Hydrogen filling identification method and system of hydrogen filling station based on image identification technology

Abstract

The invention relates to the technical field of image recognition, in particular to a hydrogen filling recognition method and system of a hydrogen filling station based on an image recognition technology, and the method comprises the steps: multi-source image synchronous collection: synchronously collecting a visible light image and an infrared thermogram of a hydrogen filling gun connection region; segmentation of a fogging interference area: segmenting the fogging interference area in the visible light image based on the temperature distribution of the infrared thermogram to generate an effective imaging area; hydrogen flow feature extraction and matching region generation: hydrogen flow features are extracted, the hydrogen flow features comprise reflection intensity gradient distribution in the visible light image and a low-temperature diffusion region of the infrared thermogram, and a candidate matching region set is constructed; and spatial coincidence degree judgment and leakage trend identification: monitoring the change of the spatial coincidence degree of the reflection intensity gradient and the low-temperature diffusion region, and generating a hydrogen leakage alarm when the coincidence degree exceeds a preset coincidence degree threshold value and is in a continuous expansion trend. The method introduces a spatial coincidence degree evolution discrimination mechanism, and has the characteristics of high resolution, high sensitivity and quick response.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for identifying hydrogen filling at hydrogen stations based on image recognition technology. Background Technology

[0002] With the continuous development of hydrogen energy infrastructure, the operational safety of hydrogen refueling stations, as a core component of the hydrogen supply system, has received high attention. Hydrogen is colorless, odorless, easily diffused, and flammable and explosive, making it highly susceptible to minute leaks during the refueling process due to poor sealing or abnormal contact. Traditional hydrogen refueling safety detection mainly relies on pressure sensors, mass flow meters, or local hydrogen concentration probes for judgment, which makes it difficult to detect instantaneous leaks in a timely manner, especially in complex environments where misjudgments or missed detections are prone to occur. In addition, conventional video monitoring systems only provide image recording functions and lack the ability to analyze and identify the characteristics of hydrogen.

[0003] Existing image recognition technologies are not well-suited for identifying hydrogen leaks in industrial inspections. Firstly, hydrogen itself is invisible, and its leakage process does not directly form a clear outline in visible light images. The accompanying reflective disturbances and thermal diffusion are easily masked by fog, strong light, or background interference. Secondly, current multi-source image fusion methods often focus on image registration and texture enhancement, lacking dynamic coupling modeling of the physical characteristics of hydrogen leaks, making it difficult to achieve high-precision, multi-modal linkage leak identification. Furthermore, the lack of an effective spatial overlap evolution judgment mechanism also prevents existing systems from providing early warnings of leak trends, resulting in delayed response and monitoring blind spots. Therefore, there is an urgent need to construct a hydrogen leak identification method that integrates visible light and infrared image features and possesses dynamic trend perception capabilities to improve the intelligence and real-time performance of hydrogen filling safety monitoring. Summary of the Invention

[0004] This invention provides a method and system for identifying hydrogen filling at hydrogen stations based on image recognition technology.

[0005] The hydrogen filling identification method for hydrogen filling stations based on image recognition technology includes the following steps: S1, Multi-source image synchronous acquisition: Synchronously acquire visible light images and infrared thermal images of the hydrogen charging gun connection area; S2, Fog Interference Region Segmentation: Based on the temperature distribution of infrared thermal images, segment the fog interference region in visible light images to generate an effective imaging region; S3, Hydrogen flow feature extraction and matching region generation: Extract hydrogen flow features within the effective imaging area, including the reflectance intensity gradient distribution in the visible light image and the low-temperature diffusion region in the infrared thermal image, and construct a candidate matching region set based on the initial spatial relationship between the two. S4, Spatial overlap determination and leakage trend identification: Continuously monitor the spatial overlap changes between the reflective intensity gradient and the low-temperature diffusion area in the candidate matching area. When the overlap exceeds the preset overlap threshold and shows a continuous expansion trend, a hydrogen leakage alarm is generated.

[0006] Optionally, S1 includes: S11, Image sensor calibration and alignment initialization: The visible light camera and infrared thermal imager are calibrated using the Zhang Zhengyou calibration method to obtain their respective intrinsic parameter matrices and distortion coefficients. The two are then calibrated using extrinsic parameters to obtain the relative pose matrix and complete the viewpoint alignment initialization. S12, Time Synchronization and Image Frame Matching: Set the synchronization acquisition period Every Two types of image acquisition modules are triggered every second, and image frames are aligned using a timestamp matching algorithm. If there are frame differences, a linear interpolation alignment strategy is adopted. S13, Multi-source image joint correction output: Aligned image data is fused and displayed through pseudo-color mapping. The infrared image is mapped to a pseudo-color channel according to the temperature distribution and superimposed on the visible light image to output a joint observation frame.

[0007] Optionally, S2 includes: S21, Infrared temperature analysis to extract fogging areas: By analyzing the average temperature around each pixel in the infrared thermal image, areas with significantly lower temperatures are identified as suspected fogging areas. S22, Mapping and masking to generate effective imaging area: Map the fogged mask area onto the visible light image, mask the corresponding position, and retain the rest as the effective imaging area.

[0008] Optionally, S21 includes: S211, Local Temperature Average Calculation: For each pixel in the infrared thermal image, define a local window centered on it, and calculate the average temperature within the window; S212, Low-temperature region extraction and mask generation: Based on the difference between the temperature value of each pixel and its corresponding local average temperature, determine whether the point belongs to a suspected fogging area.

[0009] Optionally, S22 includes: S221, Masking Area Projection and Masking Processing: Utilizing the spatial registration relationship already completed between the infrared thermal image and the visible light image, the fogged masking area is mapped onto the coordinate system of the visible light image; S222, Registration Error Compensation and Buffer Expansion: By setting the buffer radius, each fogged pixel is expanded into a circular region within its neighborhood, thereby constructing a weighted expansion mask.

[0010] Optionally, S3 includes: S31, Identify reflective gradient and low temperature region: Within the effective imaging area, extract the reflective intensity gradient features in the visible light image, analyze the temperature distribution in the infrared thermal image, and identify the features of the low temperature diffusion region. S32, Generate candidate hydrogen flow regions: Spatially compare the reflective intensity gradient features and the low-temperature diffusion region features to select locations that have both strong reflective gradients and are in low-temperature regions, thus forming candidate matching regions.

[0011] Optionally, S31 includes: S311, Visible light image reflectance intensity gradient extraction: Within the effective imaging area, reflectance intensity is enhanced and gradient is calculated for the visible light image, and the Sobel operator is used to extract the pixel gradient magnitude map. S312, Low-temperature diffusion region extraction of infrared image: In the same effective imaging area, the temperature map of the infrared thermal image is analyzed, and the low-temperature region mask is extracted based on the average temperature of the local window and the set low-temperature diffusion threshold.

[0012] Optionally, S32 includes: S321, Feature Coupling Condition Setting: Compare the spatial position of the reflectance intensity gradient map in the visible light image with the low-temperature diffusion region extracted from the infrared image, and set the discrimination threshold for the reflectance intensity gradient. Only pixels that simultaneously meet the conditions of being in the low-temperature diffusion region and having a reflectance intensity exceeding the discrimination threshold are considered to have hydrogen perturbation features. S322, Candidate matching region selection and construction: Based on the set feature coupling conditions, pixels that satisfy dual features are selected to form a candidate matching region set.

[0013] Optionally, S4 includes: S41, Spatial overlap calculation and dynamic monitoring: Evaluate the spatial overlap of candidate matching region sets within consecutive time frames; S42, Determining the Expansion Trend and Triggering the Leakage Alarm: Under the premise that the overlap exceeds the overlap threshold, further analyze its expansion trend over time. If two conditions are met, including that there is a number of consecutive frames greater than or equal to the minimum number of consecutive trigger frames that makes the overlap index greater than the overlap threshold, and that the area of ​​the candidate region shows an upward trend during the time period, then it is considered that a hydrogen leakage trend has occurred, and a leakage alarm signal is generated.

[0014] The hydrogen filling identification system for hydrogen filling stations based on image recognition technology, used to implement the aforementioned hydrogen filling identification method for hydrogen filling stations based on image recognition technology, includes the following modules: Image acquisition module: Simultaneously acquires visible light images and infrared thermal images of the hydrogen charging gun connection area to form a multi-source image data stream; Interference region segmentation module: Based on the temperature distribution in the infrared thermal image, it identifies and blocks fog interference regions in the visible light image to extract the effective imaging region; Feature extraction and matching construction module: Extracts the reflective intensity gradient distribution and low-temperature diffusion region within the effective imaging area, and constructs a candidate matching region set; Leakage detection module: continuously monitors the spatial overlap changes of candidate matching areas, determines whether it exceeds the preset overlap threshold and shows an expanding trend, and then generates a hydrogen leak alarm.

[0015] The beneficial effects of this invention are: This invention provides a hydrogen filling identification method for hydrogen refueling stations based on image recognition technology. It fully integrates the complementary characteristics of visible light images and infrared thermal images, and significantly improves image clarity and observation effectiveness through a fog interference elimination mechanism guided by temperature distribution. Compared with traditional sensor monitoring methods, this invention not only avoids the limitations of point-based detector deployment, but also covers a wider detection field and more complex environmental interference conditions. It proposes a coupled feature extraction method that integrates reflective intensity gradient and low-temperature diffusion region, and constructs candidate matching regions based on the initial spatial correlation between the two. This method can sensitively capture subtle disturbances caused by hydrogen leakage and effectively compensate for the blind spots caused by the invisibility of hydrogen itself.

[0016] This invention introduces a spatial overlap evolution discrimination mechanism. By continuously monitoring the overlap trend of reflective and thermal imaging features in candidate regions, and combining the overlap threshold and extended dynamics, it achieves accurate leakage trend identification and early warning, significantly improving the system's timeliness and robustness. The proposed multi-source image fusion recognition system not only has the characteristics of high resolution, high sensitivity and fast response, but is also easy to integrate into the existing hydrogen refueling station monitoring architecture. It can be widely used in hydrogen transportation, energy storage and refueling fields, and has important engineering promotion value and safety assurance significance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0021] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0022] like Figure 1 As shown, the hydrogen filling identification method for hydrogen filling stations based on image recognition technology includes the following steps: S1, Multi-source image synchronous acquisition: Synchronously acquire visible light images and infrared thermal images of the hydrogen charging gun connection area; S2, Fog Interference Region Segmentation: Based on the temperature distribution of infrared thermal images, segment the fog interference region in visible light images to generate an effective imaging region; S3, Hydrogen flow feature extraction and matching region generation: Extract hydrogen flow features within the effective imaging area, including the reflectance intensity gradient distribution in the visible light image and the low-temperature diffusion region in the infrared thermal image, and construct a candidate matching region set based on the initial spatial relationship between the two. S4, Spatial overlap determination and leakage trend identification: Continuously monitor the spatial overlap changes between the reflective intensity gradient and the low-temperature diffusion area in the candidate matching area. When the overlap exceeds the preset overlap threshold and shows a continuous expansion trend, a hydrogen leakage alarm is generated.

[0023] S1 includes: S11, Image Sensor Calibration and Alignment Initialization: The intrinsic parameters of the visible light camera and the infrared thermal imager are calibrated using Zhang Zhengyou's calibration method to obtain their respective intrinsic parameter matrices. and distortion coefficient The relative pose matrix is ​​obtained by calibrating the extrinsic parameters of the two. Complete viewpoint alignment initialization; in, The intrinsic parameter matrix of the visible light image. , Visible light images Focal length along the axial direction (unit: pixels). Visible light images Focal length along the axial direction (unit: pixels). The principal point of the visible light image is at The coordinates of the direction (the x-coordinate of the image center). The principal point of the visible light image is at The coordinates of the direction (the vertical coordinate of the image center). It is the intrinsic parameter matrix of the infrared image. , Infrared images Focal length along the axial direction (unit: pixels). Infrared images Focal length along the axial direction (unit: pixels). The main point of the infrared image is The coordinates of the direction (the x-coordinate of the image center). The main point of the infrared image is The coordinates of the direction (the vertical coordinate of the image center). It is the extrinsic parameter matrix (rotation matrix) for converting an infrared image to a visible light image. Translation vector The extrinsic parameter matrix is ​​used to describe the relative attitude (rotation and translation) between the two cameras. Its function is to convert spatial points in the infrared camera coordinate system to representations in the visible light camera coordinate system. The complete expression is as follows: ; in, yes The rotation matrix represents the rotation relationship between the infrared camera and the visible light camera. yes The translation vector represents the translation displacement of the infrared camera in the visible light camera coordinate system. It is a transformation relationship. , It is a three-dimensional point (unit: m) in the coordinate system of the infrared camera. It is a three-dimensional point in the visible light camera coordinate system after transformation; S12, Time Synchronization and Image Frame Matching: Set the synchronization acquisition period Every Two types of image acquisition modules are triggered every second, and image frames are aligned using a timestamp matching algorithm. If frame differences exist, a linear interpolation alignment strategy is used, as shown below: ; in, It is a synchronized fused reference image. , It is a moment Infrared images and visible light images, It is the image temporal alignment weight coefficient, and its value range is... The frame time difference is dynamically adjusted. When the infrared image acquisition time is slightly later than the visible light image acquisition time, the frame time is adjusted accordingly. A smaller value (e.g., 0.2–0.4) is used to improve the temporal consistency of visible light images. When the infrared image acquisition time is earlier than the visible light image acquisition time, a smaller value is used. A larger value (e.g., 0.6–0.8) is used to balance the temporal reconstruction of the infrared images, so that when the timestamps of two frames are basically the same, It can be set to 0.5, indicating that the two images are fused with equal weights; S13, Multi-source image joint correction output: Aligned image data is fused and displayed through pseudo-color mapping; infrared images are mapped to pseudo-color channels according to temperature distribution. Superimposed on a visible light image Output joint observation frames , is represented as: ; in, These are image pixel coordinates, representing the two-dimensional position of a pixel in the image, with the horizontal direction being... Vertical direction is , It is the temperature value of the corresponding pixel in the infrared image. It is a pseudo-color mapping function that maps temperature values ​​to corresponding colors in a pseudo-color image. This is an image channel-level overlay operation, which means fusing the pseudo-color channel of the infrared image with the visible light image pixel by pixel. It can be in the form of image blending or layer overlay. It is a joint observation image at the pixel level The fusion result represents the fused image, which is used for subsequent feature recognition and visual analysis, and contains both temperature and texture information.

[0024] S2 includes: S21, Infrared temperature analysis to extract fogging areas: By analyzing the average temperature around each pixel in the infrared thermal image, areas with significantly lower temperatures are identified as suspected fogging areas. S22, Mapping and masking to generate effective imaging area: Map the fogged mask area onto the visible light image, mask the corresponding position, and retain the rest as the effective imaging area.

[0025] S21 includes: S211, Calculation of local average temperature: for infrared thermal images For each pixel, define a local window centered on it, calculate the average temperature within the window, and denot it as . This is used to characterize the local temperature environment of the pixel. in, These are the temperature values ​​of pixels in an infrared thermal image, representing the actual temperature range (typically -20°C). Up to 80 (This is related to the range of the infrared camera) Therefore The average temperature of the local window centered on the pixel depends on the dynamic changes in the values ​​of neighboring pixels. S212, Low-temperature region extraction and mask generation: Based on the temperature value of each pixel and its corresponding local average temperature Based on the differences, determine whether the point belongs to a suspected fogging area. If the following conditions are met, include it in the preliminary fogging masking area. , is represented as: ; in, This is the temperature offset threshold, with a value range of [value range missing]. In infrared thermal images, fogged areas often have significantly lower temperatures than surrounding areas due to evaporation or condensation. Setting a temperature offset threshold aims to identify these anomalous areas from the local temperature distribution. When the temperature difference exceeds [a certain threshold], [the threshold is set]. At that time, the characteristics of fog formation were already somewhat identifiable, and when the temperature difference exceeded [a certain threshold], [further details needed]. By then, most of the foggy areas had become fully visible, therefore, It is considered to be a reasonable trade-off between accuracy and robustness.

[0026] S22 includes: S221, Masking Area Projection and Masking Processing: Utilizing the spatial registration relationship already established between the infrared thermal image and the visible light image, the fogged masking area is... Mapping the coordinates of the visible light image to the visible light image coordinate system, the position corresponding to the mask area in the visible light image is masked, while the rest is preserved, generating the effective imaging area of ​​the image. , is represented as: ; in, Represents the original visible light image. These are low-temperature areas detected from infrared images, used to identify locations where fog may occur. It is the effective observation area after removing fog interference, representing the area in the visible light image that is not affected by fog. It is the key analysis range for subsequent hydrogen feature identification. By first masking and then retaining, the reliability and accuracy of subsequent image processing can be significantly improved. S222, Registration Error Compensation and Buffer Expansion: To enhance the robustness of mask occlusion, in scenarios with pixel-level registration errors, a pixel buffer can be introduced at the mask edge. This buffer radius can be set... Each fogged pixel is expanded into a circular region within its neighborhood, thereby constructing a weighted expanded mask. , is represented as: ; in, This is the buffer expansion radius, ranging from 3 to 5 pixels. During image registration, there is often a sub-pixel error between the infrared image and the visible light image, especially in edge areas or during rapid movement. To avoid incomplete fogging due to slight deviations, the mask edges need to be expanded. A value of 3 to 5 pixels is used to minimize interference with clear areas while ensuring the masking effect, and is suitable for the typical registration error range (approximately 0.5 to 1.5 pixels) at common image resolutions. The weighted expanded fogging mask area, These are the expanded pixel coordinates within the buffer.

[0027] S3 includes: S31, Identify reflective gradient and low temperature region: Within the effective imaging area, extract the reflective intensity gradient features in the visible light image to capture the brightness changes caused by hydrogen disturbance, analyze the temperature distribution in the infrared thermal image, and identify the features of the low temperature diffusion region. S32, Generate candidate hydrogen flow regions: Spatially compare the reflective intensity gradient features and the low-temperature diffusion region features to select locations that have both strong reflective gradients and are in low-temperature regions, thus forming candidate matching regions.

[0028] S31 includes: S311, Visible light image reflectance intensity gradient extraction: in the effective imaging area Inside, for visible light images Reflection intensity enhancement and gradient calculation are performed, and the Sobel operator is used to extract pixel gradient magnitude maps. Used to detect regions of abrupt changes in brightness, denoted as: ; in, This represents the gradient magnitude of the reflected light intensity of the pixel, ranging from 0 to 150. In the early stages of a leak or in areas of gas disturbance, visible light images often show weak but clear light reflection edges. By calculating the brightness gradient, the detectability of this feature can be enhanced. The Sobel operator, as an efficient edge detection operator, is suitable for real-time calculation, and the extracted gradient values ​​typically vary within the range of 0–150. It is the brightness gradient of the image in the horizontal / vertical direction; S312, Low-temperature diffusion region extraction in infrared images: within the same effective imaging area In the middle, the temperature map of the infrared thermogram. Analysis was performed based on the average temperature of the local window. With the set low-temperature diffusion threshold Extracting the mask in the low-temperature region , is represented as: ; in, Infrared images at the pixel level The temperature value at that location, depending on the infrared camera configuration, is typically as follows: to , Therefore The average temperature of the center window. This is the low-temperature diffusion identification threshold; a recommended value range is [value missing]. Hydrogen leakage is often accompanied by rapid gas diffusion, resulting in a small but stable local cooling effect, typically within a certain range. to The difference is most significant between, and below It is difficult to distinguish natural temperature waves, which are higher than This could lead to a misjudgment of a normal heat dissipation area as an abnormal area. Therefore, Settings The range of parameters balances sensitivity and false positive control, making it suitable for real-time on-site detection needs. This indicates the region of low-temperature disturbance that may be caused by hydrogen diffusion.

[0029] S32 includes: S321, Feature Coupling Condition Setting: The reflectance intensity gradient map in the visible light image... Low-temperature diffusion region extracted from infrared image Perform spatial location comparison and set a discrimination threshold for reflectance intensity gradient. It is used to identify locations with significant brightness changes. Only pixels that simultaneously meet the conditions of being in a low-temperature diffusion area and having a reflective intensity exceeding the discrimination threshold are considered to have hydrogen perturbation characteristics. S322, Candidate Matching Region Selection and Construction: Based on the set feature coupling conditions, pixels that satisfy dual features are selected to form a candidate matching region set. , is represented as: ; in, This is the reflectivity intensity gradient threshold, with a recommended value range of 10-30. It's used to filter areas with significant brightness changes and is an important basis for identifying optical reflection edges. Under hydrogen flow disturbance, visible light images often produce weak but continuous brightness gradients. Setting... A value of 10–30 can effectively distinguish between real disturbance reflections and background texture changes. Values ​​below 10 are prone to introducing image noise, while values ​​above 30 may miss weak leaks. This indicates a suspected low-temperature region determined by infrared image temperature analysis. It is a set of candidate matching regions, which is Based on this, the target area selected by further superimposing the reflective features of the visible light image can be constructed by the intersection of two independent feature spaces, which can improve the accuracy and robustness of leakage area identification and prevent misjudgment by a single feature.

[0030] S4 includes: S41, Spatial overlap calculation and dynamic monitoring: Calculation and dynamic monitoring of candidate matching region sets. The degree of spatial overlap within consecutive time frames is evaluated, and the first... Overlap index at frame time , is represented as: ; in, Indicates the number of pixels in the set. For a moment The candidate matching region set satisfies the intersection of low temperature and reflective features. It is a moment The low-temperature diffusion region, For reflective intensity gradient greater than The set of pixels, It is the spatial overlap degree, and its value range is... , This is the spatial overlap threshold. Hydrogen leakage can simultaneously cause visible light reflection and infrared cryogenic features to overlap. To determine whether there is significant overlap, an overlap threshold is set. It can reliably distinguish between normal disturbances and abnormal leakage in most engineering scenarios. If the threshold is too low, it will misjudge the background texture overlap, and if it is too high, it may miss weak leakage. S42, Expansion Trend Determination and Leakage Alarm Triggering: Given that the overlap exceeds the overlap threshold, further analyze its expansion trend over time, defining the observation window length as... A frame is considered valid if it meets two conditions: there exists a number of consecutive frames greater than or equal to the minimum number of consecutive trigger frames. This causes the overlap index to exceed the overlap threshold. And during this period, the area of ​​the candidate region showed an upward trend, that is If this occurs, it is considered that there is a hydrogen leakage trend, and a leakage alarm signal is generated. ; in, The minimum number of consecutive trigger frames, with a value range of [value range missing]. Using multi-frame judgment can avoid false alarms caused by instantaneous noise or short-term disturbances. The recommended value is 3-5 frames, which means that 0.1-0.3 seconds of continuous abnormal behavior is required to trigger an alarm. This is suitable for general industrial field video frequencies (15-30fps). This is an alarm output signal; 1 indicates triggering, and 0 indicates no triggering. This is the length of the time window, used for smooth judgment. The value range is 10-20. The time window is used to evaluate the changing trend of the candidate region and prevent over-response to short-term interference. Setting it to 10-20 frames is equivalent to a dynamic tracking interval of 0.5-1 seconds, which can effectively balance response speed and trend robustness. It is suitable for hydrogen scenarios where the leakage boundary expands slowly.

[0031] like Figure 2 As shown, the hydrogen filling identification system for hydrogen filling stations based on image recognition technology is used to implement the aforementioned hydrogen filling identification method for hydrogen filling stations based on image recognition technology, and includes the following modules: Image acquisition module: Simultaneously acquires visible light images and infrared thermal images of the hydrogen charging gun connection area to form a multi-source image data stream; Interference region segmentation module: Based on the temperature distribution in the infrared thermal image, it identifies and blocks fog interference regions in the visible light image to extract the effective imaging region; Feature extraction and matching module: Extracts the reflectance intensity gradient distribution and low-temperature diffusion region within the effective imaging area, and constructs a candidate matching region set; Leakage detection module: continuously monitors the spatial overlap changes of candidate matching areas, determines whether it exceeds the preset overlap threshold and shows an expanding trend, and then generates a hydrogen leak alarm.

[0032] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0033] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying hydrogen filling at hydrogen refueling stations based on image recognition technology, characterized in that, Includes the following steps: S1, Multi-source image synchronous acquisition: Synchronously acquire visible light images and infrared thermal images of the hydrogen charging gun connection area; S2, Fog Interference Region Segmentation: Based on the temperature distribution of infrared thermal images, segment the fog interference region in visible light images to generate an effective imaging region; S3, Hydrogen flow feature extraction and matching region generation: Extract hydrogen flow features within the effective imaging area, including the reflectance intensity gradient distribution in the visible light image and the low-temperature diffusion region in the infrared thermal image, and construct a candidate matching region set based on the initial spatial relationship between the two. S4, Spatial overlap determination and leakage trend identification: Continuously monitor the spatial overlap changes between the reflective intensity gradient and the low-temperature diffusion area in the candidate matching area. When the overlap exceeds the preset overlap threshold and shows a continuous expansion trend, a hydrogen leakage alarm is generated.

2. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 1, characterized in that, S1 includes: S11, Image sensor calibration and alignment initialization: The visible light camera and infrared thermal imager are calibrated using the Zhang Zhengyou calibration method to obtain their respective intrinsic parameter matrices and distortion coefficients. The two are then calibrated using extrinsic parameters to obtain the relative pose matrix and complete the viewpoint alignment initialization. S12, Time Synchronization and Image Frame Matching: Set the synchronization acquisition period to trigger two types of image acquisition modules, and use the timestamp matching algorithm to align image frames. If there are frame differences, use a linear interpolation alignment strategy. S13, Multi-source image joint correction output: Aligned image data is fused and displayed through pseudo-color mapping. Infrared images are mapped to pseudo-color channels according to temperature distribution and superimposed on visible light images to output joint observation frames.

3. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 2, characterized in that, S2 includes: S21, Infrared temperature analysis to extract fogging areas: By analyzing the average temperature around each pixel in the infrared thermal image, areas with significantly lower temperatures are identified as suspected fogging areas. S22, Mapping and masking to generate effective imaging area: Map the fogged mask area onto the visible light image, mask the corresponding position, and retain the rest as the effective imaging area.

4. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 3, characterized in that, S21 includes: S211, Local Temperature Average Calculation: For each pixel in the infrared thermal image, define a local window centered on it, and calculate the average temperature within the window; S212, Low-temperature region extraction and mask generation: Based on the difference between the temperature value of each pixel and its corresponding local average temperature, determine whether the point belongs to a suspected fogging area.

5. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 3, characterized in that, S22 includes: S221, Masking Area Projection and Masking Processing: Utilizing the spatial registration relationship already completed between the infrared thermal image and the visible light image, the fogged masking area is mapped onto the coordinate system of the visible light image; S222, Registration Error Compensation and Buffer Expansion: By setting the buffer radius, each fogged pixel is expanded into a circular region within its neighborhood, thereby constructing a weighted expansion mask.

6. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 5, characterized in that, S3 includes: S31, Identify reflective gradient and low temperature region: Within the effective imaging area, extract the reflective intensity gradient features in the visible light image, analyze the temperature distribution in the infrared thermal image, and identify the features of the low temperature diffusion region. S32, Generate candidate hydrogen flow regions: Spatially compare the reflective intensity gradient features and the low-temperature diffusion region features to select locations that have both strong reflective gradients and are in low-temperature regions, thus forming candidate matching regions.

7. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 6, characterized in that, S31 includes: S311, Visible light image reflectance intensity gradient extraction: Within the effective imaging area, reflectance intensity is enhanced and gradient is calculated for the visible light image, and the Sobel operator is used to extract the pixel gradient magnitude map. S312, Low-temperature diffusion region extraction of infrared image: In the same effective imaging area, the temperature map of the infrared thermal image is analyzed, and the low-temperature region mask is extracted based on the average temperature of the local window and the set low-temperature diffusion threshold.

8. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 6, characterized in that, S32 includes: S321, Feature coupling condition setting: Compare the spatial position of the reflectance intensity gradient map in the visible light image with the low temperature diffusion region extracted from the infrared image, and set the discrimination threshold of the reflectance intensity gradient. Only pixels that simultaneously meet the conditions of being in the low temperature diffusion region and having a reflectance intensity exceeding the discrimination threshold are considered to have hydrogen perturbation features. S322, Candidate matching region selection and construction: Based on the set feature coupling conditions, pixels that satisfy dual features are selected to form a candidate matching region set.

9. The hydrogen filling identification method for hydrogen filling stations based on image recognition technology according to claim 8, characterized in that, S4 includes: S41, Spatial overlap calculation and dynamic monitoring: Evaluate the spatial overlap of candidate matching region sets within consecutive time frames; S42, Determining the Expansion Trend and Triggering the Leakage Alarm: Under the premise that the overlap exceeds the overlap threshold, further analyze its expansion trend over time. If two conditions are met, including that there is a number of consecutive frames greater than or equal to the minimum number of consecutive trigger frames that makes the overlap index greater than the overlap threshold, and that the area of ​​the candidate region shows an upward trend during the time period, then it is considered that a hydrogen leakage trend has occurred, and a leakage alarm signal is generated.

10. A hydrogen filling identification system for a hydrogen filling station based on image recognition technology, used to implement the hydrogen filling identification method for a hydrogen filling station based on image recognition technology as described in any one of claims 1-9, characterized in that, Includes the following modules: Image acquisition module: Simultaneously acquires visible light images and infrared thermal images of the hydrogen charging gun connection area to form a multi-source image data stream; Interference region segmentation module: Based on the temperature distribution in the infrared thermal image, it identifies and blocks fog interference regions in the visible light image to extract the effective imaging region; Feature extraction and matching construction module: Extracts the reflective intensity gradient distribution and low-temperature diffusion region within the effective imaging area, and constructs a candidate matching region set; Leakage detection module: continuously monitors the spatial overlap changes of candidate matching areas, determines whether it exceeds the preset overlap threshold and shows an expanding trend, and then generates a hydrogen leak alarm.

Images