CN-121366334-B - Fire hidden danger identification method and system based on image enhancement

CN121366334BCN 121366334 BCN121366334 BCN 121366334BCN-121366334-B

Abstract

The invention relates to the technical field of fire monitoring, in particular to a method and a system for identifying hidden fire hazards based on image enhancement. The method comprises the steps of synchronously acquiring thermal images through an infrared thermal imager array arranged on the top of the inner side wall of a silo, forming a panoramic thermal image sequence through registration and splicing, performing enhancement treatment, extracting candidate thermal spot areas, acquiring space-time dynamic evolution features of the thermal spot areas through cross-frame tracking, carrying out weighted fusion on the space dynamic evolution features and the space morphological features to obtain comprehensive confidence scores, and finally judging hidden danger of smoldering by combining with smoldering physical diffusion model verification. According to the invention, through fusing the dynamic evolution characteristics of the hot spot area with the verification of the physical model, the early and accurate identification of the hidden burning hidden trouble in the silo is realized, and the problems of high false alarm rate and early warning lag in the traditional static temperature threshold monitoring method are effectively solved.

Inventors

LIU YUXIN
ZHU YUE
FAN HAOHAO
SONG LINA
GAO JINGWEI

Assignees

城安盛邦(北京)网络科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20251010

Claims (8)

1. The method for identifying the fire hidden trouble based on image enhancement is characterized by comprising the following steps: an infrared thermal imager array is deployed at the top of the inner side wall of the silo, and comprises The infrared thermal imaging system comprises an infrared thermal imaging system; the infrared thermal imager array is used for synchronously collecting the interior of the silo A thermal image sub-sequence for the said Carrying out space-time registration and splicing on the sub-sequences of the thermal images to obtain a panoramic thermal image sequence covering the surface of the whole silo material, carrying out image enhancement treatment on the panoramic thermal image sequence to obtain an enhanced thermal image sequence, extracting candidate thermal spot areas from the enhanced thermal image sequence to obtain a candidate thermal spot area image sequence; The method comprises the steps of carrying out cross-frame tracking and matching on the same candidate hot spot region in the candidate hot spot region image sequence, calculating the spatial morphological feature and the time sequence dynamic evolution feature of the tracked candidate hot spot region according to a single frame of the candidate hot spot region image, carrying out weighted fusion on the spatial morphological feature and the time sequence dynamic evolution feature, and obtaining a comprehensive confidence coefficient score; when the comprehensive confidence score exceeds a preset confidence threshold and the time sequence dynamic evolution feature comprises a speed distribution feature deduced by a smoldering physical diffusion model, judging the candidate hot spot area as a smoldering hidden danger; The method for calculating the spatial morphological characteristics and the time sequence dynamic evolution characteristics of the tracked candidate hot spot area according to the single-frame candidate hot spot area image comprises the following steps: obtaining gray gradient distribution according to the single frame of the candidate hot spot area image through Gaussian filtering and second derivative operation, and obtaining local gray change intensity characteristic elements according to the gray gradient distribution Obtaining a contour point set of the candidate hot spot region through a contour extraction algorithm according to the single-frame candidate hot spot region image, and obtaining a contour fractal dimension feature element through calculation by a box counting method according to the contour point set And the local gray level change intensity characteristic element And profile fractal dimension feature element Combining to obtain a space feature vector; calculating the area of the candidate hot spot area in each frame to obtain an area change sequence according to the candidate hot spot area image sequence, and calculating to obtain an equivalent radius change rate characteristic element according to the area change sequence Obtaining a boundary moving speed sequence according to the candidate hot spot area image sequence through cross-frame contour point matching and average displacement calculation, and obtaining a boundary expansion speed feature element by adopting median filtering processing according to the boundary moving speed sequence And the equivalent radius change rate feature element And boundary flare speed feature element Combining to obtain a time sequence evolution feature vector; The method for obtaining the boundary moving speed sequence through cross-frame contour point matching and average displacement calculation according to the candidate hot spot region image sequence comprises the following steps: acquiring a hot spot area contour point set of the previous frame from the candidate hot spot area image sequence Hot spot area contour point set corresponding to current frame And according to the outline point set of the hot spot area of the current frame Calculating an external normal unit vector set of each contour point of the current frame, and carrying out hot spot area contour point set on the current frame A set of hot spot area contour points for each point in the previous frame Performing local nearest neighbor search to obtain a candidate corresponding point set, screening the candidate corresponding point set according to a preset maximum allowable displacement constraint and a maximum direction deviation constraint to obtain an effective corresponding point set, calculating according to the effective corresponding point set and an external normal unit vector set to obtain a normal displacement set, calculating according to the normal displacement set to obtain a median, calculating according to the median and an inter-frame time interval to obtain an instantaneous boundary expansion speed The instantaneous boundary expansion speed is increased And carrying out median filtering on the speed time sequence to obtain a boundary moving speed sequence.
2. The method for identifying fire hidden danger based on image enhancement according to claim 1, wherein the method for deploying the thermal infrared imager array on the top of the inner side wall of the silo comprises the following steps: Obtaining structural parameters of a target silo, wherein the structural parameters comprise silo inner diameter and silo wall height, obtaining device parameters of infrared thermal imagers, wherein the device parameters comprise horizontal view angles and vertical view angles, and calculating according to the silo inner diameter and the horizontal view angles to obtain the minimum number of infrared thermal imagers required for realizing 360-degree surrounding coverage Calculating the number of the thermal infrared imagers at the minimum according to the inner diameter of the silo, the height of the silo wall and the vertical field angle The required initial pitch angle is set down; determining the number of thermal infrared imagers at the minimum according to the structural parameters and the device parameters If the coverage blind area exists, the number of the thermal infrared imagers is increased iteratively and the pitch angle is recalculated until the coverage blind area is eliminated, so as to obtain the actually required number of the thermal infrared imagers Optimizing pitch angle ; According to the number of the thermal infrared imagers Calculated to obtain A deployment azimuth sequence of each thermal infrared imager, and optimizing a pitch angle according to the deployment azimuth sequence And arranging a thermal infrared imager array on the top of the inner side wall of the silo.
3. The method for identifying fire hidden trouble based on image enhancement according to claim 2, wherein the method for extracting the candidate hot spot area from the enhanced hot image sequence to obtain the candidate hot spot area image sequence comprises the following steps: dividing the silo material surface into the following parts according to the deployment azimuth sequence and the optimized pitch angle A non-uniform grid area for the said Each grid region in the grid regions is counted according to a panoramic thermal image sequence in a preset historical time period to obtain corresponding temperature data, and a reference temperature value of each grid region is calculated according to the corresponding temperature data And temperature fluctuation statistics Wherein Acquiring a frame of enhanced thermal image at the current moment, and calculating the real-time average temperature of each grid area in the frame of enhanced thermal image According to the reference temperature value Statistics of temperature fluctuations And real-time average temperature Calculating to obtain self-adaptive segmentation threshold values of M grid areas The adaptive segmentation threshold The calculation formula of (2) is as follows: ; Wherein, the And Is a preset weighting coefficient according to the following steps Adaptive segmentation threshold for individual grid regions The method comprises the steps of obtaining a local self-adaptive threshold segmentation image with the same size as the one-frame enhanced thermal image by adopting a bilinear interpolation algorithm, comparing the one-frame enhanced thermal image with the local self-adaptive threshold segmentation image pixel by pixel to obtain a binary image, wherein if the temperature value of a pixel point is higher than the threshold value of the corresponding position in the threshold segmentation image, the binary image is marked as a foreground pixel, otherwise, the binary image is marked as a background pixel, carrying out connected domain analysis on the binary image to obtain connected regions of all the foreground pixels, defining an circumscribed rectangular region of each connected region as a candidate thermal spot region, extracting the candidate thermal spot region from each frame enhanced thermal image in the enhanced thermal image sequence, and obtaining a candidate thermal spot region image sequence corresponding to the enhanced thermal image sequence.
4. The method for identifying fire hidden danger based on image enhancement according to claim 3, wherein the silo material surface is divided into according to the deployment azimuth sequence and the optimized pitch angle The method for the non-uniform grid area comprises the following steps: Obtaining internal parameters of the thermal infrared imager, wherein the internal parameters comprise focal length and pixel size, calculating the actual projection size of each pixel on the silo material surface on the image plane of the thermal infrared imager through inverse projection transformation according to the deployment azimuth angle, the optimized pitch angle and the internal parameters, determining a spatial resolution distribution diagram of the silo material surface according to the actual projection size of each pixel, dividing the region with the spatial resolution larger than a preset resolution threshold by adopting fine-grained grids according to the spatial resolution distribution diagram, and dividing the region with the spatial resolution not larger than the preset resolution threshold by adopting coarse-grained grids to obtain A non-uniform grid region.
5. The fire hidden danger identification method based on image enhancement according to claim 1, wherein the method for calculating the fractal dimension feature element of the contour according to the contour point set by adopting a box counting method comprises the following steps: The method comprises the steps of carrying out linear fitting on a contour point set according to a window with a preset length to obtain a direction variance and a fitting residual error of each window, marking the corresponding window as a regular linear segment when the direction variance is not greater than a preset variance threshold value and the fitting residual error is not greater than a preset residual error threshold value, removing the regular linear segment from the contour point set to obtain a purified contour point set, covering the purified contour point set by a series of boxes with preset sizes, carrying out statistics to obtain a coverage number sequence, and calculating the fractal dimension of the purified contour point set through linear fitting according to the coverage number sequence to serve as a contour fractal dimension feature element.
6. The method for identifying fire hidden danger based on image enhancement according to claim 1, wherein the method for obtaining the comprehensive confidence score by performing weighted fusion on the spatial feature vector and the time-sequence evolution feature vector comprises the following steps: sequentially splicing the space feature vector and the time sequence evolution feature vector to obtain a fusion feature vector according to a preset weight coefficient , , , Performing linear weighted calculation on the fusion feature vector to obtain a comprehensive confidence score 。
7. The method for identifying fire hidden danger based on image enhancement according to claim 1, wherein the time sequence dynamic evolution feature comprises a velocity distribution feature deduced from a smoldering physical diffusion model, comprising: Obtaining a boundary moving speed sequence of continuous frames, smoothing the boundary moving speed sequence to obtain a stable speed sequence, and when the stable speeds in the stable speed sequence are all in a preset theoretical speed interval And when the velocity distribution characteristics are met, judging the velocity distribution characteristics to be consistent.
8. Fire hidden danger identification system based on image enhancement, characterized in that the system comprises: The imaging module is used for deploying a thermal infrared imager array on the top of the inner side wall of the silo, and the thermal infrared imager array comprises The infrared thermal imaging system comprises an infrared thermal imaging system; the infrared thermal imager array is used for synchronously collecting the interior of the silo A thermal image sub-sequence for the said Carrying out space-time registration and splicing on the sub-sequences of the thermal images to obtain a panoramic thermal image sequence covering the surface of the whole silo material, carrying out image enhancement treatment on the panoramic thermal image sequence to obtain an enhanced thermal image sequence, extracting candidate thermal spot areas from the enhanced thermal image sequence to obtain a candidate thermal spot area image sequence; the characterization module is used for carrying out cross-frame tracking and matching on the same candidate hot spot region in the candidate hot spot region image sequence, calculating the spatial morphological characteristics and the time sequence dynamic evolution characteristics of the tracked candidate hot spot region according to the single-frame candidate hot spot region image, and carrying out weighted fusion on the spatial morphological characteristics and the time sequence dynamic evolution characteristics to obtain the comprehensive confidence score; The judging module is used for judging the candidate hot spot area as a smoldering hidden trouble when the comprehensive confidence score exceeds a preset confidence threshold and the time sequence dynamic evolution characteristic comprises a speed distribution characteristic deduced by a smoldering physical diffusion model; The method for calculating the spatial morphological characteristics and the time sequence dynamic evolution characteristics of the tracked candidate hot spot area according to the single-frame candidate hot spot area image comprises the following steps: obtaining gray gradient distribution according to the single frame of the candidate hot spot area image through Gaussian filtering and second derivative operation, and obtaining local gray change intensity characteristic elements according to the gray gradient distribution Obtaining a contour point set of the candidate hot spot region through a contour extraction algorithm according to the single-frame candidate hot spot region image, and obtaining a contour fractal dimension feature element through calculation by a box counting method according to the contour point set And the local gray level change intensity characteristic element And profile fractal dimension feature element Combining to obtain a space feature vector; calculating the area of the candidate hot spot area in each frame to obtain an area change sequence according to the candidate hot spot area image sequence, and calculating to obtain an equivalent radius change rate characteristic element according to the area change sequence Obtaining a boundary moving speed sequence according to the candidate hot spot area image sequence through cross-frame contour point matching and average displacement calculation, and obtaining a boundary expansion speed feature element by adopting median filtering processing according to the boundary moving speed sequence And the equivalent radius change rate feature element And boundary flare speed feature element Combining to obtain a time sequence evolution feature vector; The method for obtaining the boundary moving speed sequence through cross-frame contour point matching and average displacement calculation according to the candidate hot spot region image sequence comprises the following steps: acquiring a hot spot area contour point set of the previous frame from the candidate hot spot area image sequence Hot spot area contour point set corresponding to current frame And according to the outline point set of the hot spot area of the current frame Calculating an external normal unit vector set of each contour point of the current frame, and carrying out hot spot area contour point set on the current frame A set of hot spot area contour points for each point in the previous frame Performing local nearest neighbor search to obtain a candidate corresponding point set, screening the candidate corresponding point set according to a preset maximum allowable displacement constraint and a maximum direction deviation constraint to obtain an effective corresponding point set, calculating according to the effective corresponding point set and an external normal unit vector set to obtain a normal displacement set, calculating according to the normal displacement set to obtain a median, calculating according to the median and an inter-frame time interval to obtain an instantaneous boundary expansion speed The instantaneous boundary expansion speed is increased And carrying out median filtering on the speed time sequence to obtain a boundary moving speed sequence.

Description

Fire hidden danger identification method and system based on image enhancement Technical Field The invention relates to the technical field of fire monitoring, in particular to a method and a system for identifying hidden fire hazards based on image enhancement. Background In the storage and processing links of bulk materials such as coal, grain, wood dust and the like, silos are commonly used as storage containers. In the long-term accumulation process, the materials are affected by factors such as slow oxidation, fermentation and the like, and are easy to generate continuous temperature rise locally and enter a smoldering state. The smoldering masking performance is strong, and once the outside air enters, deflagration is extremely easy to be caused. The existing monitoring means mainly depend on two technologies, namely a point type sensor scheme, the problems of large monitoring blind area and alarm lag exist, and the smoldering starting point is difficult to capture in time. The second type is a thermal imaging technical scheme, although surface monitoring is realized, most of the existing methods only rely on a static temperature threshold of a single frame image for judgment. However, the internal environment of the silo is complex, various thermal interferences such as material waste heat, mechanical heat dissipation and the like exist, and the false alarm is extremely easy only by static temperature characteristics. The method cannot effectively distinguish the developed smoldering hot spot area from a stable static heat source in the early stage, so that the early-stage early-warning capability is insufficient. Therefore, a method capable of effectively identifying smoldering is needed to meet the early and reliable identification requirements of hidden danger of silo smoldering. Disclosure of Invention (1) Technical problem to be solved The invention aims to provide a fire hidden danger identification method and a fire hidden danger identification system based on image enhancement, so as to realize early, comprehensive and reliable identification of hidden danger in a silo. (2) Technical proposal In order to achieve the above object, in one aspect, the present invention provides a method for identifying fire hidden danger based on image enhancement, the method comprising: Step S1, disposing a thermal infrared imager array on the top of the inner side wall of the silo, wherein the thermal infrared imager array comprises The infrared thermal imaging system comprises an infrared thermal imaging system; the infrared thermal imager array is used for synchronously collecting the interior of the siloA thermal image sub-sequence for the saidThe method comprises the steps of carrying out space-time registration on each thermal image sub-sequence, splicing to obtain a panoramic thermal image sequence covering the surface of a whole silo material, carrying out image enhancement processing on the panoramic thermal image sequence to obtain an enhanced thermal image sequence, and extracting candidate thermal spot areas from the enhanced thermal image sequence to obtain a candidate thermal spot area image sequence. Step S2, performing cross-frame tracking and matching on the same candidate hot spot region in the candidate hot spot region image sequence, calculating the spatial morphological characteristics and the space-time dynamic evolution characteristics of the tracked candidate hot spot region according to the single-frame candidate hot spot region image, and performing weighted fusion on the spatial morphological characteristics and the space-time dynamic evolution characteristics to obtain the comprehensive confidence score. And step S3, when the comprehensive confidence score exceeds a preset confidence threshold and the space-time dynamic evolution feature comprises a speed distribution feature deduced by a smoldering physical diffusion model, judging the candidate hot spot area as a smoldering hidden trouble. Preferably, the method for deploying the thermal infrared imager array on the top of the inner side wall of the silo comprises the following steps: Obtaining structural parameters of a target silo, wherein the structural parameters comprise silo inner diameter and silo wall height, obtaining device parameters of infrared thermal imagers, wherein the device parameters comprise horizontal view angles and vertical view angles, and calculating according to the silo inner diameter and the horizontal view angles to obtain the minimum number of infrared thermal imagers required for realizing 360-degree surrounding coverage Calculating the number of the thermal infrared imagers at the minimum according to the inner diameter of the silo, the height of the silo wall and the vertical field angleThe desired initial pitch angle. Determining the number of thermal infrared imagers at the minimum according to the structural parameters and the device parametersIf the coverage blind area exists, the number of the