Gas imaging method with visual sensor data fusion

Abstract

The application belongs to the technical field of visual sensing data fusion, and specifically provides a gas imaging method for visual sensing data fusion. The method mainly comprises the following steps: acquiring original sensor signals, performing multi-source synchronization and feature extraction on the original sensor signals to obtain synchronization feature data; performing anti-disturbance space registration processing based on the synchronization feature data set to generate a space mapping data set; performing multi-source enhanced image fusion processing on the space mapping data set to obtain a fusion image; and outputting the fusion image to generate an output result. The application can effectively improve the multi-source signal synchronization accuracy and feature extraction reliability, enhance the anti-disturbance ability of space registration, optimize the multi-source data fusion quality, make the gas concentration distribution analysis more accurate, and meet the high-precision detection requirements of industrial scenes.

Description

Technical Field

[0001] The present application belongs to the technical field of visual sensor data fusion, and in particular relates to a gas imaging method of visual sensor data fusion. Background Art

[0002] Gas detection is to monitor the gas composition and concentration in the environment in real time to avoid explosions and fires caused by combustible gas leakage, or poisoning of personnel and damage to equipment due to accumulation of toxic gases, so as to ensure industrial production safety, environmental quality and human health.

[0003] Current gas detection usually uses sensors to collect raw signals and then outputs results through signal synchronization, feature extraction, spatial alignment, image fusion and other steps. However, due to differences in response characteristics and environmental interference, sensors are prone to timing deviations in the signal synchronization step, resulting in distortion of characteristic parameters, deviations in spatial coordinate mapping, and insufficient utilization of the complementarity of multi-source data. Ultimately, the accuracy of gas concentration distribution analysis results is insufficient, making it difficult to meet the needs of fine chemical industry, environmental monitoring and other scenarios for accurate assessment of gas status. Summary of the Invention

[0004] This application provides a gas imaging method based on visual sensor data fusion, which effectively solves the problem in the prior art that sensors are prone to insufficient accuracy in gas concentration distribution analysis results due to differences in response characteristics and environmental interference. It can improve the synchronization accuracy of multi-source signals and the reliability of feature extraction, enhance the anti-disturbance capability of spatial alignment, optimize the quality of multi-source data fusion, make gas concentration distribution analysis more accurate, and meet the high-precision detection needs of industrial scenarios.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a gas imaging method based on visual sensor data fusion, comprising:

[0007] The original sensor signals are acquired, multi-source synchronization and feature extraction are performed on the original sensor signals to obtain synchronized feature data.

[0008] Anti-disturbance spatial registration processing is performed based on the synchronized feature dataset to generate a spatial mapping dataset.

[0009] Multi-source enhanced image fusion processing is performed on the spatial mapping dataset to obtain a fused image.

[0010] Output the fused image to generate an output result.

[0011] Furthermore, the original sensor signals include camera video frames, gas sensor array signals, and laser ranging signals.

[0012] Further, multi-source synchronization and feature extraction are performed on the original sensor signals to obtain synchronized feature data, including:

[0013] The camera video frames, gas sensor array signals, and laser ranging signals are time-stamped and aligned to generate synchronized video frames, synchronized concentration signals, and synchronized distance signals.

[0014] The current frame of the synchronized video frames is extracted, and scene key feature points are identified to generate a reference image and scene feature coordinates.

[0015] The physical location parameters of the gas sensor array are received, and the synchronized concentration signals are combined to generate a location-bound concentration dataset.

[0016] Further, anti-disturbance spatial registration processing is performed based on the synchronized feature dataset to generate a spatial mapping dataset, including:

[0017] Based on the scene feature coordinates and sensor array physical location parameters, the relative position relationship between the feature points and the detection units is calculated to generate a dynamic registration matrix.

[0018] The device gyroscope attitude data is obtained, and the dynamic registration matrix and device gyroscope attitude data are used to compensate for the matrix offset caused by vibration to generate an anti-shake registration matrix.

[0019] The anti-shake registration matrix and the reference image are used to map the detection unit positions to the image pixel space to generate an anti-shake mapping region.

[0020] Further, multi-source enhanced image fusion processing is performed on the spatial mapping dataset to obtain a fused image, including:

[0021] The spatial mapping dataset is subjected to multi-physical field concentration calibration to obtain a calibrated concentration dataset.

[0022] Based on the calibrated concentration dataset, a structure-guided field reconstruction is performed to obtain a structure-optimized concentration field.

[0023] The structure-optimized concentration field is subjected to trace-enhanced image fusion to generate a fused image.

[0024] Further, multi-physical field concentration calibration is performed on the spatial mapping dataset to obtain a calibrated concentration dataset, including:

[0025] According to the synchronized distance signals and the anti-shake mapping region, the field of view distance of each region is calculated to obtain a field of view correction factor.

[0026] According to the concentration value and the field of view correction factor, a distance-dependent calibration is performed to generate a spatially calibrated concentration value.

[0027] Obtain temperature and humidity sensor data, and establish a concentration-temperature-humidity coupling model based on the temperature and humidity sensor data and the spatial calibration concentration value to generate a multi-field coupling concentration value.

[0028] Further, structure-guided field reconstruction is performed based on the calibration concentration dataset to obtain a structure-optimized concentration field, including:

[0029] Based on the anti-shake mapping area and the multi-field coupling concentration value, analyze the concentration gradient of the adjacent area to generate a spatial interpolation parameter set.

[0030] According to the spatial interpolation parameter set and the scene feature coordinates, an interpolation path is established along the scene structure edge to obtain a structure-guided interpolation parameter set.

[0031] Concentration surface fitting is performed using the structure-guided interpolation parameter set to output the structure-optimized concentration field.

[0032] Further, according to the spatial interpolation parameter set and the scene feature coordinates, an interpolation path is established along the scene structure edge to obtain a structure-guided interpolation parameter set, including:

[0033] All continuous edge point coordinates are identified from the scene feature coordinates, and the neighborhood interpolation weight is extracted from the spatial interpolation parameter set according to the edge point coordinates.

[0034] The interpolation weight is redistributed with the edge direction as the axis to output the structure-guided interpolation parameter set.

[0035] Further, trace enhancement image fusion is performed on the structure-optimized concentration field to generate a fused image, including:

[0036] According to the scene feature coordinates, a segmentation area is performed on the reference image to generate a partition fusion template.

[0037] According to the partition fusion template, the structure-optimized concentration field is converted to generate a semi-transparent pseudo-color layer.

[0038] The semi-transparent pseudo-color layer is superimposed on the reference image to generate a visually enhanced image.

[0039] According to the visually enhanced image and the multi-field coupling concentration value, a concentration gradient mutation point is detected to generate a leakage source positioning coordinate.

[0040] Further, the fused image is output to generate an output result, including:

[0041] Device state information is embedded in the fused image to generate a final display image.

[0042] The final display image is sent to a display module for real-time rendering to generate a visual output.

[0043] In a second aspect, the present application provides a gas imaging system for visual sensing data fusion, comprising:

[0044] Original signal acquisition and synchronization module: obtain original sensor signals, perform multi-source synchronization and feature extraction on the original sensor signals, and obtain synchronized feature data.

[0045] Anti-disturbance space registration processing module: perform anti-disturbance space registration processing based on the synchronized feature data set, and generate a space mapping data set.

[0046] Multi-source enhanced image fusion module: perform multi-source enhanced image fusion processing on the space mapping data set, and obtain a fused image.

[0047] Fused image output and result generation module: output the fused image and generate an output result.

[0048] In a third aspect, the present application provides a gas imaging device for visual sensing data fusion, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to realize the steps of the gas imaging method for visual sensing data fusion according to the first aspect.

[0049] In a fourth aspect, the present application provides a storage medium, wherein the storage medium stores computer program instructions, and the computer program instructions are read and executed by a processor to perform the steps of the gas imaging method for visual sensing data fusion according to the first aspect.

[0050] Advantages of the present application:

[0051] The present application realizes high-precision integration and visual output of gas multi-point detection data by performing synchronization and feature extraction, anti-disturbance space registration, and multi-source enhanced image fusion processing on the original signals of multi-source sensors, effectively solves the problem that the sensor is easily affected by response characteristic differences and environmental interference in the prior art, and the precision of the gas concentration distribution analysis result is not enough, can improve the multi-source signal synchronization precision and feature extraction reliability, enhance the anti-disturbance ability of space registration, optimize the multi-source data fusion quality, make the gas concentration distribution analysis more accurate, and meet the high-precision detection requirements of industrial scenes.

[0052] Other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be understood by those skilled in the art from the description. The purpose and other advantages of the present application can be realized and obtained by the structure indicated in the description and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0054] Figure 1 A flowchart of a gas imaging method of visual sensing data fusion of the present application is shown. DETAILED DESCRIPTION

[0055] In order to solve the problems raised in the background art, the present application realizes high-precision integration and visual output of gas multi-point detection data by synchronizing and extracting features of multi-source sensor original signals, anti-disturbance spatial registration and multi-source enhanced image fusion processing, which can improve the synchronization accuracy and feature extraction reliability of multi-source signals, enhance the anti-disturbance ability of spatial registration, optimize the quality of multi-source data fusion, make the gas concentration distribution analysis more accurate, and meet the high-precision detection requirements of industrial scenes.

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0057] In some embodiments, as shown in Figure 1 The present application provides a gas imaging method of visual sensing data fusion, comprising:

[0058] S1. Obtain original sensor signals, perform multi-source synchronization and feature extraction on the original sensor signals, and obtain synchronization feature data.

[0059] S2. Perform anti-disturbance spatial registration processing based on the synchronization feature data set, and generate a spatial mapping data set.

[0060] S3. Perform multi-source enhanced image fusion processing on the spatial mapping data set, and obtain a fusion image.

[0061] S4. Output the fusion image to generate an output result.

[0062] In some embodiments, the original sensor signals include camera video frames, gas sensor array signals, and laser ranging signals.

[0063] The camera video frame is a sequence of continuous images collected by the camera, and each frame contains two-dimensional pixel information of the detected environment, which is used to obtain the visual features of the scene.

[0064] The gas sensor array signal is generated by a plurality of gas detection units distributed in a grid, and each detection unit independently outputs a concentration value of the target gas.

[0065] For example, the gas sensor array adopts a 4x4 physical layout and is installed on the detection plane. When the methane gas reaches the lower explosive limit in the air, each detection unit outputs a corresponding electrical signal, which is converted to a 0-1000 digital value after analog-to-digital conversion.

[0066] The laser ranging signal is generated by the laser ranging module emitting and receiving laser pulses, and the distance between the detection plane and the module is determined by calculating the flight time of the laser.

[0067] In some embodiments, the original sensor signals in S1 are subjected to multi-source synchronization and feature extraction to obtain synchronized feature data, including:

[0068] S11. Time stamp alignment of the camera video frame, gas sensor array signal, and laser ranging signal is performed to generate synchronized video frames, synchronized concentration signals, and synchronized distance signals.

[0069] The camera video frame, gas sensor array signal, and laser ranging signal are subjected to time stamp alignment processing. Through the hardware clock synchronization module, the collection time of the three signals is locked to the rising edge of the same trigger pulse, so that the time deviation is controlled within the allowable range, thereby generating synchronized video frames, synchronized concentration signals, and synchronized distance signals that are strictly aligned in time.

[0070] For example, in the monitoring of a chemical plant reactor, when the synchronization pulse is triggered, the camera records the real-time image of the valve as the synchronized video frame, the gas sensor records the set of methane concentration values at the corresponding time as the synchronized concentration signal, and the laser ranging module records the target surface distance as the synchronized distance signal.

[0071] S12. Extract the current frame of the synchronized video frame, identify the key feature points of the scene, and generate a reference image and scene feature coordinates.

[0072] The current frame is extracted from the synchronized video frame, and the gradient amplitude of each pixel in the frame is calculated. When the gradient amplitude of a pixel point is greater than or equal to a preset gradient threshold, it is determined to be a candidate feature point. The non-maximum suppression algorithm is used to remove redundant points in the neighborhood, and the final key feature points of the scene are output. The gradient threshold can be determined by statistical analysis of an industrial scene image library.

[0073] The current frame of the synchronized video frame is directly stored as the reference image.

[0074] Extract all the scene key feature points' two-dimensional pixel coordinates and store them as a set to obtain the scene feature coordinates.

[0075] S13. Receive the physical position parameters of the gas sensor array, and bind the synchronous concentration signal to generate a position-bound concentration data set.

[0076] Bind the spatial coordinates of the i-th sensor ( , , ) with the concentration value in the synchronous concentration signal to generate a position-bound concentration data set.

[0077] In some embodiments, the anti-disturbance spatial registration processing in S2 based on the synchronous feature data set generates a spatial mapping data set, including:

[0078] S21. Based on the scene feature coordinates and the physical position parameters of the sensor array, calculate the relative position relationship between the feature points and the detection unit, and generate a dynamic registration matrix.

[0079] The scene feature coordinates provide a set of two-dimensional pixel coordinates of key feature points in the image space (such as the pixel coordinates of the flange bolt center point), and the physical position parameters of the sensor array provide the actual installation coordinates of the detection unit in the three-dimensional space (such as the three-dimensional position of the sensor on the pipe surface).

[0080] By establishing the mapping relationship between the image pixel coordinate system and the actual space coordinate system, the geometric transformation matrix is solved, and the dynamic registration matrix is generated by fitting the corresponding relationship between the image feature points and the actual detection unit position through the least squares method.

[0081] The dynamic registration matrix describes the geometric transformation relationship between the image feature points and the actual detection unit position.

[0082] S22. Obtain the device gyroscope attitude data, and compensate the matrix offset caused by vibration based on the dynamic registration matrix and the device gyroscope attitude data to generate an anti-shake registration matrix.

[0083] The gyroscope attitude data records the angular velocity and acceleration of the device in real time, and the instantaneous tilt angle of the device in the three-dimensional space is calculated by integration. The dynamic registration matrix is rotated and transformed according to the tilt angle to compensate for the matrix parameter offset caused by vibration, and an anti-shake registration matrix is generated.

[0084] S23. Map the detection unit position to the image pixel space using the anti-shake registration matrix and the reference image to generate an anti-shake mapping region.

[0085] The reference image provides a visual reference of the scene without superimposed concentration information, and the anti-shake registration matrix defines the conversion relationship from the three-dimensional space coordinates to the image pixel coordinates.

[0086] The three-dimensional installation coordinates of each detection unit are converted into pixel coordinates in the reference image by matrix multiplication operation to generate the anti-shake mapping area.

[0087] The anti-shake mapping area describes the coverage range of each detection unit in the image.

[0088] Exemplarily, in the gas pipeline monitoring, the three-dimensional coordinates of 16 sensors are mapped to the pipeline reference image by the anti-shake registration matrix to generate 16 rectangular areas as the anti-shake mapping areas, each of which corresponds to the position of a sensor in the image.

[0089] In some embodiments, the multi-source enhanced image fusion processing is performed on the spatial mapping data set in S3 to obtain a fusion image, including:

[0090] S31. The multi-physical field concentration calibration is performed on the spatial mapping data set to obtain a calibrated concentration data set.

[0091] S32. The structure-guided field reconstruction is performed based on the calibrated concentration data set to obtain a structure-optimized concentration field.

[0092] S33. The structure-optimized concentration field is subjected to source enhanced image fusion to generate a fusion image.

[0093] In some embodiments, the multi-physical field concentration calibration is performed on the spatial mapping data set in S31 to obtain a calibrated concentration data set, including:

[0094] S311. The field of view distance of each area is calculated according to the synchronization distance signal and the anti-shake mapping area to obtain a field of view correction factor.

[0095] The synchronization distance signal provides a physical distance measurement value L of the target surface, and the anti-shake mapping area describes the pixel coverage range of each detection unit in the reference image . The actual distance of each area is calculated through the geometric projection relationship , ; wherein, represents the included angle between the normal direction of the pixel coverage range and the camera sight direction, which can be calculated through the conversion relationship between the image coordinate system and the spatial coordinate system.

[0096] The field of view correction factor is: ; wherein, represents the calibration reference distance of the gas sensor, which is used to compensate for the concentration measurement attenuation caused by the distance change.

[0097] Exemplarily, in the gas pipeline monitoring, the physical distance measurement value L = 2.5 m, the flange area , then The calculation result is about 2.41 m, at which about 0.172, if 2.47 m, then about 0.164 m.

[0098] S312. Perform distance-dependent calibration according to the concentration value and the field-of-view correction factor to generate a spatially calibrated concentration value.

[0099] Concentration value The original concentration measurement value in the position-bound concentration dataset, the spatially calibrated concentration value is: .

[0100] S313. Obtain the temperature and humidity sensor data, and establish a concentration-temperature-humidity coupling model according to the temperature and humidity sensor data and the spatially calibrated concentration value to generate a multi-field coupled concentration value.

[0101] The temperature and humidity sensor data provide the ambient temperature T and the relative humidity H, and the concentration value is calculated by a linear compensation model , wherein, , represent the reference temperature and humidity at the time of sensor calibration, represent the temperature compensation coefficient, represent the humidity compensation coefficient, , All of which can be determined by sensor calibration experiments.

[0102] In some embodiments, the structure-guided field reconstruction based on the calibrated concentration dataset in S32 obtains a structure-optimized concentration field, including:

[0103] S321. Analyze the concentration gradient between adjacent regions based on the anti-shake mapping region and the multi-field coupled concentration value to generate a spatial interpolation parameter set.

[0104] The spatial position relationship of the detection unit in the image is determined according to the anti-shake mapping region, including the boundary and center coordinate points of each region.

[0105] For any two regions adjacent in space, the concentration gradient between them is calculated, including gradient value and direction information.

[0106] The gradient parameters are combined with the initial interpolation weight to form a spatial interpolation parameter set.

[0107] S322. Establish an interpolation path along the edge of the scene structure according to the spatial interpolation parameter set and the scene feature coordinates to obtain a structure-guided interpolation parameter set.

[0108] S323. Perform concentration surface fitting with the structure-guided interpolation parameter set, and output the structure-optimized concentration field.

[0109] The standard radial basis function interpolation method in computer graphics is adopted, the optimized weight provided by the structure-guided interpolation parameter set and the region center coordinates of the anti-shake mapping region are taken as input parameters, the characteristic scale is calculated in combination with the neighborhood average distance, the gas concentration value of each pixel point in the reference image is calculated through spatial distance weighted superposition, and continuous concentration distribution data covering the entire scene is generated, which is marked as the structure-optimized concentration field.

[0110] In some embodiments, the interpolation path is established along the edge of the scene structure according to the spatial interpolation parameter set and the scene feature coordinates in S322, and the structure-guided interpolation parameter set is obtained, including:

[0111] S3221. Identify all continuous edge point coordinates from the scene feature coordinates, and extract the neighborhood interpolation weight from the spatial interpolation parameter set according to the edge point coordinates.

[0112] All continuous edge point coordinates are identified from the scene feature coordinates, which are the point set (such as the pipeline center line or the tank contour line) constituting the continuous edge of the device structure in the scene feature coordinates, and then the interpolation weight value corresponding to the detection region in the spatial neighborhood is extracted from the spatial interpolation parameter set according to the position of the edge point coordinates.

[0113] S3222. Redistribute the interpolation weight with the edge direction as the axis, and output the structure-guided interpolation parameter set.

[0114] The axis is calculated through the geometric relationship of the continuous edge point sequence, and when the interpolation weight is redistributed, the fixed multiple of the original interpolation weight can be set through experimental calibration based on the device structure type and the detection accuracy requirement, and the redistributed interpolation weight is obtained, such as enhancing the region weight value in the direction of parallel edge direction and weakening the region weight value in the direction perpendicular to the edge direction, and finally outputting the set containing all region optimization weight values as the structure-guided interpolation parameter set.

[0115] In some embodiments, the structure-optimized concentration field is subjected to trace enhancement image fusion in S33, and a fused image is generated, including:

[0116] S331. Perform segmentation region on the reference image according to the scene feature coordinates, and generate a partition fusion template.

[0117] The image is divided into multiple independent sub-regions through the polygonal region formed by connecting the feature points, each sub-region corresponds to a functional unit (such as a valve group or a flange sealing surface) of the device, and the partition fusion template containing the region boundary information is output.

[0118] S332. Generate a semi-transparent false color layer according to the partition fusion template conversion structure optimized concentration field.

[0119] By mapping different concentration intervals to color blocks with specific transparency (e.g., 0-50 ppm light blue with 70% transparency, 50-100 ppm green with 50% transparency), a visually recognizable false color layer is generated by coloring within each partition. The semi-transparent false color layer is output and superimposed on the original scene.

[0120] S333. Superimpose the semi-transparent false color layer on the reference image to generate a visual enhancement image.

[0121] The standard alpha blending method of image processing is used to realize layer superimposition. The reference image provides the details of the underlying scene, and the semi-transparent false color layer provides concentration visualization information. The visual enhancement image after superimposition retains both the device structure details and the gas distribution characteristics.

[0122] S334. Detect concentration gradient mutation points according to the visual enhancement image and multi-field coupled concentration values, and generate leakage source positioning coordinates.

[0123] By identifying the positions where the concentration difference between adjacent regions exceeds the threshold, and determining the maximum mutation point coordinates in combination with the gradient direction, the spatial position information of the leakage source, i.e., the leakage source positioning coordinates, is obtained.

[0124] In some embodiments, the fusion image in S4 is output to generate an output result, including:

[0125] S41. Embed device status information in the fusion image to generate a final display image.

[0126] The device status information embedded in the fusion image to generate the final display image includes but is not limited to real-time parameters such as battery power, device number, and time stamp.

[0127] The embedding operation is realized through image superimposition technology. Specifically, the status information text or icons are drawn in the reserved area of the image (such as the top status bar), and are fused with the fusion image at the pixel level to ensure that the status information and scene content are visible at the same time and do not block each other. The final display image containing both gas distribution information and device status is output.

[0128] S42. Send the final display image to the display module for real-time rendering to generate a visual output.

[0129] The display module is specifically a hardware display system, such as an LCD screen or an AR glasses display screen. Real-time rendering uses the industry standard graphics library to realize texture mapping and rasterization processing, converting the pixel matrix of the final display image into physical driving signals of the display device, and generating a visual output that can be directly observed by the human eye.

[0130] In some embodiments, the present application also provides a gas imaging system for visual sensing data fusion, comprising:

[0131] A raw signal acquisition and synchronization module: obtaining raw sensor signals, performing multi-source synchronization and feature extraction on the raw sensor signals, and obtaining synchronized feature data.

[0132] A disturbance-resistant spatial registration processing module: performing disturbance-resistant spatial registration processing based on the synchronized feature data set, and generating a spatial mapping data set.

[0133] A multi-source enhanced image fusion module: performing multi-source enhanced image fusion processing on the spatial mapping data set, and obtaining a fused image.

[0134] A fused image output and result generation module: outputting the fused image and generating an output result.

[0135] In some embodiments, the present application also provides a gas imaging device for visual sensing data fusion, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to realize the steps of a gas imaging method for visual sensing data fusion.

[0136] In some embodiments, the present application also provides a storage medium, which stores computer program instructions; when the computer program instructions are read and run by a processor, the steps of a gas imaging method for visual sensing data fusion are executed.

[0137] In some embodiments, any reference to memory, storage, databases, or other media in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory.

[0138] It should be noted that, in the present document, relational terms such as "first" and "second", and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0139] Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood that modifications can be made to the foregoing embodiments, or additional implementations can be implemented, without departing from the spirit and scope of the inventive subject matter. Accordingly, the present application is not limited to the implementations described herein, but is intended to cover all modifications and equivalents falling within the spirit and scope of the inventive subject matter.

Claims

1. A gas imaging method based on visual sensor data fusion, characterized in that: include: Acquire raw sensor signals, perform multi-source synchronization and feature extraction on the raw sensor signals, and obtain synchronized feature data; wherein the synchronized feature data includes a reference image, scene feature coordinates, physical position parameters of the sensor array, and a concentration data set; Based on the scene feature coordinates and the physical position parameters of the sensor array, the relative position relationship between the feature points and the detection unit is calculated to generate a dynamic registration matrix; Obtain the device's gyroscope attitude data, and based on the dynamic registration matrix and the device's gyroscope attitude data, compensate for the matrix offset caused by vibration to generate an anti-shake registration matrix. The anti-shake registration matrix and the reference image are used to map the detection unit position to the image pixel space to generate the anti-shake mapping area; Calculate the field of view distance of each area based on the synchronized distance signal and the anti-shake mapping area to obtain the field of view correction factor; performing distance-dependent calibration according to the concentration values ​​in the concentration data set and the field of view correction factor to generate spatially calibrated concentration values; Acquire temperature and humidity sensor data, establish a concentration-temperature-humidity coupling model based on the temperature and humidity sensor data and spatial calibration concentration values, and generate multi-field coupled concentration values; Analyze the concentration gradients in adjacent regions based on the anti-shake mapping area and multi-field coupling concentration values ​​to generate a spatial interpolation parameter set; An interpolation path is established along the edge of the scene structure according to the spatial interpolation parameter set and the scene feature coordinates to obtain a structure-guided interpolation parameter set; Perform concentration surface fitting using the structure-guided interpolation parameter set and output the structure-optimized concentration field; Perform source-tracing enhanced image fusion on the structure-optimized concentration field to generate a fused image; Output the fused image to generate an output result.

2. The gas imaging method based on visual sensor data fusion according to claim 1, characterized in that: The raw sensor signals include camera video frames, gas sensor array signals, and laser ranging signals.

3. The gas imaging method based on visual sensor data fusion according to claim 2, characterized in that: Perform multi-source synchronization and feature extraction on the original sensor signals to obtain synchronized feature data, including: Perform multi-source time stamp alignment on camera video frames, gas sensor array signals, and laser ranging signals to generate synchronized video frames, synchronized concentration signals, and synchronized distance signals; Extract the current picture of the synchronized video frame, identify the key feature points of the scene, and generate the reference image and scene feature coordinates; The physical position parameters of the gas sensor array are received and combined with the synchronized concentration signal to generate a position-bound concentration dataset.

4. The gas imaging method based on visual sensor data fusion according to claim 1, characterized in that: An interpolation path is established along the edge of the scene structure according to the spatial interpolation parameter set and the scene feature coordinates to obtain a structure-guided interpolation parameter set, including: Identify all continuous edge point coordinates from the scene feature coordinates, and extract neighborhood interpolation weights from the spatial interpolation parameter set based on the edge point coordinates; The interpolation weights are redistributed along the edge direction, and the structure-guided interpolation parameter set is output.

5. The gas imaging method based on visual sensor data fusion according to claim 1, characterized in that: Perform source-enhanced image fusion on the structure-optimized concentration field to generate a fused image, including: Segment the reference image into regions according to the scene feature coordinates and generate a partition fusion template; According to the partition fusion template conversion structure, the concentration field is optimized to generate a semi-transparent pseudo-color layer; Superimposing a semi-transparent pseudo-color layer onto a reference image to generate a visually enhanced image; The concentration gradient mutation point is detected based on the visually enhanced image and multi-field coupled concentration values, and the leakage source positioning coordinates are generated.

6. The gas imaging method based on visual sensor data fusion according to claim 1, characterized in that: Output the fused image and generate output results, including: Embed device status information in the fused image to generate the final display image; The final display image is sent to the display module for real-time rendering to generate visual output.

Images