Explosion flame image recognition and multi-parameter analysis method

Abstract

The invention relates to the technical field of explosion flames, in particular to an explosion flame image recognition and multi-parameter analysis method, which comprises the following steps: acquiring and storing an explosion flame video; extracting image frames containing effective flame features from the video; sequentially processing the image frame, normalizing the image and suppressing noise; analyzing the flame profile to identify the front position of the flame, calculating the flame propagation speed in combination with the pixel physical scale and the time interval, and identifying the flame temperature distribution according to the mapping relation between the gray value and the temperature; and the extracted speed and temperature parameters are analyzed and processed, and various visual results are automatically generated. In result verification, manual calculation has uncertainty such as subjective visual deviation of an operator, system errors of different time periods or calculation of different operators and the like. The relative deviation between the flame propagation speed calculated by the method and the manual calculation result is less than or equal to 10%, and efficient automatic extraction and visual display of multiple parameters of the explosion flame are realized.

Description

Technical Field

[0001] This invention relates to the field of explosion flame technology, and in particular to a method for explosion flame image recognition and multi-parameter analysis. Background Technology

[0002] The study of the dynamic characteristics of explosion flames is of vital importance in fields such as aerospace, energy and chemical engineering, public safety, and disaster prevention. Accurately obtaining key parameters such as flame propagation speed and temperature distribution is the core foundation for assessing explosion power, understanding combustion mechanisms, and building safety models.

[0003] Currently, video image-based detection technology has been applied to explosion flame analysis. For example, patent application number 202310232317.7, entitled "A Method for Image Recognition of Dust Explosion Flames," discloses the following steps: capturing a video of a dust explosion flame; extracting a valid target video from the dust explosion flame video; extracting an initial image of the dust explosion flame within the valid target video; preprocessing the initial image of the dust explosion flame; processing the initial image of the dust explosion flame; performing contour processing on the image; selecting the range of the dust explosion flame, outlining the flame contour, and obtaining the horizontal and vertical data of the dust explosion flame image. However, this existing technology is mainly aimed at dust explosion flames and has significant limitations in terms of versatility, multi-parameter extraction, and automated analysis.

[0004] Specifically, existing technologies have not yet solved core problems in explosion flame analysis, such as reliance on manual experience, limited parameter extraction, and poor scene adaptability. Specifically, this manifests as: a lack of universal recognition capabilities for other types of explosion flame images; failure to automatically acquire explosion flame propagation speed; failure to automatically extract and visualize explosion flame temperature parameters; and failure to automatically generate analytical charts showing the changes in flame propagation speed and temperature over time. Therefore, there is an urgent need to propose a comprehensive method that integrates image recognition and multi-parameter analysis to achieve efficient, automatic extraction and visualization of multiple parameters such as explosion flame propagation speed and temperature distribution. Summary of the Invention

[0005] The purpose of this invention is to provide a method for image recognition and multi-parameter analysis of explosion flames, which solves the problems of existing technologies that rely on manual experience, extract only single parameters, have poor scene adaptability, and cannot automatically and comprehensively obtain multi-parameter information such as the propagation speed and temperature of explosion flames.

[0006] To achieve the above objectives, the present invention provides a method for explosion flame image recognition and multi-parameter analysis, comprising the following steps: High-speed cameras were used to capture video footage of the explosion flames, and the video storage path was set. Based on frame-by-frame analysis or fixed-time interval sampling, image frames containing valid explosion flame features are extracted from the video data, and invalid interference frames are filtered out. The extracted valid image frames are preprocessed and feature normalization operations are performed sequentially to generate image frame data samples that meet the input specifications of the recognition model. In this process, the feature recognition and extraction of the propagation speed and temperature of the explosion flame are completed in a coordinated manner. Based on the extracted propagation speed and temperature characteristics, analysis and processing are performed to generate at least one visualization result among flame propagation speed curve, flame temperature change curve, and flame temperature cloud map.

[0007] Specifically, based on frame-by-frame analysis or fixed-time interval sampling, image frames containing valid explosion flame features are extracted from the video data, and invalid interference frames are filtered out. The duration and frame rate of the video data are obtained, the total number of video frames is calculated, and the corresponding image frame sequence is extracted based on the total number of frames.

[0008] The process involves preprocessing and feature normalization of the extracted valid image frames to generate image frame data samples that conform to the input specifications of the recognition model. During this process, the feature recognition and extraction of the explosion flame propagation speed and temperature are collaboratively completed, specifically including: The preprocessing and feature normalization operations include image grayscale conversion, binarization and morphological processing. The image grayscale conversion is to fuse the pixel values ​​of the red, green and blue channels of a color image according to a preset weight and convert them into a single-channel grayscale image. The binarization is achieved by setting a threshold, setting pixels with values ​​below the threshold to 0, and pixels with values ​​above the threshold to 255, thus obtaining a binary image. The morphological processing includes first performing a closing operation on the binary image to eliminate noise in dark areas, and then performing an opening operation to eliminate noise in bright areas. The closing operation is a process of dilation followed by erosion, and the opening operation is a process of erosion followed by dilation.

[0009] The process involves preprocessing and feature normalization of the extracted valid image frames to generate image frame data samples that conform to the input specifications of the recognition model. During this process, the feature recognition and extraction of the explosion flame propagation speed and temperature are collaboratively completed. Specifically, this also includes: For the morphologically processed image, a minimum contour area threshold is set to filter out the valid flame contours; Calculate the rightmost coordinate of the effective flame profile point set as the position of the flame front in this frame; The instantaneous physical propagation speed of the flame is calculated based on the pixel displacement of the flame front position in two adjacent frames, the pre-calibrated pixel physical scale, and the time interval between the two frames. Based on a preset grayscale-temperature conversion coefficient and a base temperature, the grayscale value of each pixel in the grayscale image is linearly converted into the corresponding physical temperature value.

[0010] Specifically, based on the extracted propagation speed and temperature characteristics, analysis and processing are performed to generate at least one visualization result among flame propagation speed curve, flame temperature change curve, and flame temperature cloud map, including: The calculated series of instantaneous physical propagation velocities are smoothed using a moving average. The temperature field data extracted from all image frames is normalized, and the temperature values ​​are mapped to a preset numerical range. The normalized values ​​are converted into color pixel values ​​through a pseudo-color mapping scheme to generate a color temperature cloud map that reflects the spatial distribution of temperature.

[0011] This invention discloses a method for image recognition and multi-parameter analysis of explosion flames. First, it acquires and stores video footage of explosion flames using a high-speed camera. Then, it extracts image frames containing effective flame features from the video. Next, it performs grayscale conversion, binarization, and morphological processing on the image frames to regularize the image and suppress noise. Based on this, it identifies the flame front position by analyzing the flame contour, calculates the flame propagation speed by combining pixel physical scale and time interval, and identifies the flame temperature distribution based on the mapping relationship between grayscale values ​​and temperature. Finally, it analyzes and processes the extracted speed and temperature parameters, automatically generating various visualization results such as flame propagation speed curves, temperature change curves, and temperature cloud maps. This achieves comprehensive, automatic analysis and intuitive display of multiple parameters of explosion flames, providing comprehensive and accurate analytical results for related analysis and research on explosion flames. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0013] Figure 1 This is a flowchart of the explosion flame image recognition and multi-parameter analysis method of the present invention.

[0014] Figure 2 This is a screenshot of the flame from the present invention.

[0015] Figure 3 This is a comparison chart of the instantaneous flame propagation speed curves calculated manually and calculated by the present invention.

[0016] Figure 4 This is a graph showing the change of flame temperature over time according to the present invention.

[0017] Figure 5This is a schematic diagram of the temperature cloud map sequence of the present invention.

[0018] Figure 6 This is a flowchart of the steps of the explosion flame image recognition and multi-parameter analysis method of the present invention. Detailed Implementation

[0019] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0020] Please see Figures 1 to 6 This invention provides a method for image recognition and multi-parameter analysis of explosion flames, comprising the following steps: S101: Use a high-speed camera to collect video data of the explosion flames and set the video storage path; Specifically, the explosion flame video data is collected by a high-speed camera, the storage location of the explosion flame video data collected by the high-speed camera is set, the storage location is obtained and the video capture object is initialized, and then the video frame rate F (1000FPS), the total number of video frames P (232 frames) and the video resolution (1024×720 pixels) are obtained to provide a basis for subsequent processing.

[0021] S102: Based on frame-by-frame analysis or fixed time interval sampling, extract image frames containing valid explosion flame features from the video data, and filter out invalid interference frames; Specifically, it reads images in a loop by frame index, supports frame skipping, and efficiently filters out image frames containing valid explosion flame features, thus improving data processing efficiency.

[0022] S103: The extracted valid image frames are preprocessed and feature normalization operations are performed sequentially to generate image frame data samples that conform to the input specifications of the recognition model. In this process, the feature recognition and extraction of the propagation speed and temperature of the explosion flame are completed in a coordinated manner. Specifically, the effective image frames are preprocessed and feature normalization is performed. First, the image is converted into a grayscale image. Then, the pixel values ​​of the red (R), green (G), and blue (B) channels of the flame color image are fused according to specific weights to obtain the grayscale value of a single channel, ultimately presenting a "black, white, and gray" visual effect.

[0023] The grayscale image is binarized by setting a threshold to divide the pixels into two categories: pixels below the threshold are set to 0 (black), and pixels above the threshold are set to 255 (white). This process simplifies the image information and enhances the contrast between the target flame and the background.

[0024] Morphological operations are performed on the binary image. First, a closing operation is used to eliminate noise in dark areas. Then, a structuring element is used to dilate the original flame image, filling in small holes within the effective flame area or connecting adjacent flame areas. Next, an erosion operation is performed on the dilated image to remove redundant boundaries generated during dilation, eliminating dark spots and other dark noise areas that shouldn't be present within the flame target, making the original flame-related bright areas more complete while preserving the shape of the main flame structure. The formula can be expressed as A•B=(A⊕B)⊖B, where A is the input image and B is the structuring element. Next, an opening operation is used to eliminate noise in bright areas. Then, an erosion operation is performed on the image after the closing operation using the structuring element to remove bright spots and other interfering objects smaller than the structuring element. Finally, a dilation operation is performed on the eroded image to restore the eroded flame edges, making the original flame-related dark areas more complete while preserving the shape and structure of the main flame. The formula can be expressed as A∘B=(A⊖B)⊕B.

[0025] Finally, the position of the flame front is detected. First, a minimum area threshold is set, retaining only the contours of individual target flames with an area greater than this threshold, ignoring interference contours with too small an area, thus accurately locking the contour region corresponding to the main body of the flame. After determining the effective flame contour, the rightmost coordinates of the contour point set are further calculated. Finally, the flame propagation speed is calculated by obtaining the rightmost coordinates of the flame front in two adjacent effective images, calculating the difference between the two in the pixel dimension to obtain the pixel displacement of the flame front, and then combining it with a pre-calibrated "pixel physical scale" to convert the pixel displacement into the actual physical displacement of the flame front. Then, the time interval between two adjacent images is determined, and finally, the actual physical displacement is divided by the time interval to obtain the instantaneous physical propagation speed of the flame within that time period. Simultaneously, grayscale values ​​are converted into temperature values. Using the grayscale values ​​of pixels in the image, combined with a pre-determined "grayscale-temperature conversion coefficient" and "base temperature," the physical temperature at the corresponding pixel position is linearly calculated, and the highest temperature in each frame is recorded. The formula is as follows: in This is the temperature conversion coefficient. Base temperature.

[0026] S104: Based on the extracted propagation speed and temperature characteristics, perform analysis and processing, and generate at least one visualization result among the flame propagation speed curve, flame temperature change curve, and flame temperature cloud map.

[0027] Specifically, the calculated instantaneous physical propagation velocity of the flame is smoothed using a moving average. Each data point in the time series is replaced by the arithmetic mean of a fixed number of neighboring data points. The core mechanism of this method is to select a window containing surrounding data (e.g., 5 data points before and after) for each target point, and calculate the mean of all values ​​within the window as the smoothed value for that point.

[0028] Then, each image frame is processed individually to extract and record the maximum and minimum temperature values. The temperature field is then normalized to a global range. Normalization is a linear transformation that maps data to a specified range, removing the dimensions and units of different data dimensions, making it easier to compare and weight indicators of different units or magnitudes. This is achieved through formulas... Map the original temperature value to the range of 0-1. This represents the highest temperature in the temperature field. This is the lowest temperature in the temperature field; after this processing, all temperature values ​​are "compressed" to 0 (corresponding to...). ) to 1 (corresponding to This achieves dimensional uniformity, facilitating subsequent color mapping. Multiplying the normalized result by 255 expands the numerical range from 0-1 to 0-255. 255 is the maximum pixel brightness value in the most common image formats; for example, in grayscale images, 0 represents pure black. 255 represents pure white, and the R, G, and B channels of a color image also take values ​​between 0 and 255. The purpose of this step is to adapt to the hardware / software standards for image display, so that the values ​​can be recognized by the image system as valid pixel values.

[0029] Finally, pseudo-color mapping is applied to generate a flame temperature cloud map. Pseudo-color image processing is a technique that converts grayscale images into color images. By mapping different grayscale values ​​to different colors, the visual effect of the image can be enhanced, highlighting details within the image. The core principle of pseudo-color processing is to map each pixel value in a grayscale image to a color space. Using the COLORMAP_JET color mapping scheme, values ​​from 0 to 255 are converted into color pixel values. The numerical values ​​are associated with colors according to fixed rules. For example, the low-temperature region of the flame (flame core) corresponds to blue, the medium-temperature region (inner flame) corresponds to green / yellow, and the high-temperature region (outer flame) corresponds to red. In this way, the originally abstract temperature data is visualized as a color image, with different colors intuitively reflecting the temperature differences; for example, the red area has the highest temperature, and the blue area has the lowest temperature. The formula is as follows: Finally, the results are visualized in real time, and the flame propagation speed curve, temperature change curve, and temperature cloud map sequence are output.

[0030] Result verification: The evaluation of the core parameters output by this invention, namely flame propagation speed and flame temperature, will strictly adhere to the technical specifications. This evaluation will combine actual measurement data with recognized physical properties to verify the accuracy and rationality of each parameter.

[0031] To verify the accuracy and reliability of the method of the present invention, this embodiment uses manual calculation as one of the comparison benchmarks. It should be noted that manual calculation is inherently affected by the operator's subjective factors and has certain uncertainties. For example, there may be visual bias when manually selecting the position of the flame front; calculations performed by different operators or by the same operator at different times may contain systematic errors.

[0032] Nevertheless, this invention achieves the following through a fully automated image processing and parameter extraction process: continuous inter-frame processing based on frame-by-frame or fixed-interval sampling, avoiding the subjectivity of manual selection; algorithm consistency, with all calculation steps automatically executed based on a preset algorithm, ensuring repeatable results; and quantifiable errors, as comparative experiments show that even with fluctuations in manual calculations, the method of this invention still exhibits stable calculation accuracy, verifying its reliability in practical applications.

[0033] Verification of flame propagation speed: Verification basis: Through formula To calculate the instantaneous physical propagation speed, it is necessary to verify the consistency between the calculated value and the manually calculated instantaneous propagation speed of the flame.

[0034] Verification process: Obtain manual calculation Instantaneous propagation speed and the instantaneous propagation speed of the flame calculated by the patented method.

[0035] Table 1. Calculation of instantaneous flame propagation speed using manual calculation and patented methods. Deviation calculation and judgment: according to the formula Calculate the relative deviation.

[0036] The calculations show that the deviations are all ≤10%, indicating that the flame propagation speed calculated by the patented method conforms to the actual scene.

[0037] Key points to note: (1) Actual time It must be consistent with the time range of "frame time interval Δt" in the patented method; (2) The scale must be parallel to the direction of flame propagation to avoid inaccurate data measurement due to angular deviation.

[0038] Flame temperature verification: Verification basis: Through formula Converting grayscale values ​​to temperature requires verifying the calculated "highest temperature". "The degree of matching with the known temperature characteristics of the combustible material."

[0039] Verification steps: Step 1: Determine the temperature reference value: The combustion substance used in the experiment was hydrogen gas, and its theoretical maximum combustion temperature is known to be the baseline value. .

[0040] Step 2: Extract the temperature data calculated by this invention: The "flame maximum temperature versus time curve" generated by this invention (corresponding to) Figure 5 In the study, the highest average temperature during the stable combustion phase of the flame was extracted. At the same time, view the "Temperature Cloud Map Sequence" (corresponding to...) Figure 6 ), confirm the temperature value of the "high temperature area" in the cloud map and Consistent.

[0041] Step 3: Error Calculation and Judgment According to the formula If the relative error is calculated and found to be ≤10%, it indicates that the temperature conversion formula of the present invention is effective, the temperature calculation result is reliable, and it can accurately reflect the actual combustion temperature characteristics of the combustible substance.

[0042] This invention can analyze the flames in uploaded videos and detect the position of the flame front, thereby acquiring the flame propagation speed, temperature cloud map, and time change trend map of the explosion flame image, providing comprehensive and accurate analysis results for engineering.

[0043] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

[0044] This invention can analyze the flames in uploaded videos and detect the position of the flame front, thereby acquiring the flame propagation speed, temperature cloud map, and time change trend map of the explosion flame image, providing comprehensive and accurate analysis results for engineering.

[0045] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for image recognition and multi-parameter analysis of explosion flames, characterized in that, Includes the following steps: High-speed cameras were used to capture video footage of the explosion flames, and the video storage path was set. Based on frame-by-frame analysis or fixed-time interval sampling, image frames containing valid explosion flame features are extracted from the video data, and invalid interference frames are filtered out. The extracted valid image frames are preprocessed and feature normalization operations are performed sequentially to generate image frame data samples that meet the input specifications of the recognition model. In this process, the feature recognition and extraction of the propagation speed and temperature of the explosion flame are completed in a coordinated manner. Based on the extracted propagation speed and temperature characteristics, analysis and processing are performed to generate at least one visualization result among flame propagation speed curve, flame temperature change curve, and flame temperature cloud map.

2. The explosion flame image recognition and multi-parameter analysis method as described in claim 1, characterized in that, Based on frame-by-frame analysis or fixed-time interval sampling, image frames containing valid explosion flame features are extracted from the video data, and invalid interference frames are filtered out. Specifically, this includes: The duration and frame rate of the video data are obtained, the total number of video frames is calculated, and the corresponding image frame sequence is extracted based on the total number of frames.

3. The explosion flame image recognition and multi-parameter analysis method as described in claim 1, characterized in that, The extracted valid image frames are preprocessed and feature normalization operations are performed sequentially to generate image frame data samples that conform to the input specifications of the recognition model. During this process, the feature recognition and extraction of the explosion flame propagation speed and explosion flame temperature are completed collaboratively, specifically including: The preprocessing and feature normalization operations include image grayscale conversion, binarization and morphological processing. The image grayscale conversion is to fuse the pixel values ​​of the red, green and blue channels of a color image according to a preset weight and convert them into a single-channel grayscale image. The binarization is achieved by setting a threshold, setting pixels with values ​​below the threshold to 0, and pixels with values ​​above the threshold to 255, thus obtaining a binary image. The morphological processing includes first performing a closing operation on the binary image to eliminate noise in dark areas, and then performing an opening operation to eliminate noise in bright areas. The closing operation is a process of dilation followed by erosion, and the opening operation is a process of erosion followed by dilation.

4. The explosion flame image recognition and multi-parameter analysis method as described in claim 3, characterized in that, The extracted valid image frames are preprocessed and feature normalization operations are performed sequentially to generate image frame data samples that conform to the input specifications of the recognition model. During this process, the feature recognition and extraction of the explosion flame propagation speed and explosion flame temperature are completed collaboratively. Specifically, this also includes: For the morphologically processed image, a minimum contour area threshold is set to filter out the valid flame contours; Calculate the rightmost coordinate of the effective flame profile point set as the position of the flame front in this frame; The instantaneous physical propagation speed of the flame is calculated based on the pixel displacement of the flame front position in two adjacent frames, the pre-calibrated pixel physical scale, and the time interval between the two frames. Based on a preset grayscale-temperature conversion coefficient and a base temperature, the grayscale value of each pixel in the grayscale image is linearly converted into the corresponding physical temperature value.

5. The explosion flame image recognition and multi-parameter analysis method as described in claim 1, characterized in that, Based on the extracted propagation speed and temperature characteristics, analysis and processing are performed to generate at least one visualization result among flame propagation speed curve, flame temperature change curve, and flame temperature cloud map, specifically including: The calculated series of instantaneous physical propagation velocities are smoothed using a moving average. The temperature field data extracted from all image frames is normalized, and the temperature values ​​are mapped to a preset numerical range. The normalized values ​​are converted into color pixel values ​​through a pseudo-color mapping scheme to generate a color temperature cloud map that reflects the spatial distribution of temperature.

Images