Millimeter wave radar gas leakage detection method and system based on wavelet scattering characteristics

Abstract

The invention discloses a millimeter wave radar gas leakage detection method and system based on wavelet scattering characteristics. Comprising the following steps: monitoring a target container by using a millimeter-wave radar and filtering an echo signal of a static target to obtain a radar echo of a reserved dynamic target; fourier transform, spectrum peak search and phase unwrapping are carried out on the radar echo signals, and phase information is extracted; inputting the extracted phase information into a wavelet scattering network, and outputting wavelet scattering characteristics of the millimeter wave radar gas leakage signal; the wavelet scattering features are input into a pre-trained support vector machine classifier to recognize the leakage state, and a gas leakage state recognition result is output. The method is simple in model structure, good in real-time performance and high in anti-noise capacity, compared with a conventional time-frequency feature method, the accuracy of gas leakage detection can be remarkably improved, and the method is suitable for popularization and application. And the method is especially suitable for the requirements of high precision and real-time detection on industrial sites.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of millimeter wave radar signal detection, and more particularly relates to a millimeter wave radar gas leakage detection method and system based on wavelet scattering features. BACKGROUND

[0002] Gas leakage detection is a very important task. Once gas leakage occurs, it may cause property loss or even large-scale casualties. Millimeter wave radar is very suitable for gas leakage detection tasks due to its high detection accuracy, long detection distance, and all-weather operation. The micro-Doppler effect of radar can be used to detect the container vibration signal caused by the airflow impact on the container during gas leakage. The vibration signal features are extracted and fed into a classifier to identify the leakage state. The time-frequency characteristics of millimeter wave radar gas leakage signals are very complex, and the commonly used time-frequency features cannot effectively identify the leakage state.

[0003] The prior art patent with publication number CN202411977245.X proposes a high-precision biological perception method and system based on millimeter wave radar. The method decomposes and analyzes the signal intrinsic modal of the millimeter wave detection transmission signal and the millimeter wave detection echo signal to capture the signal component frequency domain difference key semantic aggregation features between the millimeter wave detection transmission and echo signals, and expresses the perception result based on the aggregated representation of the signal component frequency domain difference features. The biological perception recognition result is obtained based on the perception result, thereby improving the accuracy of biological perception using millimeter wave radar detection. However, the signal intrinsic modal decomposition is prone to modal aliasing, i.e., different time scale signal components may be mixed together, resulting in inaccurate decomposition results. In addition, the endpoint effect in the decomposition process may also affect the integrity of the signal. The algorithm process of the method is complex and is easily affected by modal aliasing and endpoint effect, and requires high hardware and data volume. SUMMARY

[0004] To overcome the problems of weak anti-interference ability and high computational complexity of the gas leakage detection method in the prior art, the application provides a millimeter wave radar gas leakage detection method and system based on wavelet scattering features.

[0005] The primary object of the present application is to solve the above technical problems. The technical solution of the present application is as follows: The first aspect of the present application provides a millimeter wave radar gas leakage detection method based on wavelet scattering features, comprising the following steps: The millimeter wave radar is used to monitor the target container and filter out the echo signal of the stationary target to obtain the radar echo of the dynamic target; The radar echo signal is subjected to Fourier transform, spectral peak search, and phase unwrapping to extract the phase information; The extracted phase information is input into a wavelet scattering network, and a wavelet scattering feature of a millimeter wave radar gas leakage signal is output. The wavelet scattering feature is input into a pre-trained support vector machine classifier to identify the leakage state, and a gas leakage state identification result is output.

[0006] Further, the method for monitoring the target container by using the millimeter wave radar and filtering out the echo signal of the stationary target comprises the following steps: The state of the target container is monitored by using the millimeter wave radar to emit a frequency-modulated continuous wave signal, and the echo signal is received to obtain a monitoring signal. The monitoring signal is subjected to static clutter removal to filter out the echo signal of the stationary target and only keep the echo signal of the dynamic target.

[0007] Further, the method for removing the static clutter from the monitoring signal is a dynamic target display algorithm or a phasor mean cancellation algorithm.

[0008] Further, the Fourier transform is a two-dimensional fast Fourier transform.

[0009] Further, the method for extracting phase information from the radar echo signal comprises the following steps: The signal after the static clutter removal is subjected to a two-dimensional fast Fourier transform to obtain a range-Doppler map. The range-Doppler map is subjected to a spectrum peak search to determine the position of the target in the range-Doppler map. The radar signal corresponding to the target position is subjected to phase unwrapping to extract phase information.

[0010] Further, the method for calculating the wavelet scattering feature by using the extracted phase information comprises the following steps: The extracted phase information is input into a wavelet scattering network, and a wavelet scattering feature matrix of a millimeter wave radar gas leakage signal is calculated by using two layers of wavelet scattering transformation, combining a modulo operation and a low-pass filter. The feature importance scores of the elements in the wavelet scattering feature matrix are calculated, and the N most representative wavelet scattering features are selected as the wavelet scattering features of the millimeter wave radar gas leakage signal according to the feature importance scores.

[0011] Further, the method for calculating the feature importance scores of the elements in the wavelet scattering feature matrix is a maximum correlation minimum redundancy algorithm.

[0012] Further, the pre-trained support vector machine is used for gas leakage state identification, the kernel function of the support vector machine is a cubic polynomial kernel function, the high-dimensional space is mapped through the kernel function, and the expression of the kernel function is as follows:

[0013] wherein, is the transpose of the feature is a constant.

[0014] The second aspect of the present application provides a millimeter wave radar gas leakage detection system based on wavelet scattering features, comprising a memory and a processor, the memory comprising a millimeter wave radar gas leakage detection method based on wavelet scattering features program, the millimeter wave radar gas leakage detection method based on wavelet scattering features program is executed by the processor to realize the steps of a millimeter wave radar gas leakage detection method based on wavelet scattering features.

[0015] The third aspect of the present application provides a computer readable storage medium, the computer readable storage medium comprising a millimeter wave radar gas leakage detection method based on wavelet scattering features program, the millimeter wave radar gas leakage detection method based on wavelet scattering features program is executed by the processor to realize the steps of a millimeter wave radar gas leakage detection method based on wavelet scattering features.

[0016] Compared with the prior art, the beneficial effects of the technical scheme of the present application are: The present application makes full use of the characteristics of millimeter wave radar detection container vibration signal, and accurately extracts and identifies the micro-Doppler effect caused by gas leakage state; through wavelet scattering transform, the problems of modal aliasing and end effect in noise strong and multi-component complex signal in intrinsic mode decomposition are effectively overcome, the calculation redundancy and interference are greatly reduced while ensuring the accuracy of key feature expression; the model structure is simple, real-time and has strong anti-noise ability, compared with the conventional time-frequency feature method, the accuracy of gas leakage detection can be significantly improved, especially suitable for the demand of high precision and real-time detection in industrial field. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to make the purpose, technical scheme of the present application more clear, the present application provides the following drawings and makes the following description: Figure 1 The method flowchart provided for the embodiment of the present application; Figure 2 The radar frame data graph before static clutter removal provided for the embodiment of the present application; Figure 3 The radar frame data graph after static clutter removal provided for the embodiment of the present application; Figure 4 The target position graph obtained after the spectrum peak search provided for the embodiment of the present application; Figure 5 The radar signal graph before phase unwrapping provided for the embodiment of the present application; Figure 6 ​A phase-unwrapped radar signal graph provided by the embodiment of the present application; Figure 7 A wavelet scattering feature matrix comparison graph under different leakage conditions provided by the embodiment of the present application; Figure 8 A feature importance score difference graph provided by the embodiment of the present application; Figure 9 An ROC curve graph of different leakage conditions provided by the embodiment of the present application; Figure 10 A confusion matrix graph of the classification model prediction result provided by the embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below in combination with the drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0019] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the protection scope of the present application is not limited by the specific embodiments disclosed below.

[0020] Embodiment 1: The present application provides a millimeter wave radar gas leakage detection method based on wavelet scattering features, as shown in Figure 1 The flow chart of the millimeter wave radar gas leakage detection method based on wavelet scattering features is shown, and the specific steps are as follows: S1: Use the millimeter wave radar to monitor the target container and filter out the echo signal of the stationary target, to obtain the radar echo of the remaining dynamic target.

[0021] The specific process is as follows: Use the millimeter wave radar to emit a frequency-modulated continuous wave (FMCW) signal to monitor the state of the target container, receive the echo signal, and obtain the monitoring signal; Remove the static clutter of the monitoring signal, filter out the echo signal of the stationary target, and only retain the echo signal of the dynamic target.

[0022] More specifically, the method for removing the static clutter of the monitoring signal is a moving target indication (MTI) algorithm or a phasor mean cancellation algorithm.

[0023] Since the radar signal can contain interference from static background, which can mask the signal of dynamic targets (such as micro-vibration caused by gas leakage). Through the moving target indication (MTI) algorithm or the phasor mean cancellation algorithm, the signal of static targets can be filtered out, and only the echo signal of dynamic targets is reserved. For example, the MTI algorithm utilizes the Doppler frequency difference between clutter and moving targets to suppress the phase of static targets through pulse cancellation. In this embodiment, as shown in FIG. 2, it is a radar frame data graph before static clutter removal, as shown in FIG. 3, it is a radar frame data graph after static clutter removal. Figure 2 Figure 3

[0024] S2: Fourier transform, spectrum peak search and phase unwrapping are performed on the radar echo signal to extract phase information.

[0025] More specifically, the Fourier transform is a two-dimensional fast Fourier transform.

[0026] The specific process is as follows: The signal after static clutter removal is subjected to two-dimensional fast Fourier transform to obtain a Range-Doppler graph; The Range-Doppler graph is subjected to spectrum peak search to determine the position of the target in the Range-Doppler graph; The radar signal corresponding to the target position is subjected to phase unwrapping to extract phase information.

[0027] By searching for a spectrum peak in the Range-Doppler graph, the position of the dynamic target can be determined. The spectrum peak search algorithm can identify the significant feature points in the signal, thereby accurately positioning the position of the target in the distance and velocity dimensions. In this embodiment, the target position graph obtained after spectrum peak search is shown in FIG. 4. Figure 4

[0028] After determining the target position, the signal of the target in the Range-Doppler graph is subjected to phase unwrapping processing. Phase unwrapping is to remove phase ambiguity and recover the true phase information of the signal. This process can extract the phase change of the target, which provides key information for subsequent leakage signal analysis. The radar data is a complex signal composed of real and imaginary parts, in which the real and imaginary parts of the complex signal reflect the phase information of the intermediate frequency signal. The phase signal in the distance unit of the target contains the most leakage information, so after determining the distance unit of the target, the complex signal in the distance unit is subjected to arctangent operation processing, and the expression of the phase signal of the target at time t is as follows:

[0029] ​​​​wherein, and respectively represent the imaginary part and the real part of the complex signal at the mth distance unit where the target is located. By applying the inverse tangent function, the range of the phase change is However, during the computer running process, the phenomenon of phase wrapping often occurs. Establishing the coordinate axis with the real axis as the horizontal axis and the imaginary axis as the vertical axis, the range of the angle in the first and second quadrants is and the range of the angle in the third and fourth quadrants is When the angle changes from 0 to , that is, the value from the previous time to the next time crosses the real axis, the phenomenon of phase wrapping occurs, that is, first from and then It can be seen that the angle at jumps, and the amplitude of the jump is

[0030] In order to solve the problem of phase wrapping and obtain the real phase value, we need to phase unwrap the inverse tangent operation. First, determine the difference between the phase value at the next time and the previous time. If it is greater than , the phase value at the next time is subtracted by ; if it is less than , the phase value at the next time is added by ; otherwise, it remains unchanged. The specific phase unwrapping formula is:

[0031] As shown in Figure 5 , the radar signal diagram before phase unwrapping, and as shown in Figure 6 , the radar signal diagram after phase unwrapping. It can be observed that the phase difference of the phase signal before unwrapping changes greatly in many times, lacks periodicity and discontinuity, and cannot extract the leakage vibration signal. After unwrapping, the phase signal, that is, the unwrapping signal, converts the limited phase to a continuous form, thereby improving the accuracy and accuracy of phase measurement. However, the unwrapping signal not only contains the gas leakage signal, but also contains various forms of noise interference.

[0032] S3: input the extracted phase information into the wavelet scattering network, and output the wavelet scattering features of the millimeter wave radar gas leakage signal.

[0033] The specific process is as follows: input the extracted phase information into the wavelet scattering network, and use two layers of wavelet scattering transformation combined with the modulo operation and the low-pass filter to calculate the wavelet scattering feature matrix of the millimeter wave radar gas leakage signal; The feature importance scores of each element in the wavelet scattering feature matrix are calculated, and N wavelet scattering features with the highest representativeness are selected as the wavelet scattering features of the millimeter wave radar gas leakage signal according to the feature importance scores.

[0034] More specifically, the method for calculating the feature importance scores of each element in the wavelet scattering feature matrix is the maximum relevance minimum redundancy (MRMR) algorithm.

[0035] Specifically, first, the radar signal is subjected to low-pass filtering, and the radar gas leakage signal is convolved with the low-pass filter to obtain the zero-order scattering coefficient, which is expressed as follows:

[0036] where x is the original radar gas leakage signal, is a convolution operation, is a low-pass filter, and the extracted is the low-frequency part of the signal, which has translational invariance; The high-frequency part is subjected to nonlinear transformation (modulus), which is expressed as follows:

[0037] where, is a wavelet filter at scale j, which is specifically used to capture high-frequency or local abrupt change components. Since the modulus is nonlinear, it can suppress phase changes and maintain certain deformation stability.

[0038] The nonlinear output is then convolved with the low-pass filter to obtain the first-order scattering coefficient.

[0039] In order to continue to extract the translational invariant component, the first-order scattering coefficient is convolved with , which is expressed as follows:

[0040] to obtain the first-order scattering coefficient. If the same operation (convolution of the high-frequency part with the same or different wavelet and modulus) is performed again, the second-order scattering coefficient can be obtained, and the process is repeated to form a multi-layer wavelet scattering network.

[0041] In this embodiment, the extracted leakage signal is subjected to wavelet scattering change, and a two-layer wavelet scattering network is adopted, with a quality factor of [8, 1]. Different numbers of wavelet filter sets are selected at different scales to perform multi-layer convolution and modulus, and then the wavelet scattering coefficient matrix is generated for subsequent feature selection and classification.

[0042] The feature importance score of each element in the wavelet scattering coefficient matrix is ​​calculated using the maximum effect minimum redundancy (MRMR) algorithm. Based on the feature importance score, feature selection is performed on the wavelet scattering coefficients to remove redundant wavelet scattering coefficients. The N most representative wavelet scattering coefficients are extracted as wavelet scattering features of the millimeter-wave radar gas leak signal. In this embodiment, N is set to 90.

[0043] In the MRMR method, the first step is to maximize the overall correlation between the selected feature set J and the target variable y. This correlation is measured using mutual information. Reflects characteristics With target variable The degree of dependence between them. If and The stronger the linear or nonlinear relationship, the greater the mutual information. The expression for maximum correlation is shown below:

[0044] in, For feature set, The number of features, mutual information The expression is as follows:

[0045] in, for The marginal probability density function, , for , The joint probability density function, the greater the mutual information, the better. and The closer the relationship.

[0046] Given the same set of features The minimum redundancy R(J) is calculated as follows:

[0047] in, , Belonging to the same feature set Any two features, The smaller the value, the less overlapping information there is within the feature set.

[0048] The comprehensive index MRMR is defined as follows:

[0049] Maximize features by searching or iteratively selecting them. This allows us to obtain the optimal feature subset. Finally, according to... Select the top N features from largest to smallest values.

[0050] S4: Input the wavelet scattering features into a pre-trained support vector machine classifier to identify the leak status and output the gas leak status identification result.

[0051] More specifically, a pre-trained Support Vector Machine (SVM) is used for gas leak status identification. The kernel function of the SVM is a cubic polynomial kernel function, which maps high-dimensional spaces. The expression is as follows:

[0052] in, Features transpose, It is a constant.

[0053] This means that SVM seeks the optimal hyperplane in a high-dimensional mapping space of cubic polynomials. A kernel function is a mapping technique used to map samples originally in a low-dimensional feature space to a higher-dimensional or infinite-dimensional feature space, in order to find a hyperplane in this high-dimensional space that more easily separates different categories. If we directly search for a linear separating surface in the low-dimensional space, we may not be able to separate the data well; however, by using a kernel function, we do not need to explicitly write the function form mapping to the high-dimensional space, yet we can still achieve the effect of classification in the high-dimensional space.

[0054] To find the separating hyperplane that maximizes the inter-class margin, the final decision function of SVM is formulated as follows:

[0055] in, The support vectors for the extracted wavelet scattering features, 'as a feature transpose, Let be the label of the i-th support vector. It is a Lagrange operator. is the kernel function, n is the number of support vectors, and b is the bias term.

[0056] To verify the effectiveness of this invention, millimeter-wave radar was used to collect data on four types of gas leaks. Static target clutter removal, spectral peak search, and phase information extraction were performed to obtain the gas leak signal. Wavelet scattering coefficients were obtained using wavelet scattering transformation, and feature selection was performed using MRMR to obtain the final wavelet scattering features. Leakage status was identified using triple SVM. Figure 7The wavelet scattering feature matrix contrast chart of the millimeter wave radar gas leakage signal under different leakage conditions is shown. As can be seen from the figure, there is a clear difference in the wavelet scattering coefficient chart, wherein, Figure 7 (a) is no leakage, Figure 7 (b) is leakage condition 1, Figure 7 (c) is leakage condition 2, Figure 8 (d) is leakage condition 3. As Figure 9 The feature importance score difference chart is shown. The importance score gradually decreases from top to bottom, showing the contribution of key features, indicating that the features above contribute more to the prediction ability of the model. The importance scores of the first few features are significantly higher than those of other features, indicating that these features play a key role in the model. The range of feature importance scores is approximately 0 to 0.03, and the importance scores of the first few features are close to 0.03. As Figure 10 The ROC curve chart of different leakage conditions has four curves, corresponding to four different leakage conditions (0, 1, 2, 3), and each curve has a corresponding AUC (area under the curve) value. The AUC value of leakage condition 0 is the highest, which is 0.9993, indicating that its classification performance is the best. Other AUC values are also high, which are 0.9977 (condition 1), 0.9982 (condition 2) and 0.9991 (condition 3), indicating that the performance of leakage classification recognition is very good. The ROC curves of all models are close to the upper left corner, indicating that these models perform well in distinguishing positive and negative samples. As ​ As shown, the confusion matrix is used to evaluate the accuracy of the classification model, and the classification accuracy of the model in each class is detailed. The horizontal axis represents the predicted class, and the vertical axis represents the true class. The values on the diagonal line represent the number of correctly classified samples, and the values off the diagonal line represent the number of incorrectly classified samples. Overall, the classification accuracy in each class is high, but the confusion between class 1 and class 2 is more obvious. The method performs well in feature selection, classification performance and confusion matrix.

[0057] The present application makes full use of the characteristics of millimeter wave radar detection of container vibration signal, and accurately extracts and identifies the micro-Doppler effect caused by gas leakage state; through wavelet scattering transformation, the problems of modal aliasing and end effect in intrinsic mode decomposition in strong noise and multi-component complex signal are effectively overcome, the calculation redundancy and interference are greatly reduced while ensuring the accuracy of key feature expression; the model structure is simple, real-time and has strong anti-noise ability, compared with conventional time-frequency feature method, the accuracy of gas leakage detection can be significantly improved, especially suitable for the demand of high precision and real-time detection in industrial field.

[0058] Example 2: The embodiment provides a millimeter wave radar gas leakage detection system based on wavelet scattering features, including a memory and a processor, the memory includes a millimeter wave radar gas leakage detection method program based on wavelet scattering features, and the millimeter wave radar gas leakage detection method program based on wavelet scattering features is used to realize the steps of the millimeter wave radar gas leakage detection method based on wavelet scattering features as described in embodiment 1 when the processor executes the millimeter wave radar gas leakage detection method program based on wavelet scattering features.

[0059] Embodiment 3: The embodiment provides a computer readable storage medium, the computer readable storage medium includes a millimeter wave radar gas leakage detection method program based on wavelet scattering features, and the millimeter wave radar gas leakage detection method program based on wavelet scattering features is used to realize the steps of the millimeter wave radar gas leakage detection method based on wavelet scattering features as described in embodiment 1 when the processor executes the millimeter wave radar gas leakage detection method program based on wavelet scattering features.

[0060] Obviously, the above embodiment of the present application is only an example for clearly illustrating the present application, and is not a limitation on the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments cannot be exhausted. Any modification, equivalent replacement and improvement within the spirit and principle of the present application should be included in the protection scope of the claims of the present application.

Claims

1. A millimeter-wave radar gas leak detection method based on wavelet scattering characteristics, characterized in that, Includes the following steps: Millimeter-wave radar is used to monitor target containers and filter out the echo signals of stationary targets, thus obtaining radar echoes that retain dynamic targets. The radar echo signal is subjected to Fourier transform, spectral peak search, and phase unwrapping to extract phase information; The extracted phase information is input into a wavelet scattering network, and the wavelet scattering features of the millimeter-wave radar gas leak signal are output. The wavelet scattering features are input into a pre-trained support vector machine classifier to identify the leak status and output the gas leak status identification result.

2. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 1, characterized in that, A method for monitoring target containers and filtering out echo signals from stationary targets using millimeter-wave radar includes the following steps: The monitoring signal is obtained by transmitting frequency-modulated continuous wave signals using millimeter-wave radar and receiving the echo signals. Static clutter removal is performed on the monitoring signal to filter out the echo signals of stationary targets and retain only the echo signals of moving targets.

3. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 2, characterized in that, The methods for static clutter removal of monitoring signals are moving target display algorithm or phasor mean cancellation algorithm.

4. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 1, characterized in that, The Fourier transform is a two-dimensional fast Fourier transform.

5. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 4, characterized in that, The method for extracting phase information using radar echo signals includes the following steps: The signal after removing static clutter is subjected to a two-dimensional fast Fourier transform to obtain the range-Doppler map; Perform a spectral peak search on the range-Doppler plot to determine the target's position within the range-Doppler plot; Phase unwrapping is performed on the radar signal corresponding to the target location to extract phase information.

6. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 1, characterized in that, The method for calculating wavelet scattering features using extracted phase information includes the following steps: The extracted phase information is input into a wavelet scattering network. By combining two layers of wavelet scattering transform with modulus taking and a low-pass filter, the wavelet scattering feature matrix of the millimeter-wave radar gas leak signal is calculated. Calculate the feature importance score of each element in the wavelet scattering feature matrix, and select the N most representative wavelet scattering features as the wavelet scattering features of the millimeter-wave radar gas leak signal based on the feature importance scores.

7. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 6, characterized in that, The method for calculating the feature importance score of each element in the wavelet scattering feature matrix is ​​the maximum correlation minimum redundancy algorithm.

8. The millimeter-wave radar gas leak detection method based on wavelet scattering characteristics according to claim 1, characterized in that, Gas leak status identification is performed using a pre-trained support vector machine. The kernel function of the support vector machine is a cubic polynomial kernel function, which maps a high-dimensional space. The expression of the kernel function is as follows: in, Features transpose, It is a constant.

9. A millimeter-wave radar gas leak detection system based on wavelet scattering characteristics, characterized in that, The system includes a memory and a processor. The memory includes a program for a millimeter-wave radar gas leak detection method based on wavelet scattering characteristics. When the processor executes the program for the millimeter-wave radar gas leak detection method based on wavelet scattering characteristics, it implements the steps of a millimeter-wave radar gas leak detection method based on wavelet scattering characteristics as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a program for a millimeter-wave radar gas leak detection method based on wavelet scattering features. When the program is executed by a processor, it implements the steps of a millimeter-wave radar gas leak detection method based on wavelet scattering features as described in any one of claims 1 to 8.

Images