A nitrogen-hydrogen leak detection leak discrimination algorithm based on time-frequency scale and image recognition

Abstract

The application discloses a nitrogen-hydrogen leak detection leakage discrimination algorithm based on time-frequency scale and image recognition, and comprises the following steps: S1, performing zero-phase low-pass filtering on a signal; S2, performing wavelet transform on the signal; S3, drawing a time-frequency scale coefficient graph; S4, using a peak feature recognition algorithm to determine whether there is leakage; S5, obtaining a leakage feature peak time domain coordinate to obtain a leak point position; and S6, obtaining an integral of a time-frequency scale coefficient in an interval above a certain frequency near the leak point to evaluate a leakage amount. In the application, continuous information in two spaces of a time domain and a frequency domain is used for leak point recognition and leakage determination, zero-phase low-pass filtering can be used to avoid position deviation caused by filtering phase delay, a peak feature recognition algorithm can be used to eliminate position deviation related to a leak rate, and leak point positioning precision is improved and a misjudgment rate is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sealing detection, in particular to a nitrogen-hydrogen leak detection leakage discrimination algorithm based on time-frequency scale and image recognition. BACKGROUND

[0002] Due to the advantages of economy and high precision, the leak detection method using nitrogen-hydrogen gas as a leak detection gas has been applied. At present, in the dynamic leak detection application, the positioning of the leak in the high-speed scanning of the suspected leak area such as the weld is basically carried out by the following method: the leak signal is collected during the scanning process (in some cases, some advanced filtering is carried out), and in the case that the amplitude of the leak signal or its derived signal is higher than a certain threshold, it is determined that there is a leak point at this position.

[0003] The existing leak point positioning algorithm has some of the following shortcomings: 1. The positioning error is affected by the scanning direction. Due to the response characteristics of the sensor, the same leak point will have different direction delays in forward and reverse scanning, thereby causing positioning error; 2. The positioning error is affected by the leakage level. Due to the response characteristics of the sensor, the leak holes of different levels at the same position will have deviation in the positioning position by using the traditional discrimination method; 3. Due to the influence of environmental background noise, the leak misjudgment rate is high.

[0004] Therefore, it is necessary to seek a method for improving the leak point positioning accuracy and reducing the misjudgment rate under high beat conditions. SUMMARY

[0005] The present application relates to the technical field of sealing detection, in particular to a nitrogen-hydrogen leak detection leakage discrimination algorithm based on time-frequency scale and image recognition.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme:

[0007] A nitrogen-hydrogen leak detection leakage discrimination algorithm based on time-frequency scale and image recognition, comprising the following steps:

[0008] S1. Zero-phase low-pass filtering the signal;

[0009] S2. Wavelet transform of the signal;

[0010] S3. Drawing a wavelet time-frequency scale diagram;

[0011] S4. Based on the wavelet time-frequency scale diagram, using image recognition algorithm, using spike feature recognition to judge whether there is a leak or not;

[0012] S5. Obtaining the leak feature spike time domain coordinate to obtain the leak point position;

[0013] S6. Obtaining the integral of the scale coefficient in the time-frequency domain above a certain frequency near the leak point to evaluate the size of the leak.

[0014] Preferably, the method for determining whether there is a leakage in step S4 specifically comprises the following steps:

[0015] S41. Obtain the image data of the nitrogen-hydrogen pipeline;

[0016] S42. Binaryzation preprocessing of the image;

[0017] S43. LBP feature vector extraction of the preprocessed image;

[0018] S44. Finally, use SVM for classification to determine whether there is a leakage point.

[0019] Preferably, the LBP feature vector extraction method in step S43 specifically comprises the following steps:

[0020] S431. First, divide the detection picture into m*n small areas;

[0021] S432. For a pixel in each small area, compare the gray values of the adjacent 8 pixels with it. If the surrounding pixel values are greater than the center pixel value, the position of the pixel point is marked as 1, otherwise as 0. In this way, 8 bits of binary numbers can be generated by comparing the 8 points in the 3*3 neighborhood, that is, the LBP value of the center pixel point of the small area is obtained;

[0022] S433. Then calculate the histogram of each small area, that is, the frequency of each number appearing; then normalize the histogram;

[0023] S434. Finally, connect the statistical histogram of each small area obtained to become a feature vector, that is, the LBP texture feature vector of the whole image.

[0024] Preferably, the specific method for obtaining the leakage feature peak time domain coordinate to obtain the position of the leakage point in step S5 is: using the peak feature to obtain the displacement s1 of the position corresponding to the peak value, obtaining the corresponding displacement Ds through the time domain deviation Dt, and obtaining the position s0 of the leakage point through compensation s1-Ds.

[0025] Preferably, the specific method for obtaining the time-frequency domain scale coefficient integral of a certain frequency interval near the leakage point to evaluate the size of the leakage in step S6 is: selecting a frequency ws, which is the frequency of the leakage feature, and which is higher than the background noise frequency; selecting the lower limit of the frequency ws, where ws should be able to distinguish the background noise from the leakage signal; performing time-frequency scale coefficient integration on the region below the peak feature curve and above ws formed by the intersection of the peak feature curve and the ws straight line, and evaluating the size of the leakage through calibration.

[0026] In conclusion, by using the above technical solutions, the application has the following advantages:

[0027] In the application, the continuous information in the time domain and the frequency domain is used for leak point identification and leakage determination, zero-phase low-pass filtering can be used to avoid position deviation caused by filtering phase delay, and a peak feature identification algorithm can be used to eliminate position deviation related to the size of the leakage rate, thereby improving the leak point positioning accuracy and reducing the misjudgment rate. BRIEF DESCRIPTION OF DRAWINGS

[0028] Fig. 1 A nitrogen and hydrogen leak detection leakage determination algorithm based on time-frequency scale and image recognition is shown.

[0029] The upper left graph is a time domain leakage signal waveform, the lower left graph is a time-frequency scale coefficient intensity graph, and the right graph is a time domain displacement curve.

[0030] Fig. 2 A leak point identification diagram of a nitrogen and hydrogen leak detection leakage determination algorithm based on time-frequency scale and image recognition is shown. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the application will be described in detail below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0032] Please refer to Figs. 1-2 The application provides a technical solution:

[0033] A nitrogen and hydrogen leak detection leakage determination algorithm based on time-frequency scale and image recognition includes the following steps:

[0034] S1. Zero-phase low-pass filtering of the signal;

[0035] S2. Wavelet transform of the signal;

[0036] S3. Drawing a wavelet time-frequency scale graph;

[0037] S4. Based on the wavelet time-frequency scale graph, using an image recognition algorithm, using a peak feature identification algorithm to determine whether there is a leakage;

[0038] S5. Obtaining the leakage feature peak time domain coordinates to obtain the leak point position;

[0039] S6. Obtaining the integral of the time-frequency domain scale coefficient in the vicinity of the leak point above a certain frequency to evaluate the size of the leakage.

[0040] Specifically, the method for determining whether there is a leakage in step S4 specifically includes the following steps:

[0041] S41. Obtain the image data of the nitrogen-hydrogen pipeline;

[0042] S42. Binaryzation preprocessing of the image;

[0043] S43. LBP feature vector extraction of the preprocessed image;

[0044] S44. Finally, use SVM for classification to determine whether there is a leakage point.

[0045] Specifically, the LBP feature vector extraction method in step S43 specifically includes the following steps:

[0046] S431. First, divide the detection picture into m*n small areas;

[0047] S432. For a pixel in each small area, compare the gray values of the adjacent 8 pixels with it. If the surrounding pixel value is greater than the center pixel value, the position of the pixel point is marked as 1, otherwise as 0. In this way, 8 bits of binary numbers can be generated by comparing the 8 points in the 3*3 neighborhood, that is, the LBP value of the center pixel point of the small area is obtained;

[0048] S433. Then calculate the histogram of each small area, that is, the frequency of each number. Then normalize the histogram;

[0049] S434. Finally, connect the statistical histogram of each small area to obtain a feature vector, which is the LBP texture feature vector of the whole image.

[0050] Specifically, as shown in Fig. 2 the specific method for obtaining the leakage feature peak time domain coordinate to obtain the position of the leakage point in step S5 is: using the peak feature for feature matching to obtain the displacement s1 of the position corresponding to the peak value, obtaining the corresponding displacement Ds through the time domain deviation Dt, and obtaining the position s0 of the leakage point through compensation s1-Ds.

[0051] Specifically, the specific method for obtaining the frequency domain scale coefficient integral of a certain frequency interval near the leakage point to evaluate the size of the leakage in step S6 is: selecting a frequency ws, which is the frequency of the leakage feature and is higher than the background noise frequency, as the frequency domain limit;

[0052] As shown in Fig. 2, select the lower limit of frequency ws=0.4Hz, wherein 4Hz can obviously distinguish background noise and leakage signal, the peak characteristic curve and the straight line formed by ω=0.4Hz cross, the area below the peak characteristic curve and above ws, carry out time-frequency scale coefficient integration in time-frequency domain, through calibration, evaluate the size of leakage.

[0053] The nitrogen and hydrogen leak detection leakage discrimination algorithm based on time-frequency scale and image recognition provided by the embodiment utilizes the continuous information of two spaces of time domain and frequency domain to carry out leak point identification and leakage discrimination, can adopt zero-phase low-pass filtering to avoid the position deviation caused by filtering phase delay, can adopt peak characteristic identification algorithm to eliminate the position deviation related to the size of leak rate, improve the leak point positioning precision and reduce the misjudgment rate.

[0054] The above description of the embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A nitrogen-hydrogen leak discrimination algorithm based on time-frequency scale and image recognition, characterized in that, It comprises the following steps: S1. Zero-phase low-pass filtering the signal; S2. Wavelet transforming the signal; S3. Drawing a wavelet time-frequency scale map; S4. Based on the wavelet time-frequency scale map, using an image recognition algorithm, using a peak feature recognition judgment to determine whether there is a leakage; S5. Obtaining the leakage feature peak time domain coordinate to obtain the leak point position; S6. Obtaining the scale coefficient integral of the interval above a certain frequency near the leak point to evaluate the size of the leakage.

2. The nitrogen-hydrogen leak detection leak discrimination algorithm based on time-frequency scale and image recognition according to claim 1, characterized in that, The method for judging whether there is leakage in step S4 specifically comprises the following steps: S41. Obtain the image data of the nitrogen-hydrogen pipeline; S42. Binaryzation preprocessing of the image; S43. LBP feature vector extraction on the preprocessed image; S44. Finally, use SVM for classification to determine whether there is a leak point.

3. The nitrogen-hydrogen leak detection leak discrimination algorithm based on time-frequency scale and image recognition of claim 2, wherein, The LBP feature vector extraction method in step S43 specifically comprises the following steps: S431. First, divide the detection picture into m*n small areas; S432. For a pixel in each small area, compare the gray values of the adjacent 8 pixels with it. If the surrounding pixel value is greater than the center pixel value, the position of the pixel point is marked as 1, otherwise as 0. In this way, 8 bits of binary number can be generated by comparing the 8 points in the 3*3 neighborhood, that is, the LBP value of the center pixel point of the small area is obtained; S433. Then calculate the histogram of each small area, and then normalize the histogram; S434. Finally, the statistical histogram of each small area is connected to form a feature vector, that is, the LBP texture feature vector of the whole image.

4. The nitrogen-hydrogen leak detection leak discrimination algorithm based on time-frequency scale and image recognition of claim 1, wherein, The specific method for obtaining the leakage feature peak time domain coordinate to obtain the leak point position in step S5 is to use the peak feature for feature matching to obtain the displacement s1 of the peak value corresponding position, obtain the corresponding displacement Ds through the time domain deviation Dt, and obtain the leak point position s0 through compensation s1-Ds.

5. The nitrogen-hydrogen leak detection leak discrimination algorithm based on time-frequency scale and image recognition of claim 1, wherein, The specific method for obtaining the scale coefficient integral of the interval above a certain frequency near the leak point to evaluate the size of the leakage in step S6 is to select a frequency ws, which is the frequency of the leakage feature, and the frequency is higher than the background noise frequency, and ws is used as the frequency domain limit.

Images