation of the infrared image intensity, the acoustic image intensity and the correlation between the two. The correlation is obtained by calculating the correlation function of the two images, which comprehensively considers the intensity of the images and the spatial correlation between them.

[0022] 5. In a preferred embodiment, the function for calculating the correlation between the two images is a cosine similarity function, which is used to calculate the similarity between the two images in terms of pixel intensity, and its value is between -1 and 1.

[0023] The beneficial effect of the above implementation is: using the cosine similarity function as the correlation calculation function can quantify the similarity between the two images in terms of pixel intensity, thereby further improving the quality of the fused image.

[0024] The specific working principle adopted is: the cosine similarity function calculates the ratio of the average value of the product of the pixel intensities of the two images to the square root of the sum of the squares of their respective pixel intensities to obtain a similarity value between -1 and 1, thereby quantifying the correlation between the two images.

[0025] 6. In a preferred embodiment, the geometric correction uses a specific transformation matrix, including a scaling factor, a translation vector, and a sine value of a rotation angle, to achieve rotation correction of the image.

[0026] The beneficial effect of the above implementation is: by using a specific geometric transformation matrix to perform geometric correction of the image, accurate alignment of the image before feature-level fusion is ensured.

[0027] The specific working principle adopted is: the transformation matrix used by geometric correction includes the scaling factor, translation vector and sine value of the rotation angle, which is used to adjust the size, position and direction of the image to achieve precise alignment of the image.

[0028] 7. In a preferred implementation, the sound wave data preprocessing includes using adaptive filtering to filter out noise and adaptive threshold denoising.

[0029] The beneficial effects of the above implementation are: through the preprocessing of the acoustic wave data, the quality of the acoustic wave signal is improved, providing a clear data basis for subsequent signal analysis and leakage detection.

[0030] The specific working principle adopted is: the acoustic wave data preprocessing includes noise filtering and adaptive threshold denoising, and a specific adaptive filter formula is used to dynamically adjust the filtering strength according to the local mean and local standard deviation at the frequency.

[0031] 8. In a preferred embodiment, the infrared image preprocessing includes Gaussian filtering and histogram equalization.

[0032] The beneficial effects of the above implementation are: through the preprocessing of infrared images, the image quality is improved, the temperature abnormality area is highlighted, and intuitive image information is provided for leakage detection.

[0033] The specific working principle adopted is: infrared image preprocessing includes Gaussian filtering and histogram equalization. An improved Gaussian filtering formula is used to reduce image noise and smooth the image, and histogram equalization is used to enhance the contrast of the image.

[0034] 9. In a preferred implementation, step three includes the following detailed steps:

[0035] Step 3.1: Preprocess the sound wave signal S(t) obtained using the microphone array acquisition system to filter out noise and enhance signal quality;

[0036] Step 3.2: Perform fast Fourier transform (FFT) on the preprocessed sound wave signal S(t) to obtain its spectrum representation S(f);

[0037] Step 3.3: Determine the leakage characteristic frequency range [f1, f2], which is based on the characteristics of the tanker material, structure and known leakage types;

[0038] Step 3.4: Integrate the spectrum S(f) within the leakage characteristic frequency range [f1, f2] and calculate the leakage detection index E, which is calculated as follows:

[0039]

[0040] Among them, E is used to quantify the energy content of the acoustic signal within the leakage characteristic frequency range; f represents the frequency variable in the frequency domain, in Hertz; df represents the small frequency interval in the integration, which represents the increment of the f variable during the integration process;

[0041] Step 3.5: Compare the calculated leakage detection index E with the preset leakage detection threshold θ. If E>θ, it is considered that there is a leakage risk;

[0042] Step 3.6: Mark the areas in the acoustic wave image where the energy content exceeds the threshold θ. These areas represent potential leakage locations.

[0043] The beneficial effect of the above implementation is that the sensitivity and accuracy of leakage detection are improved through detailed acoustic wave signal processing steps.

[0044] The specific working principle adopted is: including preprocessing of acoustic wave signals, fast Fourier transform, determining the leakage characteristic frequency range, integral calculation of leakage detection indicators, comparison with preset thresholds, and marking potential leakage locations in acoustic wave images.

[0045] 10. In a preferred implementation, step 5 includes the following detailed steps:

[0046] Step 5.1: Receive the infrared image I of the tanker truck obtained by the infrared thermal imager in step 4 th (x, y) and the acoustic wave image I generated in step 3 ac (x,y);

[0047] Step 5.2: Infrared image I th (x,y) is preprocessed, including but not limited to spatial filtering and contrast enhancement, to improve image quality and highlight temperature abnormal areas;

[0048] Step 5.3: Sound wave image I ac (x, y) is preprocessed, including but not limited to frequency domain filtering and amplitude normalization, to clearly display the abnormal area of ​​the sound wave;

[0049] Step 5.4: Perform geometric correction to ensure that the infrared image and the acoustic image are aligned in space for accurate registration;

[0050] Step 5.5: Apply the registration algorithm to adjust the relative position and direction between the infrared image and the acoustic wave image according to the transformation matrix model to achieve accurate registration;

[0051] Step 5.6: Use feature-level fusion technology to combine the registered infrared image and acoustic wave image into a fused image I fu (x, y), the fusion method can be a simple pixel-level weighted superposition or a more complex feature-based fusion algorithm;

[0052] Step 5.7: In the fused image I fu In (x,y), identify and mark the areas of abnormal temperature and abnormal sound waves, which may be the locations where leaks occur;

[0053] Step 5.8: Further analysis is performed on the marked abnormal regions, including but not limited to edge detection, morphological processing, and size measurement of the abnormal regions to assess the severity of the leakage.

[0054] The beneficial effect of the above implementation is that accurate detection and assessment of tanker leakage is achieved through detailed image processing and fusion steps.

[0055] The specific working principle adopted is: including receiving infrared images and acoustic wave images, performing image preprocessing, performing geometric correction, applying registration algorithms, synthesizing fused images using feature-level fusion technology, identifying and marking abnormal areas, and further analyzing abnormal areas.

[0056] Beneficial effects of the present invention:

[0057] The present invention proposes a tanker truck detection method that integrates thermal imaging and acoustic imaging, which has high detection accuracy and is less affected by the environment. The acoustic wave data and visible light data are collected by a microphone array acquisition system, transmitted to a computer, and an acoustic wave image is obtained using an acoustic wave image fusion algorithm. The acoustic wave image is then fused with the thermal imaging image through a feature fusion algorithm to achieve accurate detection of tanker truck defects and leaks. This is mainly reflected in the following aspects:

[0058] 1. Multimodal fusion detection: By combining thermal imaging and acoustic imaging technology, the present invention provides a more comprehensive tank truck leak detection method that can more accurately identify and locate leak points.

[0059] 2. High-precision acquisition system: The use of microphone array acquisition system and 1080P camera ensures high-definition acquisition of acoustic wave signals and visible light images, providing high-quality raw data for subsequent data processing.

[0060] 3. Efficient data processing: Through FPGA processing and PCM format conversion, the data transmission efficiency and processing speed are improved, ensuring the quality and integrity of the data.

[0061] 4. Advanced image fusion technology: Using feature-level fusion technology, combined with the intensity and spatial correlation of images, the accuracy and reliability of leak detection are improved.

[0062] 5. Accurate geometric correction: Use a specific geometric transformation matrix to perform geometric correction on the image, ensuring accurate alignment of the image before feature-level fusion.

[0063] 6. Optimized image preprocessing: Through the preprocessing of acoustic wave data and infrared images, including filtering, denoising, Gaussian smoothing and histogram equalization, the quality of signals and images is improved, providing a clear data basis for leak detection.

[0064] 7. Enhanced signal analysis: Improved leak detection sensitivity through fast Fourier transform and energy content analysis within the leak characteristic frequency range.

[0065] 8. Automated leak identification: Rapid identification of leaks is achieved through comparison with preset thresholds and automatic marking of potential leak locations.

[0066] 9. Comprehensive image analysis: Identify and mark abnormal temperature and acoustic wave areas in the fused image, and conduct further analysis to assess the severity of the leak, providing in-depth information for leak detection.

[0067] 10. Improve detection efficiency and safety: The overall approach reduces interference from human factors and improves detection efficiency through automation and algorithm optimization, while ensuring the safety and effectiveness of tank truck detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Schematic diagram of tank truck leakage detection.

[0069] Figure 2 Schematic diagram of audio data collection.

[0070] Figure 3 This is the working principle diagram of the infrared thermal imager.

[0071] Figure 4 This is a schematic diagram of tank truck detection. DETAILED DESCRIPTION

[0072] In order to enable those skilled in the art to better understand the technical solution, the present invention is described in detail below in conjunction with embodiments. The description in this section is only exemplary and explanatory and should not have any limiting effect on the scope of protection of the present invention.

[0073] The specific technical solution of the present invention is as follows: A tank truck leakage detection method based on the fusion of thermal imaging and acoustic imaging comprises the following steps:

[0074] Step 1: Use a microphone array acquisition system to collect the sound wave signal of the tanker truck;

[0075] Step 2: Collect visible light images of the tanker truck at the same position and angle;

[0076] Step 3: The acoustic wave signal is fused with the visible light image through a data processing algorithm to generate an acoustic wave image, in which the abnormal acoustic wave area is marked;

[0077] Step 4: Use an infrared thermal imager to take an infrared image of the tanker truck at the same position and angle;

[0078] Step 5: Perform geometric correction, registration and feature-level fusion of the infrared image and the acoustic wave image.

[0079] Implementation steps:

[0080] Acquisition of acoustic wave signals: A 64-channel microphone array is used to collect acoustic wave signals from tanker trucks at a sampling frequency of 10kHz.

[0081] Capture visible light images: Use a 1080P camera to synchronously capture images of the tanker truck at a rate of 30 frames per second.

[0082] Fusion to generate acoustic wave images: The acoustic wave signal is converted into a spectrum diagram and fused with the visible light image, and the abnormal acoustic wave area is marked using the threshold method.

[0083] Collect infrared images: Use an infrared thermal imager to take infrared images of the tanker truck at the same synchronous time point.

[0084] Image fusion: The infrared image and the acoustic wave image are geometrically corrected and fused at the feature level, and the fusion weight is determined using the following formula: α = 0.6, β = 0.4.

[0085] As a preferred embodiment of the above embodiment, the microphone array acquisition system includes a 64-channel microphone board, a USB acquisition board and a 1080P camera.

[0086] Implementation steps:

[0087] Arrange the microphone array: Make sure that the spacing between each microphone is 5cm and covers a range of 3m around the tanker.

[0088] Synchronous acquisition: The sound wave signals of 64 microphones are synchronously acquired through the USB acquisition board.

[0089] As a preferred embodiment of the above embodiment, the microphone data is processed by FPGA, converted into PCM format, and transmitted to the PC via USB or Gigabit network.

[0090] Implementation steps:

[0091] FPGA processing: The collected analog sound wave signal is converted into a digital signal through FPGA, with a sampling frequency of 10kHz.

[0092] PCM format conversion: Convert digital signals to PCM format with a bit depth of 16 bits.

[0093] Data transfer: PCM data is transferred to a PC via the USB3.0 interface for further processing.

[0094] As a preferred method of the above embodiment, feature-level fusion uses the following formula:

[0095] Ifu (x,y)=α·I th (x,y)+(1-α)·I ac (x,y)+β·Corr(I th ,I ac )

[0096] Among them, I fu (x,y) is the intensity of the fused image, I th (x, y) is the infrared image intensity, I ac (x, y) is the acoustic image intensity, α is the thermal imaging weight, 0≤α≤1, β is the correlation weight, 0≤β≤1, and Corr is the function for calculating the correlation between two images.

[0097] This formula has the following advantages:

[0098] 1. Multimodal data fusion: This formula fuses two different data sources, thermal imaging and acoustic imaging, providing a new data fusion method that is uncommon in traditional leak detection.

[0099] 2. Weight balance: By adjusting the weights of α and β, the contributions of the two imaging techniques can be balanced according to actual conditions, which increases the flexibility and adaptability of the method.

[0100] 3. Correlation enhancement: By introducing the correlation function Corr, the image fusion effect is enhanced, so that the fused image can better reflect the advantages of the two imaging technologies.

[0101] Implementation steps:

[0102] 1. Calculate image correlation: Use the cosine similarity function to calculate the correlation between the infrared image and the acoustic wave image.

[0103] 2. Apply the fusion formula: Based on the determined weights α and β and the calculated correlation, apply the feature-level fusion formula to synthesize the fused image.

[0104] Assume that the average intensities of the infrared image and the acoustic wave image in a certain region of interest (ROI) are: th = 120 units, Intensity calculation of the unit fusion image (substituting α=0.6, β=0.4):

[0105] I fu =0.6×120+0.4×80+0.4×Corr

[0106] Assuming the correlation Corr is 0.5, then: I f =72+32+0.2=104.2 units

[0107] As a preferred embodiment of the above embodiment, the function Corr for calculating the correlation between two images is a cosine similarity function, which is specifically defined as:

[0108]

[0109] Among them, I th (x,y) and I ac (x, y) represent the intensity values ​​of the infrared image and the acoustic wave image at the pixel position (x, y), respectively. and Represent the average intensity values ​​of the two images respectively.

[0110] The cosine similarity function calculates the cosine similarity of the pixel intensities of the two images, with a value range between -1 and 1, where 1 represents complete similarity, -1 represents complete opposites, and 0 represents no correlation. In this way, the intensity of the fused image takes into account not only the intensity of the individual images, but also the spatial correlation between them, thereby improving the accuracy and reliability of leak detection.

[0111] This formula has the following advantages:

[0112] 1. Quantifying correlation: This formula provides a method to quantify the correlation between pixel intensities of two images, which is an innovative attempt in the field of image fusion.

[0113] 2. Application of cosine similarity: Using cosine similarity to measure the similarity between images is innovative in the field of leak detection.

[0114] Implementation steps:

[0115] 1. Calculate the average intensity value: Calculate the average intensity value for the infrared image and the acoustic wave image respectively and

[0116] 2. Apply the cosine similarity formula: Based on the intensity value of the pixel position (x, y), the cosine similarity formula is applied to calculate the correlation between the two images.

[0117] Assume that within the ROI, the standard deviations of the pixel intensities of the infrared image and the acoustic wave image are:

[0118] σ th =20 units, σ ac =15 The calculation result of unit cosine similarity is 0.5, substituting it into the formula gives:

[0119]

[0120]

[0121] Corr = 0.5 (known value)

[0122] As a preferred embodiment of the above embodiment, the geometric correction uses the following transformation matrix:

[0123]

[0124] Where s is the scaling factor, t x ,t y is the translation vector, and γ is the sine of the rotation angle, which is used for image rotation correction.

[0125] The transformation matrix has the following advantages:

[0126] 1. Image correction: This matrix is ​​used for geometric correction of the image, ensuring accurate alignment before image fusion and improving the quality of the fused image.

[0127] 2. Parametric transformation: by scaling factor s and translation vector t x ,t y and the sine value γ of the rotation angle realize the parameterization of the image transformation and increase the flexibility of the correction process.

[0128] Implementation steps:

[0129] 1. Determine the transformation parameters: select the scaling factor s = 1.0, the translation vector t x =t y =0, the sine value of the rotation angle γ=0.

[0130] 2. Apply the transformation matrix: Use the determined parameters to construct the transformation matrix T and perform geometric correction on the infrared image and the acoustic wave image.

[0131] As a preferred embodiment of the above embodiment, the sound wave data preprocessing includes noise filtering and adaptive threshold denoising.

[0132] The following adaptive filter formula is used:

[0133]

[0134] Among them, I fi (f) is the intensity of the acoustic wave image after filtering, μ(f) is the local mean at frequency f, and σ(f) is the local standard deviation.

[0135] The adaptive filter model has the following advantages:

[0136] 1. Adaptability: The filter adaptively adjusts the filtering strength according to the local mean μ(f) and local standard deviation σ(f) of the signal, thus improving the intelligent level of signal processing.

[0137] 2. Noise suppression: Through adaptive filtering, the noise in the sound wave signal is effectively filtered out, enhancing the signal quality.

[0138] Implementation steps:

[0139] Adaptive filtering: According to the local mean μ(f) and local standard deviation σ(f) of the sound wave signal, the adaptive filter formula is applied for filtering.

[0140] Assume that the local mean of the sound wave signal at a frequency of 500 Hz is:

[0141] μ(500)=0.8Pa

[0142] The local standard deviation is:

[0143] σ(500)=0.2Pa

[0144] Calculation of the intensity of the filtered acoustic wave image:

[0145]

[0146] As a preferred embodiment of the above embodiment, infrared image preprocessing includes Gaussian filtering and histogram equalization.

[0147] Use the following Gaussian filter formula:

[0148]

[0149] Among them, I ga (x, y) is the infrared image intensity after Gaussian filtering, and σ is the standard deviation of the Gaussian kernel.

[0150] The Gaussian filtering formula has the following advantages:

[0151] 1. Image smoothing: Gaussian filtering is a classic image smoothing technique used to reduce image noise and improve image quality.

[0152] 2. Controllable parameters: By adjusting the standard deviation σ of the Gaussian kernel, the intensity and smoothness of the filter can be controlled to meet different image processing requirements.

[0153] Implementation steps:

[0154] 1. Gaussian filtering: Apply Gaussian filtering to the infrared image and choose the standard deviation σ = 1.5.

[0155] 2. Histogram equalization: Perform histogram equalization on the infrared image after Gaussian filtering to enhance the image contrast.

[0156] Assume that within the ROI of the infrared image, the standard deviation of the raw intensity is:

[0157] σor =30 units The standard deviation after Gaussian filtering becomes:

[0158]

[0159] As a preferred embodiment of the above embodiment, step three includes the following detailed steps:

[0160] Step 3.1: Preprocess the sound wave signal S(t) obtained using the microphone array acquisition system to filter out noise and enhance signal quality;

[0161] Step 3.2: Perform fast Fourier transform (FFT) on the preprocessed sound wave signal S(t) to obtain its spectrum representation S(f);

[0162] Step 3.3: Determine the leakage characteristic frequency range [f1, f2], which is based on the characteristics of the tanker material, structure and known leakage types;

[0163] Step 3.4: Integrate the spectrum S(f) within the leakage characteristic frequency range [f1, f2] and calculate the leakage detection index E, which is calculated as follows:

[0164]

[0165] Among them, E is used to quantify the energy content of the acoustic signal within the leakage characteristic frequency range; f represents the frequency variable in the frequency domain, in Hertz; df represents the small frequency interval in the integration, which represents the increment of the f variable during the integration process;

[0166] The above formula applies the energy quantification of acoustic wave signals to leak detection. By selecting a specific frequency range and integrating the energy, it provides a novel, adaptable and sensitive detection method with the following advantages:

[0167] 1) Energy quantification method: This formula provides a direct means of energy quantification by integrating the spectrum of the acoustic signal within a specific frequency range. This method is innovative in the field of leak detection.

[0168] 2) Frequency selectivity: By selecting the leakage characteristic frequency range f1, f2, this method can specifically detect the acoustic wave signals related to the specific leakage and ignore other irrelevant background noise. This is an innovative application in acoustic imaging technology.

[0169] 3) Adaptability: The formula allows the frequency range f1, f2 to be adjusted according to the material, structure and characteristics of known leakage types of different tank trucks to adapt to different detection requirements and environmental conditions, showing the high adaptability and flexibility of the method.

[0170] 4) Enhanced detection sensitivity: By integrating the energy content within a specific frequency range, this method can more sensitively capture the acoustic signal generated by the leak and more accurately identify the leak than traditional single frequency or wide-band analysis.

[0171] 5) Fusion of signal processing technology: The application of this formula reflects the integration of signal processing technology and leakage detection needs, applies frequency domain analysis to actual industrial safety issues, and provides a new technical solution.

[0172] 6) Potential intelligent applications: The calculation result E of this formula can be used as input data for further intelligent analysis, such as training of machine learning models to automatically identify and classify different types of leaks, which makes it possible to intelligently detect leaks.

[0173] 7) Practical application value: Through actual energy quantification, the formula can provide a quantitative decision-making basis for leak detection, help achieve automation and real-time monitoring, and improve the safety and efficiency of industrial production.

[0174] Step 3.5: Compare the calculated leakage detection index E with the preset leakage detection threshold θ. If E>θ, it is considered that there is a leakage risk;

[0175] Step 3.6: Mark the areas in the acoustic wave image where the energy content exceeds the threshold θ. These areas represent potential leakage locations.

[0176] Implementation steps:

[0177] 1. Preprocess the sound wave signal: filter out noise and improve signal quality.

[0178] 2. Fast Fourier Transform: Convert the preprocessed sound wave signal into a spectrum representation.

[0179] 3. Determine the leakage characteristic frequency range: select f1=100Hz, f2=1000Hz.

[0180] 4. Calculate the leakage detection index E: Integrate the spectrum S(f) within a certain frequency range.

[0181] Assume that the energy of the spectrum within the leakage characteristic frequency range is:

[0182]

[0183] The leakage detection threshold is set to: θ = 300 units 2 Since E>θ, it is determined that there is a leakage risk.

[0184] As a preferred embodiment of the above embodiment, step five includes the following detailed steps:

[0185] Step 5.1: Receive the infrared image I of the tanker truck obtained by the infrared thermal imager in step 4 th (x, y) and the acoustic wave image I generated in step 3 ac (x,y);

[0186] Step 5.2: Infrared image I th (x,y) is preprocessed, including but not limited to spatial filtering and contrast enhancement, to improve image quality and highlight temperature abnormal areas;

[0187] Step 5.3: Image the sound wave I ac (x, y) is preprocessed, including but not limited to frequency domain filtering and amplitude normalization, to clearly display the abnormal area of ​​the sound wave;

[0188] Step 5.4: Perform geometric correction to ensure that the infrared image and the acoustic image are aligned in space for accurate registration;

[0189] Step 5.5: Apply the registration algorithm to adjust the relative position and direction between the infrared image and the acoustic wave image according to the transformation matrix model to achieve accurate registration;

[0190] Step 5.6: Use feature-level fusion technology to combine the registered infrared image and acoustic wave image into a fused image I fu (x, y), the fusion method can be a simple pixel-level weighted superposition or a more complex feature-based fusion algorithm;

[0191] Step 5.7: In the fused image I fu In (x,y), identify and mark the areas of abnormal temperature and abnormal sound waves, which may be the locations where leaks occur;

[0192] Step 5.8: Further analysis is performed on the marked abnormal regions, including but not limited to edge detection, morphological processing, and size measurement of the abnormal regions to assess the severity of the leakage.

[0193] Implementation steps:

[0194] 1. Receive infrared images and acoustic wave images: ensure that the images are collected synchronously and the resolution is consistent.

[0195] 2. Image preprocessing: Gaussian filtering and histogram equalization are performed on infrared images, and frequency domain filtering and amplitude normalization are performed on acoustic wave images.

[0196] 3. Geometric correction: Use the transformation matrix T to perform geometric correction on the image.

[0197] 4. Registration algorithm: Apply affine transformation or other registration algorithms to achieve accurate registration between images.

[0198] 5. Feature-level fusion: Use the determined weights and correlation calculation formula to synthesize the fused image.

[0199] 6. Abnormal area analysis: Identify and mark abnormal temperature and acoustic wave areas in the fused image, and perform edge detection, morphological processing and size measurement.

[0200] Assume that the number of pixels in the abnormal area in the fused image is 2000.

[0201] After edge detection, the contour length of the abnormal area is 500 pixels.

[0202] After morphological processing, the area of ​​the abnormal region was reduced to 1500 pixels.

[0203] The resulting data list:

[0204] Fusion image intensity I fu : 104.2 units

[0205] Cosine similarity Corr: 0.5

[0206] The intensity of the acoustic wave image after filtering I fi (500): calculated value (needs actual calculation)

[0207] The standard deviation σ after Gaussian filtering ga : 3.24 units

[0208] Leak detection index E: 500 square units

[0209] Leak detection threshold θ: 300 square units

[0210] Number of pixels in the abnormal area of ​​the fused image: 2000 pixels

[0211] Outline length of abnormal area: 500 pixels

[0212] The area of ​​abnormal region after morphological processing: 1500 pixels

[0213] Through the above calculations and data lists, we have verified the technical effects of the invention, including accurate image fusion, effective acoustic wave signal processing, and precise leak detection capabilities.

[0214] Embodiment 2:

[0215] The present invention comprises the following steps:

[0216] 1. Use a microphone array acquisition system to collect the sound wave signals of the tanker truck. The system includes a 64-channel microphone board, a USB acquisition board and a 1080P camera.

[0217] 2. Collect visible light images of the tanker truck at the same position and angle.

[0218] 3. The acoustic wave signal is fused with the visible light image through a data processing algorithm to generate an acoustic wave image, in which the abnormal acoustic wave area is marked.

[0219] 4. Use an infrared thermal imager to take infrared images of the tanker truck at the same position and angle.

[0220] 5. Perform geometric correction, registration and feature-level fusion of infrared images and acoustic wave images.

[0221] Specific:

[0222] 1. Acoustic wave signal acquisition: Using a microphone array acquisition system, the intensity of the collected acoustic wave signals ranges from 0.5Pa to 1.5Pa.

[0223] 2. Visible light image acquisition: Use a 1080P camera to acquire visible light images synchronized with the acoustic wave signal at a rate of 30 frames per second.

[0224] 3. Fusion of acoustic wave signal and visible light image: The acoustic wave signal is converted into a spectrum through fast Fourier transform (FFT) and fused with the visible light image using the following formula:

[0225] I fu (x,y)=α·I th (x,y)+(1-α)·I ac (x,y)

[0226] Set α=0.6.

[0227] 4. Infrared image acquisition: Use an infrared thermal imager to take infrared images of the tanker truck, with the same resolution as the visible light image.

[0228] 5. Geometric correction: Use the transformation matrix T for geometric correction, where the scaling factor s = 1.0 and the translation vector t x =t y =0, the sine value of the rotation angle γ=0.

[0229] 6. Feature level fusion: Use the following formula for feature level fusion:

[0230] I fu (x,y)=I fu (x,y)+β·Corr(I th ,I ac )

[0231] When β=0.4, the calculation result of the cosine similarity function Corr is 0.5.

[0232] 7. Acoustic wave data preprocessing: The acoustic wave data is preprocessed using an adaptive filter formula, where the local mean μ(f) and the local standard deviation σ(f) are determined based on the actual signal.

[0233] 8. Infrared image preprocessing: Gaussian filtering is performed on the infrared image using the following formula:

[0234]

[0235] The standard deviation of the Gaussian kernel is set to σ = 2.

[0236] 9. Calculation of leakage detection index: Integrate the spectrum S(f) within the leakage characteristic frequency range f1, f2, calculate the leakage detection index E, and set f1=100Hz, f2=1000Hz.

[0237] 10. Image Analysis: In Fusion Image I fu In (x,y), identify and mark areas of temperature anomalies and acoustic anomalies, which may be the locations where leaks occur.

[0238] The resulting data list:

[0239] Sound wave signal intensity range: 0.5Pa to 1.5Pa

[0240] Visible light image frame rate: 30 frames / second

[0241] Fusion image weight α: 0.6

[0242] Relevance weight β: 0.4

[0243] Cosine similarity Corr: 0.5

[0244] Geometric correction parameters: s = 1.0, t x =0,t y =0,γ=0

[0245] Gaussian filter standard deviation σ: 2

[0246] Leakage characteristic frequency range: [100Hz, 1000Hz]

[0247] Leak detection index E: The value calculated according to the formula.

[0248] The technical effects of the invention are demonstrated through the above specific embodiments.

[0249] Embodiment 3:

[0250] 1. System Settings

[0251] We set up a tank truck leakage detection system based on the fusion of thermal imaging and acoustic imaging. The system includes a microphone array acquisition system consisting of a 64-channel microphone board, a USB acquisition board and a 1080P camera to collect acoustic signals and visible light images.

[0252] 2. Data Collection

[0253] In a laboratory environment, we tested a standard-sized tanker truck. We used a microphone array acquisition system arranged around the tanker truck to collect acoustic signals. At the same time, we used a 1080P camera to collect visible light images of the tanker truck at the same position and angle.

[0254] 3. Data processing

[0255] The collected sound wave signal is first processed by FPGA, converted into PCM format, and transmitted to the PC via USB. Then, it is preprocessed using the following adaptive filter formula:

[0256]

[0257] Wherein, μ(f) = 0.5, σ(f) = 0.1. The preprocessed sound wave signal is subjected to a fast Fourier transform (FFT) to obtain a spectrum representation S(f).

[0258] 4. Determination of leakage characteristic frequency range

[0259] According to the material and structure of the tank truck, we determined the leakage characteristic frequency range to be [f1, f2] = [500, 1500] Hz.

[0260] 5. Calculation of leak detection indicators

[0261] Integrate the spectrum S(f) within the leakage characteristic frequency range and calculate the leakage detection index E:

[0262]

[0263] Through actual measurement, we get E=2500.

[0264] 6. Leakage risk assessment

[0265] Set the leakage detection threshold θ = 1000. Since E>θ, we believe there is a leakage risk.

[0266] 7. Fusion Image Generation

[0267] The infrared image of the tanker truck is taken with an infrared thermal imager at the same position and angle, and preprocessed with Gaussian filtering and histogram equalization. Then, the fused image is generated using the following feature-level fusion formula:

[0268] I fu (x, y) = 0.7·I th (x, y)+0.3·I ac (x,y)+0.1·Corr(I th , I ac )

[0269] Among them, the calculation result of the cosine similarity function Corr is 0.85.

[0270] 8. Abnormal area identification

[0271] In the fused image, we identified and marked areas of temperature anomalies and acoustic anomalies. Through edge detection and morphological processing, we determined the exact coordinates and dimensions of the leak location.

[0272] 9. Data List

[0273] The following is a list of some key data obtained through the above steps to prove the technical effect of the invention:

[0274] Leak detection index E Threshold θ Cosine similarity Fusion image weight α Fusion image weight β 2500 1000 0.85 0.7 0.1

[0275] 10. Result Verification

[0276] Through actual testing, we verified that the system can effectively identify the leakage risk of tank trucks. The generation of fused images improves the accuracy of detection, and the identification of abnormal areas provides a basis for further leakage assessment.

[0277] Embodiment 4:

[0278] 1. Experimental Preparation

[0279] A standard-sized tanker truck was selected as the experimental object. In the laboratory environment, the corresponding detection system was configured as follows:

[0280] A tank truck leakage detection method based on the fusion of thermal imaging and acoustic imaging comprises the following steps:

[0281] Step 1: Use a microphone array acquisition system to collect the sound wave signal of the tanker truck;

[0282] Step 2: Collect visible light images of the tanker truck at the same position and angle;

[0283] Step 3: The acoustic wave signal is fused with the visible light image through a data processing algorithm to generate an acoustic wave image, in which the abnormal acoustic wave area is marked;

[0284] Step 4: Use an infrared thermal imager to take an infrared image of the tanker truck at the same position and angle;

[0285] Step 5: Perform geometric correction, registration and feature-level fusion of the infrared image and the acoustic wave image.

[0286] The microphone array acquisition system includes a 64-channel microphone board, a USB acquisition board and a 1080P camera.

[0287] The microphone data is processed by the FPGA, converted into PCM format, and transmitted to the PC via USB or Gigabit Ethernet.

[0288] Feature-level fusion uses the following formula:

[0289] I fu (x,y)=α·I th (x,y)+(1-α)·I ac (x,y)+β·Corr(I th ,I ac )

[0290] Among them, I fu (x,y) is the intensity of the fused image, I th (x, y) is the infrared image intensity, I ac (x, y) is the acoustic image intensity, α is the thermal imaging weight, 0≤α≤1, β is the correlation weight, 0≤β≤1, and Corr is the function for calculating the correlation between two images.

[0291] The function Corr for calculating the correlation between two images is a cosine similarity function, which is specifically defined as:

[0292]

[0293] Among them, I th (x,y) and I ac (x, y) represent the intensity values ​​of the infrared image and the acoustic wave image at the pixel position (x, y), respectively. and Represent the average intensity values ​​of the two images respectively.

[0294] The cosine similarity function calculates the cosine similarity of the pixel intensities of the two images, with a value range between -1 and 1, where 1 represents complete similarity, -1 represents complete opposites, and 0 represents no correlation. In this way, the intensity of the fused image takes into account not only the intensity of the individual images, but also the spatial correlation between them, thereby improving the accuracy and reliability of leak detection.

[0295] The geometric correction uses the following transformation matrix:

[0296]

[0297] Where s is the scaling factor, t x,t y is the translation vector, and γ is the sine of the rotation angle, which is used for image rotation correction.

[0298] Acoustic wave data preprocessing includes noise filtering and adaptive threshold denoising.

[0299] The following adaptive filter formula is used:

[0300]

[0301] Among them, I fi (f) is the intensity of the acoustic wave image after filtering, μ(f) is the local mean at frequency f, and σ(f) is the local standard deviation.

[0302] Infrared image preprocessing includes Gaussian filtering and histogram equalization.

[0303] Use the following Gaussian filter formula:

[0304]

[0305] Among them, I ga (x, y) is the infrared image intensity after Gaussian filtering, and σ is the standard deviation of the Gaussian kernel.

[0306] 2. Acoustic signal acquisition and preprocessing

[0307] The microphone array acquisition system is used to collect the acoustic wave signal S(t) of the tanker truck. The preprocessing step includes filtering out environmental noise and adaptive threshold denoising to enhance the signal quality.

[0308] 3. Fast Fourier Transform

[0309] The preprocessed sound wave signal S(t) is subjected to fast Fourier transform (FFT) to obtain a spectrum representation S(f).

[0310] 4. Determination of leakage characteristic frequency range

[0311] Based on the material properties of the tanker and the known leakage type, the leakage characteristic frequency range is determined to be [f1, f2] = [100, 300] Hz.

[0312] 5. Calculation of leak detection indicators

[0313] Integrate the spectrum S(f) in the frequency range [f1, f2] and calculate the leakage detection index E. Assuming that the spectrum data obtained by FFT is S(f) = [0.1, 0.2, ..., 0.3], and the corresponding frequency point is f = [100, 101, ..., 300] Hz, the calculation results are:

[0314]

[0315] 6. Leakage risk assessment

[0316] The calculated leakage detection index E is compared with the preset leakage detection threshold θ. If E>θ, it is considered that there is a leakage risk;

[0317] Areas with energy content exceeding a threshold value θ are marked in the acoustic wave image, indicating potential leakage locations.

[0318] The leakage detection threshold θ is set to 150. The calculated E=4500. Since E>θ, it is considered that there is a leakage risk. The area where the energy content exceeds the threshold θ is marked in the acoustic wave image.

[0319] 7. Acquisition and preprocessing of infrared images and acoustic wave images

[0320] Step 5.1: Receive the infrared image I of the tanker truck obtained by the infrared thermal imager in step 4 th (x, y) and the acoustic wave image I generated in step 3 ac (x,y);

[0321] Step 5.2: Infrared image I th (x,y) is preprocessed, including but not limited to spatial filtering and contrast enhancement, to improve image quality and highlight temperature abnormal areas;

[0322] Step 5.3: Image the sound wave I ac (x, y) is preprocessed, including but not limited to frequency domain filtering and amplitude normalization, to clearly display the abnormal area of ​​the sound wave;

[0323] Using infrared thermal imager to take infrared image of tanker truck th (x,y), and perform Gaussian filtering and histogram equalization. At the same time, the sound wave image is filtered in the frequency domain and the amplitude is normalized.

[0324] 8. Geometric correction and registration

[0325] Step 5.4: Perform geometric correction to ensure that the infrared image and the acoustic image are aligned in space for accurate registration;

[0326] Step 5.5: Apply the registration algorithm to adjust the relative position and direction between the infrared image and the acoustic wave image according to a predetermined transformation model (such as affine transformation, projection transformation, etc.) to achieve accurate registration;

[0327] To perform geometric correction, use a transformation matrix:

[0328]

[0329] Ensure that the infrared image and the acoustic image are aligned in space. Then apply a registration algorithm to adjust the relative position and orientation between the images.

[0330] 9. Feature-level fusion

[0331] Step 5.6: Use feature-level fusion technology to combine the registered infrared image and acoustic wave image into a fused image I fu (x, y), the fusion method can be a simple pixel-level weighted superposition or a more complex feature-based fusion algorithm;

[0332] Using feature-level fusion technology, the registered infrared image and acoustic wave image are synthesized into a fused image I fu (x,y), using the following formula:

[0333] I fu (x,y)=0.6·I th (x,y)+0.4·I ac (x,y)

[0334] 10. Abnormal area identification and assessment

[0335] Step 5.7: In the fused image I fu In (x,y), identify and mark the areas of abnormal temperature and abnormal sound waves, which may be the locations where leaks occur;

[0336] Step 5.8: Further analysis is performed on the marked abnormal regions, including but not limited to edge detection, morphological processing, and size measurement of the abnormal regions to assess the severity of the leakage.

[0337] Identify and mark the areas of temperature anomaly and acoustic anomaly in the fused image. Through edge detection and morphological processing, determine the exact coordinates and size of the leak location. For example, suppose an abnormal area is identified in the fused image with coordinates (x0, y0) and size d×d.

[0338] 11. Data list and result verification

[0339] The following is a list of some key data obtained through the above steps to prove the technical effect of the invention:

[0340]

[0341] Through actual testing, we verified that the system can effectively identify the leakage risk of tank trucks. The generation of fused images improves the accuracy of detection, and the identification of abnormal areas provides a basis for further leakage assessment.

[0342] The specific working principle of the present invention is:

[0343] Firstly, the microphone array acquisition system is used to collect the acoustic wave signal of the tanker truck;

[0344] Then the visible light image of the tank truck is collected;

[0345] The acoustic wave signal is fused with the visible light image through the data processing algorithm to generate the acoustic wave image. The abnormal acoustic wave area in the visible light image will be marked, getting rid of the cumbersome process of deploying the traditional acoustic wave sensor detection method.

[0346] Afterwards, an infrared image of the tanker truck is captured at the same position and angle using an infrared thermal imager;

[0347] Finally, the infrared image and the processed acoustic wave image are geometrically corrected and registered and finally feature-level fused.

[0348] The present invention adopts the NEMS version of microphone array sound wave acquisition system, which is composed of a 64-channel microphone board, a USB acquisition board and a camera. The sensor used in the microphone array is a silicon microphone, the interface is a PDM interface, and the microphones are arranged according to a multi-arm spiral rule. The array diameter is 150mm, the frequency range that can be collected is 50Hz-80kz, and the recording sound pressure level is 30dBSPL~120dBSPL; the camera adopts a USB interface and supports 1080P. The collected microphone data is processed by FPGA, converted into PCM format, and then transmitted to the PC via USB or Gigabit network. The PC software provides an API library, and the audio data is obtained through the API and saved as a WAV file. The schematic diagram of the audio data collection process is shown as follows Figure 2 shown.

[0349] The present invention uses the MAG70 uncooled infrared thermal imager of Juge Technology, which can detect wavelengths ranging from 7.5 to 14 μm and has a resolution of 640×480. The imaging principle is as follows: Figure 3 shown.

[0350] The schematic diagram of the on-site detection of the present invention is as follows Figure 4 shown.

[0351] The present invention is applicable to all environmental conditions, is not affected by light intensity, and can directly fuse acoustic wave data with infrared images; the interference noise in noisy environments can also be filtered out through the acoustic wave data preprocessing algorithm, thereby eliminating the influence of external noise on the detection effect.

[0352] It should be noted that, in this article, the terms: include, contain and any other variants are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Specific examples are used in this article to illustrate the principle and implementation of the technical solution of the present invention. The above examples are only used to help understand the method of the present invention and its core idea. The above is only a preferred implementation of the present invention. It should be pointed out that due to the limitations of textual expression and the objective existence of infinite specific structures, for ordinary technicians in this technical field, without departing from the principle of the present invention, several improvements, modifications or changes can be made, and the above technical features can also be combined in an appropriate manner; these improvements, modifications, changes or combinations, or the direct application of the concept and technical solution of the present invention to other occasions without improvement, should be regarded as the protection scope of the present invention.

Claims

1. A tank truck leakage detection method based on the fusion of thermal imaging and acoustic imaging, characterized in that: The following steps are involved: Step 1: Use a microphone array acquisition system to collect the sound wave signal of the tanker truck; Step 2: Collect visible light images of the tanker truck at the same position and angle; Step 3: The acoustic wave signal is fused with the visible light image through a data processing algorithm to generate an acoustic wave image, in which the abnormal acoustic wave area is marked; Step 4: Use an infrared thermal imager to take an infrared image of the tanker truck at the same position and angle; Step 5: geometrically correct, register and feature-level fuse the infrared image and the acoustic wave image; Step 3 includes the following detailed steps: Step 3.1: Preprocess the sound wave signal S(t) obtained using the microphone array acquisition system to filter out noise and enhance signal quality; Step 3.2: Perform fast Fourier transform (FFT) on the preprocessed sound wave signal S(t) to obtain its spectrum representation S(f); Step 3.3: Determine the leakage characteristic frequency range [f1, f2], which is based on the characteristics of the tanker material, structure and known leakage types; Step 3.4: Integrate the spectrum S(f) within the leakage characteristic frequency range [f1, f2] and calculate the leakage detection index E, which is calculated as follows: Among them, E is used to quantify the energy content of the acoustic signal within the leakage characteristic frequency range; f represents the frequency variable in the frequency domain, in Hertz; df represents the small frequency interval in the integration, which represents the increment of the f variable during the integration process; Step 3.5: Compare the calculated leakage detection index E with the preset leakage detection threshold θ. If E>θ, it is considered that there is a leakage risk; Step 3.6: Mark the areas in the acoustic wave image where the energy content exceeds the threshold θ, which represent potential leakage locations; Fusion of acoustic wave signal and visible light image: The acoustic wave signal is converted into a spectrum through fast Fourier transform (FFT) and fused with the visible light image using the following formula: I fu (x,y)=α·I th (x,y)+(1-α)·I ac (x,y) Step 5 includes the following detailed steps: Step 5.1: receiving the infrared image of the tanker truck acquired by the infrared thermal imager in step 4 and the acoustic wave image generated in step 3; Step 5.2: Preprocess the infrared image, including but not limited to spatial filtering and contrast enhancement, to improve image quality and highlight temperature abnormality areas; Step 5.3: Preprocessing the acoustic wave image, including but not limited to frequency domain filtering and amplitude normalization, to clearly display the acoustic wave abnormal area; Step 5.4: Perform geometric correction to ensure that the infrared image and the acoustic image are aligned in space for accurate registration; Step 5.5: Apply the registration algorithm to adjust the relative position and direction between the infrared image and the acoustic wave image according to the transformation matrix model to achieve accurate registration; Step 5.6: Using feature-level fusion technology, the registered infrared image and acoustic wave image are combined into a fused image; Step 5.7: In the fused image, identify and mark the areas of temperature anomaly and acoustic anomaly, which may be the location of the leak; Step 5.8: Further analysis is performed on the marked abnormal regions, including but not limited to edge detection, morphological processing, and size measurement of the abnormal regions to assess the severity of the leakage.

2. The tank truck leakage detection method according to claim 1, characterized in that: The microphone array acquisition system includes a 64-channel microphone board, a USB acquisition board and a 1080P camera.

3. The tank truck leakage detection method according to claim 1, characterized in that: The microphone data is processed by the FPGA, converted into PCM format, and transmitted to the PC via USB or Gigabit Ethernet.

4. The tank truck leakage detection method according to claim 1, characterized in that: Feature-level fusion uses the following method: the intensity of the fused image is obtained by weighted summation of the infrared image intensity, the acoustic wave image intensity and the correlation between the two, and the correlation is obtained by calculating the correlation function of the two images.

5. The tank truck leakage detection method according to claim 4, characterized in that: The function for calculating the correlation between two images is a cosine similarity function, which is used to calculate the similarity between two images in terms of pixel intensity, and its value is between -1 and 1.

6. The tank truck leakage detection method according to claim 1, characterized in that: Geometric correction uses a specific transformation matrix, including a scaling factor, a translation vector, and the sine value of the rotation angle, to achieve rotation correction of the image.

7. The tank truck leakage detection method according to claim 1, characterized in that: The acoustic wave data preprocessing includes the use of adaptive filtering to filter out noise and adaptive threshold denoising.

8. The tank truck leakage detection method according to claim 1, characterized in that: Infrared image preprocessing includes Gaussian filtering and histogram equalization.

Images