Gas type identification method

Abstract

The invention discloses a gas type identification method. The method comprises the following steps: S10, carrying out denoising and image enhancement on a collected infrared image; s20, carrying out a gas absorption spectrum experiment; s30, constructing an absorption peak value database; s40, carrying out band selection and coding on the optical filter; and S50, performing gas type identification through nearest neighbor method matching. Aiming at the difficulty in spectrum peak value matching in gas type identification, a gas absorption spectrum experiment is designed, a special absorption peak value database is constructed, and a nearest neighbor method is introduced for automatic matching and identification of gas types.

Description

Technical Field

[0001] The invention belongs to the technical field of gas detection and relates to a gas type identification method. Background Art

[0002] With the development of artificial intelligence technology and the increasing abundance of on-site gas data, image recognition methods based on deep learning have been widely used in the field of infrared gas identification.

[0003] In a single video frame, it's often difficult to directly identify a gas due to its irregular outlines, blurred edges, and poor contrast with the background. However, the motion relationship between previous and subsequent frames can be used to determine its location. Therefore, incorporating temporal information becomes a key approach to achieving accurate identification.

[0004] Existing technology includes a video classification network model called VideoGasNet, which employs a "foreground-background separation + suspected target identification" approach. This model has been studied on the self-built GasVid dataset and achieved promising results. Alternatively, a two-stage network structure sequentially extracts temporal and spatial characteristics of gas, using a cascade approach to identify leaks. However, this cascade approach can lead to cumulative errors. To address this issue, an asymmetric 3D convolutional neural network (A3DNet) exists. The core component of the network is the asymmetric 3D convolutional A3D-CNN mechanism, which enables the network to focus on key frames containing motion information, reducing the number of parameters while mitigating the cumulative errors associated with the two-stage approach.

[0005] However, there are problems such as detecting a single gas or several types of gases within a narrow frequency band, and imaging interference caused by complex dynamic backgrounds. Summary of the Invention

[0006] To solve the above problems, the technical solution of the present invention is: a gas type identification method, comprising the following steps: S10, performing denoising and image enhancement on the collected infrared image; S20, gas absorption spectrum experiment; S30, building an absorption peak database; S40, performing wavelength selection and encoding on the filter; S50, identifying the gas type by nearest neighbor matching.

[0007] Preferably, the step S30 includes the following steps: S31, using a Fourier transform infrared spectrometer to obtain the infrared absorption peak of the target gas; S32, extract the wavelength band where the strongest absorption peak is located and construct a spectral database named after the peak wave number.

[0008] Preferably, the S40 includes: Filter setting and image acquisition; Image recognition and code generation.

[0009] Preferably, the filter setting and image acquisition include: Set 6 narrowband filters with different central bands, numbered a~f; The infrared images under filters a~f are collected separately through a multi-channel optical structure.

[0010] Preferably, the image recognition and code generation includes: Input 6 images into the YOLOv8 improved detection network respectively; If a gas leak area is detected in a certain filter image, the bit is coded as 1, otherwise it is 0; Generates a 6-bit binary detection code.

[0011] Preferably, the S50 includes the following steps: Compare the detection code with the code corresponding to each gas in the database; The Hamming distance minimum principle is used for nearest neighbor matching; Outputs the most suitable gas type.

[0012] Compared with existing technologies, the present invention has the following beneficial effects: Due to the infrared absorption characteristics of different gases and the different operating bands of various filters, infrared imaging results also vary. When the infrared absorption peak of a gas happens to fall within the operating band of a particular filter, gas leakage can be observed in the imaging results of that filter; conversely, when the infrared absorption peak of a gas does not fall within the operating band of that filter, gas leakage cannot be observed in the imaging results of that filter. To address the difficulties in spectral peak matching in gas type identification, the present invention designs a gas absorption spectrum experiment, constructs a dedicated absorption peak database, and introduces the nearest neighbor method for automatic matching and identification of gas types. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of a gas type identification method according to a specific embodiment of the present invention; Figure 2 FIG. 4 is a flow chart of gas type identification of a gas type identification method according to a specific embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0015] On the contrary, the present invention covers any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention as defined by the claims. Furthermore, to facilitate a better understanding of the present invention, certain specific details are described in detail below in the detailed description of the present invention. Those skilled in the art will be able to fully understand the present invention without these details.

[0016] See also Figure 1 , a gas type identification method, comprising the following steps: S10, performing denoising and image enhancement on the collected infrared image; S20, gas absorption spectrum experiment; S30, building an absorption peak database; S40, performing wavelength selection and encoding on the filter; S50, identifying the gas type by nearest neighbor matching.

[0017] In a specific embodiment, S20 and S30 gas absorption spectrum experiments and database construction are carried out. Spectral acquisition: A Fourier transform infrared spectrometer will be used for spectrum acquisition. Its core component is a dual-beam interferometer. As its moving mirror moves, the optical path difference between the two coherent beams passing through the interferometer changes, and the light intensity measured by the detector also changes accordingly, resulting in an interference pattern. After Fourier transform calculations, the spectrum of the incident light can be obtained.

[0018] Construction of absorption peak data: The wave value of the strongest absorption peak is bifurcated, and a separate data file is formed and named using the wave value of the strongest peak to construct a gas spectrum code database.

[0019] The filter selection and coding in S40 plans to use 6 filters af, and each filter corresponds to a different working range band. According to the infrared absorption peak database of the gas, if the infrared absorption peak of the gas is within the working range of a certain filter, it is coded as 1, otherwise it is coded as 0. The 6 filters can obtain a 6-bit code containing only 0 and 1. For example, assuming that the infrared absorption peak of sulfur hexafluoride is located in the working range band of filters d, e, and f, and the infrared absorption peak of carbon tetrafluoride is located in the working range band of filters b, c, and d, the sulfur hexafluoride library code is 000111, and the carbon tetrafluoride library code is 011100. According to this coding rule, the library coding data can be obtained by encoding the gas to be detected, which can be used for example Figure 2Gas type identification scheme shown.

[0020] S50 first feeds the imaging results of filter af into the improved YOLOv8 algorithm to detect gas leak areas. It then encodes the detection results for each filter. If the algorithm detects leaking gas in a particular filter's imaging result, the code is 1; otherwise, it is 0. This way, when a gas leak occurs, a 6-bit code is generated based on the detection results of the improved YOLOv8 network. This 6-bit code is then compared with the library code using the nearest neighbor method. If it matches the code corresponding to a gas in the library code, the gas is identified as that gas.

[0021] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying gas types, characterized in that: The following steps are involved: S10, performing denoising and image enhancement on the collected infrared image; S20, gas absorption spectrum experiment; S30, building an absorption peak database; S40, performing wavelength selection and encoding on the filter; S50, identifying the gas type by nearest neighbor matching.

2. The gas type identification method according to claim 1, characterized in that: The S30 includes the following steps: S31, using a Fourier transform infrared spectrometer to obtain the infrared absorption peak of the target gas; S32, extract the wavelength band where the strongest absorption peak is located and construct a spectral database named after the peak wave number.

3. The gas type identification method according to claim 1, characterized in that: The S40 includes: Filter setting and image acquisition; Image recognition and code generation.

4. The gas type identification method according to claim 3, characterized in that: The filter setting and image acquisition include: Set 6 narrowband filters with different central bands, numbered a~f; The infrared images under filters a~f are collected separately through a multi-channel optical structure.

5. The gas type identification method according to claim 3, characterized in that: The image recognition and code generation include: Input 6 images into the YOLOv8 improved detection network respectively; If a gas leak area is detected in a certain filter image, the bit is coded as 1, otherwise it is 0; Generates a 6-bit binary detection code.

6. The gas type identification method according to claim 1, characterized in that: The S50 includes the following steps: Compare the detection code with the code corresponding to each gas in the database; The Hamming distance minimum principle is used for nearest neighbor matching; Outputs the most suitable gas type.

Images