Workshop fire safety monitoring method, system, device and medium

Abstract

This application provides a method, system, equipment, and medium for monitoring fire safety in a workshop. The method acquires optical monitoring data for a specific area within the workshop, then obtains the optical signal intensities of at least two different spectral bands from this data. Based on the optical signal intensity of a first flame characteristic band and a second reference band, spectral characteristic data is determined. Candidate brightness regions are extracted from the optical monitoring data to obtain corresponding motion characteristic data. Finally, based on the spectral and motion characteristic data, flame suspicion is determined for the candidate brightness regions, resulting in a first flame suspicion monitoring result. This effectively distinguishes between normal process light events and abnormal flame events, reducing false alarm rates and minimizing manual shielding or desensitization operations caused by false alarms. Ultimately, this improves the overall reliability and practical application effectiveness of the workshop fire safety monitoring system under complex operating conditions.

Description

Technical Field

[0001] This application relates to safety monitoring technology, and more particularly to a method, system, equipment and medium for monitoring fire safety in a workshop. Background Technology

[0002] Existing workshop fire safety monitoring systems typically use flame optical detectors and / or video surveillance cameras to conduct optical monitoring of the workshop monitoring area. By analyzing the characteristics of the optical monitoring data, such as brightness thresholds, flicker frequency, and changes in area, it is possible to determine whether there is a flame or suspected fire event.

[0003] However, in typical industrial settings such as welding workshops, machining workshops, and metal component assembly workshops, many equipment casings, conveyor lines, tooling fixtures, metal pipes, and other components have high visible and infrared light reflectivity, and their complex surface shapes easily form strong reflective spots resulting from the superposition of specular and diffuse reflection. Furthermore, the bright reflective spots produced on these metal surfaces by strong light sources such as welding arcs, high-intensity work lights, and sunlight entering through doors and windows often exhibit brightness, flickering characteristics, and area variation characteristics highly similar to the optical characteristics of real flames.

[0004] Optical flame detection methods based solely on single-band brightness amplitude and simple time variation characteristics are insufficient to reliably distinguish real flames from interfering light sources such as welding arc light and metal reflection spots under complex metal reflection and strong process light interference conditions. This results in a high false alarm rate for optical flame monitoring in workshop environments, affecting the actual reliability and availability of fire monitoring systems. Summary of the Invention

[0005] This application provides a workshop fire safety monitoring method, system, equipment, and medium to solve the technical problem of false alarms caused by confusion between normal process light events and abnormal flame events in a workshop environment.

[0006] Firstly, this application provides a method for monitoring fire safety in a workshop, including: Acquire optical monitoring data for the monitored areas of the workshop; The optical signal intensity of at least two different spectral bands is obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band; Based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, the spectral ratio parameter and / or spectral combination characteristic parameter are determined to obtain the spectral characteristic data corresponding to the workshop monitoring area. Image analysis and / or spot analysis are performed on the optical monitoring data to extract candidate brightness regions, so as to obtain motion feature data of the morphology of the candidate brightness regions; Based on the spectral feature data and the motion feature data, flame suspicion is determined for the candidate brightness region, and a first flame suspicion monitoring result is obtained.

[0007] Secondly, this application provides a workshop fire safety monitoring system, comprising: The acquisition module is used to acquire optical monitoring data for the workshop monitoring area; The optical signal intensity of at least two different spectral bands is obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band; The determination module is used to determine the spectral ratio parameter and / or spectral combination characteristic parameter based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, so as to obtain the spectral characteristic data corresponding to the workshop monitoring area; The analysis module is used to perform image analysis and / or spot analysis on the optical monitoring data, extract candidate brightness regions, and obtain motion feature data of the morphology of the candidate brightness regions. The determination module is used to determine the flame suspicion of the candidate brightness region based on the spectral feature data and the motion feature data, and obtain the first flame suspicion monitoring result.

[0008] Thirdly, this application provides an electronic device, comprising: Processor; and, Memory for storing the executable instructions of the processor; The processor is configured to perform any of the possible methods described in the first aspect by executing the executable instructions.

[0009] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement any of the possible methods described in the first aspect.

[0010] The workshop fire safety monitoring method, system, equipment, and medium provided in this application acquire optical monitoring data for the workshop monitoring area, then obtain the optical signal intensity of at least two different spectral bands from the optical monitoring data. Based on the optical signal intensity of a first flame characteristic band and the optical signal intensity of a second reference band, spectral characteristic data is determined. Candidate brightness regions are extracted from the optical monitoring data to obtain corresponding motion characteristic data. Finally, based on the spectral characteristic data and motion characteristic data, flame suspicion is determined for the candidate brightness regions to obtain the first flame suspicion monitoring result. This effectively distinguishes between normal process strong light events and abnormal flame events, reducing false alarm rates and minimizing manual shielding or desensitization operations caused by false alarms. This improves the overall reliability and practical application effect of the workshop fire safety monitoring system under complex working conditions. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0012] Figure 1 This is a schematic flowchart illustrating a workshop fire safety monitoring method according to an example embodiment of this application; Figure 2 This is a flowchart illustrating a workshop fire safety monitoring method according to another example embodiment of this application; Figure 3 This is a schematic diagram of the structure of a workshop fire safety monitoring system according to an example embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0013] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0014] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0015] The numerous metal components in the workshop are typically made of polished steel plates, stainless steel, or coated metal materials, which have high reflectivity in the visible and infrared bands. In addition, the complex geometric structures such as bends, folds, and holes on the surface of the equipment cause external strong light sources to produce specular reflection, diffuse reflection, and multiple reflections through the metal surface, forming local bright spots or bright lines. Moreover, the brightness value of these reflected spots on the imaging or detection plane of the optical monitoring device can be comparable to or even stronger than that of a real flame.

[0016] In addition, the welding arc is a high-temperature plasma light source that outputs ultra-strong visible light and near-infrared radiation. Its brightness exhibits flicker characteristics within a certain frequency range depending on factors such as power supply fluctuations and changes in welding torch posture. In the time domain spectrum, it overlaps with the flicker frequency range of the flame.

[0017] However, existing flame optical detection methods mostly use threshold and frequency analysis of light intensity in a single channel or a few channels, and judge flames based solely on simple rules of brightening and flickering. They lack collaborative analysis of multi-band spectral distribution characteristics, regional geometric morphology changes, and non-optical quantities such as temperature and smoke. As a result, the system cannot distinguish between flame radiation generated by combustion and welding arc light and their reflection on the metal surface from the underlying mechanism, thus generating a large number of false alarms in complex optical interference scenarios in workshops.

[0018] To address the aforementioned issues, the embodiments provided in this application achieve technical differentiation through multi-channel optical detection and multi-feature fusion determination: On the one hand, by utilizing the physical mechanism that gases such as carbon dioxide produced during flame combustion have characteristic radiation peaks near the mid-infrared characteristic absorption band, the first channel of the multi-channel optical detection unit is configured to collect the infrared radiation intensity of the first flame characteristic band located near the carbon dioxide characteristic absorption band, and the second channel collects the radiation intensity of the near-infrared band and / or visible light band. By calculating the spectral ratio parameters and / or spectral combination characteristic parameters between different bands, the unique spectral energy distribution pattern of the flame is reflected, which can be distinguished from the differences in ratio and combination characteristics of light sources mainly formed by electric arcs or metal reflections.

[0019] On the other hand, based on the irregular, continuously rising, and strongly disturbed flow characteristics of flames driven by thermal buoyancy, frame-by-frame threshold segmentation, differential analysis, and connected component extraction are performed on the image sequence corresponding to the optical monitoring data to obtain the geometric centroid coordinates, area, and contour roughness of candidate brightness regions. Within a preset time window, the average brightness, area change, and centroid longitudinal displacement trend are tracked. Through time series features such as periodic fluctuations in area, high-frequency changes in contour shape, and overall upward migration of the centroid, the unstable combustion behavior of the flame in space is characterized, and it is distinguished from metal reflection spots and process light that are relatively stable in position and have slow contour changes.

[0020] Furthermore, by utilizing the thermal and flue gas diffusion mechanism that real fires are usually accompanied by a continuous rise in local ambient temperature and an increase in smoke concentration, while welding arcs or reflections often do not produce matching temperature rise and smoke characteristics, the baseline values ​​of ambient temperature and smoke concentration are extracted in the monitoring sub-region corresponding to the flame candidate area. The temperature change and smoke concentration change are calculated and compared with preset temperature rise thresholds and smoke rise thresholds to achieve cross-verification of physical quantities of optical flame suspect events by multiple sensors.

[0021] By combining the welding start and end linkage signals and the allowed time period information of welding operations obtained from the production control system, manufacturing execution system and welding equipment controller, flame-related suspected events verified by multiple sensors are classified and constrained according to working conditions. This systematically suppresses the problem of misjudgment of flame optical monitoring in complex workshop environments from the perspective of coupling the underlying physical mechanism and process state.

[0022] Figure 1 This is a schematic flowchart illustrating a workshop fire safety monitoring method according to an example embodiment of this application. Figure 1 As shown, the workshop fire safety monitoring method provided in this embodiment includes: S110. Acquire optical monitoring data for the workshop monitoring area.

[0023] In this step, optical monitoring data for the workshop monitoring area may be acquired. The optical monitoring data is acquired by an optical monitoring device deployed in the workshop monitoring area. The optical monitoring device includes at least a flame optical detector and / or a video surveillance camera.

[0024] Specifically, optical monitoring devices can be installed on the ceiling, side walls, and / or supporting structure of the workshop monitoring area according to the preset monitoring field coverage and installation spacing, so that the monitoring fields of adjacent optical monitoring devices overlap each other at least within the preset overlap ratio range, in order to avoid monitoring blind spots.

[0025] At the installation location of each optical monitoring device, adjust the installation height, installation tilt angle and / or azimuth angle of the optical monitoring device according to the equipment layout in the workshop and / or the distribution of flammable and explosive hazardous sources, so that the target monitoring sub-area in the workshop monitoring area is within the effective field of view of the optical monitoring device, and the optical axis direction of the optical monitoring device avoids large-area highly reflective metal surfaces and / or areas directly exposed to strong light.

[0026] When using a flame optical detector as an optical monitoring device, the detection sensitivity, sampling period and / or alarm threshold of the flame optical detector are configured, and the spatial range of the monitoring sub-area corresponding to each flame optical detector is determined by on-site calibration, so that the flame optical detector periodically outputs the multi-channel optical signal intensity of the corresponding monitoring sub-area as part of the optical monitoring data.

[0027] When using video surveillance cameras as optical monitoring devices, the image resolution, frame rate, exposure time, gain, and / or white balance parameters of the video surveillance cameras are configured, and automatic exposure mode and / or wide dynamic range mode are set according to the ambient lighting conditions on site, so that the video surveillance cameras periodically acquire image sequences of the workshop monitoring area as another part of the optical monitoring data.

[0028] Then, the multi-channel optical signal intensity data output by each flame optical detector and the image sequence data output by each video surveillance camera are sent to the centralized processing unit via wired and / or wireless networks, respectively, according to a preset data transmission protocol. The centralized processing unit timestamps and caches the received data to form optical monitoring data for subsequent spectral feature calculation and image analysis processing.

[0029] S120. Obtain the optical signal intensity of at least two different spectral bands from the optical monitoring data.

[0030] In this step, the optical signal intensity of at least two different spectral bands can be obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band.

[0031] It is worth noting that in workshop conditions such as welding, cutting, and heat treatment, non-fire-causing light sources such as welding sparks, electric arcs, high-temperature metal light emission, and strong reflective light spots often exhibit optical intensity characteristics similar to or even stronger than real flames in a single wavelength band. This makes flame monitoring methods based on single-wavelength intensity judgment prone to false alarms.

[0032] However, because gases such as carbon dioxide and water vapor produced during combustion exhibit significant radiation and absorption characteristics in the mid-infrared band, especially near the characteristic absorption band of carbon dioxide, the radiation distribution of a real flame in this band differs intrinsically from the thermal radiation of ordinary metals and visible welding arc light. Furthermore, the welding arc, molten metal pool, lamp light, and reflected highlights primarily exhibit strong radiation in the visible and near-infrared bands, while often lacking corresponding characteristic enhancement in the mid-infrared band near the characteristic absorption band of carbon dioxide.

[0033] Therefore, in this step, the infrared radiation intensity of the first flame characteristic band located near the carbon dioxide characteristic absorption band can be obtained through the first channel of the multi-channel optical detection unit, and the radiation intensity of the second reference band located in the near-infrared band and / or visible light band can be obtained through the second channel.

[0034] Optionally, the center wavelength of the first flame characteristic band is selected near the carbon dioxide characteristic absorption band of 4.2μm-4.6μm, and the spectral passband of the first channel is defined by a bandpass optical filter, so that the first channel has a selective response to the infrared radiation of the first flame characteristic band.

[0035] Optionally, the second reference band is set to the visible light and / or near-infrared band in the range of 0.4μm-1.1μm, and the second channel independently acquires the radiation intensity of the second reference band through corresponding optical filters and photodetectors.

[0036] Therefore, when the monitored target is a real flame, the intensity ratio between the mid-infrared radiation near the carbon dioxide characteristic absorption band and the visible / near-infrared reference band will exhibit an intensity ratio that matches the preset flame characteristics. However, when the monitored target is welding arc light, metal reflection, or ordinary lighting, the intensity ratio between the two bands will deviate significantly from the flame characteristic distribution. Thus, with the support of multi-band joint judgment, this step can reduce the false alarm rate caused by welding arc light, metal highlights, etc., while maintaining high sensitivity, thereby improving the accuracy of identifying suspected flame events in the workshop.

[0037] S130. Based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, determine the spectral characteristic data.

[0038] In this step, the spectral ratio parameter and / or spectral combination characteristic parameter can be determined based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, so as to obtain the spectral characteristic data corresponding to the workshop monitoring area.

[0039] Optionally, the intensity ratio of the optical signal intensity of the first flame characteristic band to the optical signal intensity of the second reference band can be determined as a spectral ratio parameter.

[0040] Specifically, the first channel output signal corresponding to the first flame characteristic band and the second channel output signal corresponding to the second reference band are acquired by the multi-channel optical detection unit, and the first channel output signal and the second channel output signal are converted from analog to digital to obtain the first channel digital intensity value and the second channel digital intensity value.

[0041] Then, dark current correction and / or background radiation baseline correction are performed on the digital intensity values ​​of the first and second channels to obtain the corrected optical signal intensity I1 of the first flame characteristic band and the corrected optical signal intensity I2 of the second reference band.

[0042] Within a preset time window, I1 and I2 are subjected to time moving average and / or median filtering respectively to obtain the filtered optical signal intensity Ĩ1 of the first flame characteristic band and the filtered optical signal intensity Ĩ2 of the second reference band.

[0043] Under the condition that Ĩ2 is greater than the preset lower limit threshold of reference intensity, the intensity ratio R of Ĩ1 to Ĩ2 is calculated as the spectral ratio parameter of the corresponding time window and the corresponding spatial monitoring unit.

[0044] When Ĩ2 is less than or equal to the preset lower limit threshold of reference intensity, the spectral ratio parameter is set to invalid and / or the spectral ratio calculation result at the current time is discarded, and the spectral ratio parameter is recalculated in the subsequent time window.

[0045] Optionally, the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band can be weighted and combined to obtain the spectral combination characteristic parameters.

[0046] Specifically, based on the pre-calibrated first flame characteristic band weighting coefficient w1 and second reference band weighting coefficient w2, Ĩ1 and Ĩ2 can be linearly weighted and combined to obtain the spectral combination characteristic parameters. The values ​​of w1 and w2 can be determined through calibration experiments using sample data from typical workshop environmental background radiation scenarios and actual flame radiation scenarios, and stored in the parameter configuration database of the workshop fire safety monitoring system for online retrieval.

[0047] For the calibration test mentioned above, an optical monitoring device of the same model and / or the same parameter configuration as the actual workshop monitoring area can be deployed in the simulated workshop monitoring area to construct an environmental background radiation scenario and / or an actual flame radiation scenario as calibration test scenarios. The environmental background radiation scenario includes a scenario where only workshop lighting sources and / or process equipment heat radiation exist without open flames, and the actual flame radiation scenario includes a scenario where typical combustibles are ignited under controlled and safe conditions to form a stable flame.

[0048] In an environmental background radiation scenario, multiple sets of optical monitoring data are collected within a preset time period. Based on the multiple sets of optical monitoring data, the corresponding corrected optical signal intensity of the first flame characteristic band and the corrected optical signal intensity of the second reference band are obtained. The optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band are then processed by time moving average and / or median filtering to obtain statistical characteristic parameters of the background scenario, including the background mean and standard deviation of the first flame characteristic band, the background mean and standard deviation of the second reference band.

[0049] In actual flame radiation scenarios, multiple sets of optical monitoring data are collected within a preset time period. Based on the multiple sets of optical monitoring data, the corresponding corrected first flame characteristic band optical signal intensity and the corrected second reference band optical signal intensity are obtained. The first flame characteristic band optical signal intensity and the second reference band optical signal intensity are then processed by time moving average and / or median filtering to obtain statistical characteristic parameters under the flame scenario, including the flame mean, flame standard deviation, second reference band flame mean, and flame standard deviation.

[0050] Then, based on the statistical feature parameters of the background scene and the flame scene, a target optimization function is constructed. This target optimization function is used to characterize the separability of the spectral combination feature parameters between the flame scene and the background scene. The separability includes the ratio of inter-class distance to intra-class variance and / or the discriminability index based on Fisher's discrimination criterion. Under the constraint that w1 and / or w2 are positive numbers, the target optimization function is numerically solved to obtain the optimal weight coefficients corresponding to the maximum value of the target optimization function. These are used as the weight coefficients w1 of the first flame feature band and w2 of the second reference band.

[0051] S140. Perform image analysis and / or spot analysis on the optical monitoring data to extract candidate brightness regions.

[0052] In this step, image analysis and / or spot analysis are performed on the optical monitoring data to extract candidate brightness regions. The centroid coordinates, area, contour morphology parameters, and time-varying parameters of the candidate brightness regions are obtained to acquire motion characteristic data of the candidate brightness region morphology.

[0053] Specifically, the image sequence corresponding to the optical monitoring data can be segmented frame by frame using a brightness threshold to obtain target brightness regions that meet a preset brightness threshold condition. Then, differential processing and / or connected component analysis are performed on adjacent frames to remove noise regions with an area lower than a preset area threshold, and the remaining brightness regions are retained as candidate brightness regions.

[0054] It is worth noting that under typical workshop process conditions such as welding, cutting, and grinding, a large number of welding slag splashes, grinding sparks, and metal processing sparks will generate numerous bright spots in the monitoring screen that are extremely bright but small in size and exist for a very short time. These bright spots are easily identified as brightness target areas based on frame-by-frame brightness threshold segmentation. As a result, in the subsequent extraction and analysis of candidate brightness areas, a large number of interference areas formed by short-term optical noise such as splashing sparks are inevitably introduced. This makes the existing workshop fire safety monitoring methods prone to problems such as a surge in the number of candidate targets, unstable motion feature analysis, and a high false positive rate of flame suspicion under complex working conditions.

[0055] It is important to understand that the above problems arise because during processes such as welding, cutting, and grinding, molten metal droplets and spark particles are ejected from the workpiece surface instantaneously at high speed. Their projected area in a single frame image is typically only a few to a dozen pixels, but the local brightness is extremely high, exceeding the brightness threshold multiple times. This results in numerous small, bright, connected regions in each frame. These connected regions often lack stable spatial continuity and area change continuity between adjacent frames, making it difficult to form reliably correlated regional trajectories. Consequently, time-series analysis must deal with a massive number of short-lived candidate regions whose centroid motion, area changes, and morphological evolution characteristics are highly discrete and random. This weakens the distinguishability of the relatively stable and expanding physical characteristics inherent in real flames in motion features, directly causing technical problems such as an excessive number of candidate brightness regions, unstable motion feature extraction, and an increased false positive rate for suspected flames.

[0056] To address this, spatial scale constraints can be applied to the brightness target region based on a preset first area threshold and / or a first shape threshold. Brightness target regions with an area smaller than the first area threshold and whose contour shape parameters represent a characteristic shape are marked as suspected noise regions. Brightness target regions not marked as suspected noise regions are then used as candidate brightness regions in the extraction process of region centroid coordinates, region area, and contour shape parameters. The aforementioned characteristic shape can be an elongated shape and / or a highly irregular shape. An elongated shape can be a rectangle with an aspect ratio greater than a preset ratio, and a highly irregular shape can be any shape outside the preset regular shape range, which can include rectangles, circles, ellipses, etc.

[0057] Understandably, when performing image analysis and / or spot analysis on the optical monitoring data of the workshop monitoring area, the image sequence corresponding to the optical monitoring data is first subjected to frame-by-frame brightness threshold segmentation to obtain the brightness target area that meets the preset brightness threshold conditions.

[0058] Subsequently, the area and contour morphology parameters of each brightness target region are determined, and spatial scale constraint processing is applied to the brightness target region based on the preset first area threshold and / or first morphology threshold. Brightness target regions with an area smaller than the first area threshold and whose contour morphology parameters are characterized as elongated and / or highly irregular are marked as suspected noise regions. Only brightness target regions that are not marked as suspected noise regions are used as candidate brightness regions in the extraction process of region centroid coordinates, region area and contour morphology parameters.

[0059] Based on this, a region identifier is assigned to each candidate brightness region detected in a single frame image, and region association matching is performed on the candidate brightness regions between adjacent frames based on the spatial proximity of the region centroid coordinates and / or the region area change constraint to form a continuous region trajectory.

[0060] For each continuous region trajectory, the number of valid survival frames and the number of recurrences within a preset time window are counted to form time series features such as region survival time and region recurrence count. When the region survival time is lower than the preset minimum survival time threshold and / or the region recurrence count is lower than the preset occurrence count threshold, the corresponding candidate brightness region is marked as a short-term noise region and excluded from the flame candidate region set.

[0061] By combining the above-mentioned spatial scale constraints, temporal persistence constraints, and trajectory correlation constraints, it is possible to distinguish between small-area high-speed spatter bright spots that appear only in a single frame or a very few frames and real flame areas that persist for a certain period of time and have certain spatial expansion or small-range jitter characteristics. This suppresses the interference of short-term optical noise generated by welding slag spatter, grinding sparks, etc. on the extraction of motion features of candidate brightness areas and the determination of flame suspicion.

[0062] Furthermore, the determination of motion feature data for the morphology of candidate brightness regions specifically includes: Determine the geometric centroid coordinates of the candidate brightness region in the image plane, and use them as the region centroid coordinates.

[0063] The number of pixels occupied by the candidate brightness region in a single frame image is determined as the region area.

[0064] Based on the perimeter, area, and edge roughness of the candidate brightness region, the contour morphology parameters are determined. The contour morphology parameters include the ratio of the square of the perimeter to the area and / or the edge gradient change index.

[0065] Then, time series analysis is performed on the average brightness, area, contour morphology parameters, and centroid coordinates of the candidate brightness region within a preset time window to obtain the brightness change amplitude, brightness flicker frequency, contour change rate, and centroid longitudinal displacement trend, which serve as parameters for the change of region brightness over time. Motion feature data includes parameters for the change of region brightness over time.

[0066] The aforementioned scenarios involving welding slag spatter, grinding sparks, and metalworking sparks exhibit physical characteristics of high initial velocity, short lifespan, small particle size, and strong reflection. In image sequences, this manifests as a large spatial displacement of the centroid between adjacent frames, rapid decay of the affected area over time (even disappearing within a very short period), and high brightness peaks with extremely short durations. However, real flames are affected by fuel supply and air turbulence, typically persisting within a certain area. The centroid of the affected area only fluctuates randomly within a small range, and the area generally shows a stable or slowly fluctuating growth trend over time. Furthermore, due to combustion turbulence, the flame outline gradually becomes more complex, with enhanced edge undulations.

[0067] Therefore, in order to distinguish between the two scenarios mentioned above, the time series analysis may further include: Each candidate brightness region detected in a single frame image is assigned a region identifier. Based on the spatial proximity of the region's centroid coordinates and / or the region's area change constraints in adjacent frames, region association matching is performed on the candidate brightness regions to obtain corresponding continuous region trajectories. For each continuous region trajectory, the number of valid frames within a preset time window is counted as the region's duration, and the number of times it appears within the preset time window is counted as the region's repetition count. When the region's duration is lower than a preset minimum duration threshold and / or the region's repetition count is lower than a preset repetition count threshold, the corresponding candidate brightness region is marked as a short-term noise region and is not included in the generation of the first flame suspicion monitoring result during flame suspicion determination.

[0068] Furthermore, for candidate brightness regions selected by region duration and number of recurrences, the region centroid coordinate trajectory curve and the region area change curve over time are constructed within a preset time window. Based on the region centroid coordinate trajectory curve, the centroid motion speed, centroid motion trajectory smoothness, and centroid longitudinal displacement trend are determined. Based on the region area change curve over time, the region area growth rate is determined.

[0069] When the displacement of the centroid's motion velocity between frames is greater than the preset high-speed motion threshold and the corresponding area growth rate is negative and its absolute value is greater than the preset area growth threshold, the corresponding selected brightness area is determined to be a splash-type optical anomaly event and is removed from the flame suspicion determination.

[0070] When the displacement of the centroid motion velocity between single frames is less than the preset small-range jitter threshold, the area growth rate is positive and greater than the preset area growth threshold, and the change amplitude of the contour morphology parameters within the preset time window is greater than the preset contour complexity growth threshold, the corresponding motion feature data of the corresponding candidate brightness region will be used as the motion feature data of the flame candidate region morphology with higher priority to participate in the determination of the first flame suspect monitoring result.

[0071] It is understandable that after extracting candidate brightness regions in a single frame, each candidate brightness region detected in a single frame image is assigned a unique region identifier. Between adjacent frames, using the spatial distance of the region centroid coordinates and / or the magnitude of the region area change as constraints, nearest neighbor matching or threshold matching strategies are employed to perform region association matching on the candidate brightness regions, constructing continuous region trajectories, and tracking the evolution of each region trajectory within a preset time window.

[0072] Then, the number of valid frames for each continuous region trajectory within the time window is taken as the region's duration, and the number of frames in which the region's identifier appears is taken as the region's recurrence count. Candidate regions whose duration and recurrence count are lower than their respective thresholds are directly marked as short-term noise regions and eliminated.

[0073] For the candidate brightness regions that pass the above persistence screening, their centroid coordinate trajectory curve and region area change curve are further constructed on the time axis: The centroid displacement of a single frame is obtained by differentiating the centroid coordinates between consecutive frames, and the centroid motion velocity is calculated by combining it with the image space calibration parameters. The smoothness of the centroid's trajectory is evaluated by statistically analyzing the curvature and direction change frequency of the centroid's motion path, and the longitudinal displacement trend of the centroid is extracted by performing monotonicity and trend analysis on the longitudinal coordinate components.

[0074] Simultaneously, the time-varying sequence of region area is differentially analyzed and fitted to obtain the region area growth rate and its variation pattern. Based on these time-series characteristics, when a candidate region is detected to have a centroid displacement greater than a preset high-speed motion threshold in adjacent frames, and the region area growth rate is negative and its absolute value is greater than a preset decay threshold, it can be classified as a high-speed splashing and area decaying splash-type optical anomaly event and removed from the flame candidate set.

[0075] Conversely, when the centroid displacement of a candidate region is always less than the small-range jitter threshold, the region's area growth rate is positive and exceeds the area growth threshold, and its contour morphology parameters, such as the perimeter-to-area ratio and the boundary fractal dimension, show a significant increasing trend within a preset time window, indicating that the region has the characteristics of stable combustion and gradual expansion with increasingly complex boundaries, the region is identified as a higher-priority real flame candidate region, and its motion feature data is directly used to generate the first flame suspect monitoring result.

[0076] By combining regional association matching, duration statistics, kinematic feature extraction, and area / contour evolution feature analysis, we can effectively eliminate splash-type optical anomalies and prioritize the identification of real flame areas.

[0077] Specifically, within a preset time window, the average brightness, area, contour morphology parameters, and centroid coordinates of the same candidate brightness region in consecutive image frames can be indexed in chronological order to form corresponding time series of average brightness, area, contour morphology parameters, and centroid longitudinal coordinates.

[0078] The mean brightness time series is detrended and / or bandpass filtered, and the difference between the maximum and minimum values ​​and / or standard deviation of the mean brightness within a preset time window are calculated based on the processed mean brightness time series as the brightness variation amplitude; frequency domain analysis is performed on the mean brightness time series, including fast Fourier transform and / or autocorrelation analysis, to determine the dominant frequency of brightness fluctuation and / or the energy proportion in the frequency band of the dominant frequency, as the brightness flicker frequency.

[0079] The contour morphology parameter time series is differentially calculated between adjacent frames and / or linearly fitted and / or polynomial fitted within a preset time window. Based on the average absolute value of the difference between the contour morphology parameters at adjacent times and / or the slope of the fitted curve, the average rate of change and / or peak rate of change of the contour morphology parameter within the preset time window are determined as the contour change rate.

[0080] Monotonicity detection and / or trend fitting are performed on the centroid longitudinal coordinate time series within a preset time window. Trend fitting includes least-squares linear fitting of the centroid longitudinal coordinate time series to obtain the longitudinal displacement fitting slope and / or calculating the centroid longitudinal coordinate difference between the start and end times. Based on the sign of the fitting slope, the magnitude of the slope, and / or the sign and absolute value of the longitudinal coordinate difference, it is determined whether the centroid longitudinal displacement trend is generally upward within the preset time window, as well as the distance and / or speed of the upward movement, which serve as quantitative indicators of the centroid longitudinal displacement trend.

[0081] Then, the brightness change amplitude, brightness flicker frequency, contour change rate, and centroid longitudinal displacement trend are used together as a set of regional brightness change parameters to characterize the brightness change characteristics of candidate brightness regions over time. These regional brightness change parameters are then used as part of the motion feature data for subsequent flame suspicion determination.

[0082] The technical principle behind the above steps is worth understanding. Because flames exhibit highly time-varying, irregular shapes and obvious edge fluctuations in the image plane, they are fundamentally different from interfering light sources such as welding arcs, metal reflections, and car headlight spots in terms of spatial distribution, temporal variation patterns and contour roughness. However, if only brightness is used for quantification, it is impossible to extract the unique dynamic behavioral characteristics of flames, such as brightness flicker frequency, contour deformation rate and overall upward drift.

[0083] Therefore, through the above steps, on the one hand, the contour roughness and area jitter characteristics can be used to effectively distinguish between real flames with stable contour boundaries and regular shapes, such as metallic reflections and light spots, and those with significant edge disturbances. On the other hand, by utilizing the periodic changes in brightness over time and the overall upward displacement trend of the centroid, static interference sources with basically constant brightness and fixed positions can be suppressed, thereby significantly reducing the false alarm rate and improving the reliability of flame suspicion detection results, effectively solving the problem of insufficient robustness of flame identification under complex working conditions in existing technologies.

[0084] S150. Based on spectral feature data and motion feature data, flame suspicion is determined for candidate brightness regions to obtain the first flame suspicion monitoring result.

[0085] When the spectral feature data meets the preset flame spectral threshold condition, and the brightness region morphological motion feature data meets at least one of the following conditions, the corresponding candidate brightness region is determined as a flame candidate region: The area of ​​the region exhibits periodic changes exceeding a preset area change threshold within a preset time window; The change in the contour morphology parameters within the preset time window is greater than the preset contour change threshold. The longitudinal displacement trend of the centroid is generally upward within the preset time window and the displacement distance exceeds the preset displacement threshold. To obtain the first suspected flame monitoring result corresponding to the flame candidate area.

[0086] It is worth understanding that in this step, firstly, the optical signal intensity of different spectral bands is obtained by using multi-channel optical detection or multispectral imaging, and spectral feature data is constructed based on the intensity ratio and / or weighted combination of the first flame characteristic band and the second reference band. By setting flame spectral threshold conditions, the brightness target area that deviates significantly from the flame characteristic distribution in the spectrum is eliminated in the early stage of decision-making, and the first layer of screening is carried out from the perspective of physical radiation mechanism.

[0087] Secondly, for the candidate brightness regions selected by spectral screening, the region area, contour morphology parameters, and the change curve of the longitudinal coordinate of the region centroid over time are calculated based on the region pixel set in the image sequence. Through time series analysis, motion features such as the periodic change of area, the change amplitude of contour morphology parameters, and the longitudinal displacement trend of the centroid are extracted. The overall upward behavior of random expansion and contraction of area, boundary disturbance, and buoyancy driven by airflow turbulence during flame combustion is transformed into measurable time series indicators.

[0088] Finally, in terms of judgment logic, a joint decision strategy is adopted, in which the spectral features meet the flame spectral threshold condition and the motion features meet at least one of the dynamic constraints of area expansion and contraction, contour jitter or upward drift. This is equivalent to cross-validating the target on two independent but complementary physical dimensions: spectral characteristics and spatiotemporal evolution behavior. This ensures that the flame suspicion result is output only when both the flame radiation mechanism and combustion dynamics behavior mode are met, thereby improving the reliability of the judgment.

[0089] In this embodiment, optical monitoring data for the workshop monitoring area is acquired, and then optical signal intensities of at least two different spectral bands are obtained from the optical monitoring data. Then, based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, spectral feature data is determined, and candidate brightness regions are extracted from the optical monitoring data to obtain corresponding motion feature data. Finally, based on the spectral feature data and motion feature data, flame suspicion is determined for the candidate brightness regions to obtain the first flame suspicion monitoring result. This effectively distinguishes between normal process strong light events and abnormal flame events, thereby reducing the false alarm rate and minimizing manual shielding or desensitization operations caused by false alarms. This improves the overall reliability and practical application effect of the workshop fire safety monitoring system under complex working conditions.

[0090] It is worth noting that, in Figure 1 In the illustrated embodiment, when relying solely on optical monitoring data for flame suspicion determination, even with algorithm optimization, a small probability of false alarms may still occur. Therefore, to further improve the accuracy of flame suspicion determination in workshop fire scenarios, in Figure 1 Based on the illustrated embodiment, Figure 2 This is a flowchart illustrating a workshop fire safety monitoring method according to another example embodiment of this application. Figure 2 As shown, the workshop fire safety monitoring method provided in this embodiment, in Figure 1 Based on the illustrated embodiment, it may further include: S210. Based on the temperature monitoring data and smoke monitoring data of the workshop monitoring area, combined with the temperature change and smoke concentration change within the preset time window, the first suspected flame monitoring result is jointly verified by multiple sensors to obtain the multi-sensor verification result.

[0091] Optionally, the ambient temperature baseline and smoke concentration baseline can be acquired within a monitoring sub-region corresponding to the flame candidate area. Then, within a preset time window, the temperature change between the actual temperature value and the ambient temperature baseline, and the concentration change between the actual smoke concentration value and the smoke concentration baseline, can be determined.

[0092] When the temperature change exceeds the preset temperature rise threshold and the concentration change exceeds the preset smoke rise threshold, the first suspected flame monitoring result is confirmed as a suspected flame event verified by multiple sensors.

[0093] If the judgment conditions are not met, the first flame suspect monitoring result will be confirmed as an optical anomaly event that has failed the multi-sensor joint verification.

[0094] Furthermore, the acquisition of the aforementioned ambient temperature baseline and smoke concentration baseline can be achieved by averaging the temperature monitoring data and smoke monitoring data of the target monitoring sub-region over a sliding time window during the period when no flame candidate region is detected, and then periodically updating these values.

[0095] It is worth understanding that in the above steps, the workshop monitoring area is first divided into multiple monitoring sub-areas, and the field of view of each optical monitoring device is spatially mapped with the monitoring coverage of the corresponding temperature sensor and smoke sensor to establish a one-to-one or one-to-many correspondence between the flame candidate area and the monitoring sub-area.

[0096] Secondly, during the time period when no flame candidate area is detected, the temperature monitoring data and smoke monitoring data of each monitoring sub-area are averaged and / or filtered by a sliding time window to obtain the ambient temperature baseline value and smoke concentration baseline value, and are updated at a preset period to adaptively compensate for long-term drift caused by factors such as seasonal changes, ventilation condition adjustments, and start-up and shutdown of high-temperature equipment.

[0097] Based on this, when the first suspected flame monitoring result is obtained based on spectral feature data and motion feature data, the system starts to calculate the temperature change and smoke concentration change within a preset time window in the monitoring sub-region corresponding to the flame candidate region. The temperature change is obtained by subtracting the current actual temperature value from the corresponding ambient temperature baseline value, and the concentration change is obtained by subtracting the actual smoke concentration value from the corresponding smoke concentration baseline value. By setting temperature rise threshold and smoke rise threshold, the significant heat release and smoke production characteristics that are inevitably accompanied by the physical combustion process are transformed into executable quantitative logical criteria.

[0098] When the temperature change and the smoke concentration change both exceed their respective thresholds, the suspected flame event is identified as having combustion characteristics in both the thermal field and the smoke field, thus achieving secondary confirmation of the physical mechanism of the optical flame suspected event.

[0099] This unifies heterogeneous data from multiple sensors into a single judgment framework in terms of time, space, and physical dimensions, thereby enabling reliable constraints on suspected flame events through joint verification of multiple sensors in a workshop setting without the need to introduce complex models.

[0100] S220. Based on the multi-sensor verification results and the process linkage information related to the operating status of workshop process equipment and / or the process operation time period, perform working condition constraint judgment on the multi-sensor verification results.

[0101] In this step, the process equipment operating status information and / or process operation plan information corresponding to the flame candidate area may be obtained. The operating status information includes at least the start and stop status of the welding equipment, and the process operation plan information includes at least the allowed time period for welding operations.

[0102] Optionally, the system may receive process linkage signals representing welding start and welding end commands from the production control system, manufacturing execution system, and / or welding equipment controller, and determine the operating status of the welding equipment at the corresponding workstation based on the process linkage signals.

[0103] When the welding equipment at the corresponding workstation is in operation and / or within the permitted welding operation time period, suspected flame events from the flame candidate area that have been jointly verified by multiple sensors are judged as low-level fire events constrained by process interference, and the corresponding alarm trigger threshold is raised and / or the alarm duration judgment time is shortened.

[0104] When the welding equipment at the corresponding workstation is stopped and / or not within the permitted welding operation time period, suspected flame events from the flame candidate area that have been jointly verified by multiple sensors are identified as high-priority suspected fire events, and the corresponding alarm trigger threshold is lowered and / or the alarm duration is extended.

[0105] It is worth understanding that in the above steps, firstly, the candidate flame areas obtained by optical monitoring are mapped and associated with specific workstations in the workshop in space, so that each suspected flame event can be associated with a specific process workstation.

[0106] Secondly, in the time dimension, the operating status information of the process equipment corresponding to the workstation and the pre-configured process operation plan information are obtained from the production control system, manufacturing execution system and / or welding equipment controller, and these operating condition information are transformed into discrete operating condition variables of process allowable state / process prohibitive state.

[0107] Furthermore, the aforementioned operating condition variables and multi-sensor verification results are combined to form a joint state space: when the state is "process permitted" and "flame suspicion established," a pre-set first-type alarm strategy parameter set is invoked. By increasing the trigger threshold required for fire alarms and / or shortening the duration of judgment, this type of event is restricted to the low-level fire alarm range. Conversely, when the state is "process prohibited" and "flame suspicion established," a pre-set second-type alarm strategy parameter set is invoked. By lowering the fire alarm trigger threshold and / or extending the duration of judgment, the system acknowledges and alarms this type of event with high priority.

[0108] By logically coupling the process operation status, operation time window and multi-sensor flame suspicion results, and driving the alarm threshold and judgment duration to adaptively adjust in a parameterized manner, the coupling between the fire monitoring algorithm layer and the production process control layer is realized. This enables the final fire alarm result to dynamically reflect the real production conditions, thereby technically ensuring the unity of false alarm suppression and early reliable alarm effects.

[0109] S230, Generate the final fire alarm results for the workshop monitoring area.

[0110] When the multi-sensor verification results indicate that the event does not constitute a suspected fire event, a non-fire event marker is generated, a fire alarm is not triggered, and only a log is recorded or a prompt message is generated.

[0111] When the multi-sensor verification results indicate a suspected fire event and the operating conditions determine it to be a target priority suspected fire event, a fire alarm signal is generated and output to the workshop fire alarm device and / or linkage control device.

[0112] When the multi-sensor verification results indicate a suspected flame event and the operating condition constraints determine it to be a low-level fire event constrained by process interference, a low-level alarm signal is generated for manual confirmation or as a basis for trend monitoring.

[0113] Figure 3 This is a schematic diagram of a workshop fire safety monitoring system according to an example embodiment of this application. Figure 3As shown, the workshop fire safety monitoring system 300 provided in this embodiment includes: The acquisition module 310 is used to acquire optical monitoring data for the workshop monitoring area; The optical signal intensity of at least two different spectral bands is obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band; The determination module 320 is used to determine the spectral ratio parameter and / or spectral combination characteristic parameter based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, so as to obtain the spectral characteristic data corresponding to the workshop monitoring area; The analysis module 330 is used to perform image analysis and / or spot analysis on the optical monitoring data, extract candidate brightness regions, and obtain motion feature data of the morphology of the candidate brightness regions. The determination module 340 is used to determine the flame suspicion of the candidate brightness region based on the spectral feature data and the motion feature data, and obtain the first flame suspicion monitoring result.

[0114] Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 4 As shown, the electronic device 400 provided in this embodiment includes: a processor 401 and a memory 402; wherein: Memory 402 is used to store computer programs, and the memory may also be flash memory.

[0115] Processor 401 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0116] Alternatively, the memory 402 can be either standalone or integrated with the processor 401.

[0117] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include: Bus 403 is used to connect the memory 402 and the processor 401.

[0118] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0119] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0120] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0121] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for monitoring fire safety in a workshop, characterized in that, include: Acquire optical monitoring data for the monitored areas of the workshop; The optical signal intensity of at least two different spectral bands is obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band; Based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, the spectral ratio parameter and / or spectral combination characteristic parameter are determined to obtain the spectral characteristic data corresponding to the workshop monitoring area. Image analysis and / or spot analysis are performed on the optical monitoring data to extract candidate brightness regions, so as to obtain motion feature data of the morphology of the candidate brightness regions; Based on the spectral feature data and the motion feature data, flame suspicion is determined for the candidate brightness region, and a first flame suspicion monitoring result is obtained.

2. The workshop fire safety monitoring method according to claim 1, characterized in that, The step of acquiring optical signal intensities from at least two different spectral bands from the optical monitoring data includes: The infrared radiation intensity of the first flame characteristic band located near the carbon dioxide characteristic absorption band is obtained through the first channel of the multi-channel optical detection unit, and the radiation intensity of the second reference band located in the near-infrared band and / or visible light band is obtained through the second channel.

3. The workshop fire safety monitoring method according to claim 1, characterized in that, The determination of the spectral ratio parameter and / or spectral combination characteristic parameter based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band includes: The intensity ratio of the optical signal intensity of the first flame characteristic band to the optical signal intensity of the second reference band is determined as a spectral ratio parameter. And / or, the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band are weighted and combined to obtain the spectral combination characteristic parameters.

4. The workshop fire safety monitoring method according to claim 1, characterized in that, The image analysis and / or spot analysis of the optical monitoring data includes: The image sequence corresponding to the optical monitoring data is segmented frame by frame using a brightness threshold to obtain a brightness target region that meets a preset brightness threshold condition; Differential processing and / or connected component analysis are performed on adjacent frames to remove noise regions with an area lower than a preset area threshold, and the remaining brightness regions are retained as the candidate brightness regions.

5. The workshop fire safety monitoring method according to claim 1, characterized in that, The motion feature data for obtaining the morphology of the candidate brightness region includes: Determine the geometric centroid coordinates of the candidate brightness region in the image plane, and use them as the region centroid coordinates; The number of pixels occupied by the candidate brightness region in a single frame image is determined as the region area; Based on the perimeter, area, and edge roughness of the candidate brightness region, contour morphology parameters are determined, including the ratio of the square of the perimeter to the area and / or the edge gradient change index. Time series analysis is performed on the average brightness of the candidate brightness region within a preset time window, the region area, the contour morphology parameters, and the centroid coordinates of the region to obtain the brightness change amplitude, brightness flicker frequency, contour change rate, and centroid longitudinal displacement trend, which are used as parameters for the change of region brightness over time. The motion feature data includes the parameters for the change of region brightness over time.

6. The workshop fire safety monitoring method according to claim 5, characterized in that, The step of determining flame suspicion in the candidate brightness region based on the spectral feature data and the motion feature data to obtain a first flame suspicion monitoring result includes: When the spectral feature data meets the preset flame spectral threshold condition, and the motion feature data of the candidate brightness region morphology meets at least one of the following, the corresponding candidate brightness region is determined as a flame candidate region: The area of ​​the region exhibits periodic changes exceeding a preset area change threshold within a preset time window; The change in the contour morphology parameter within the preset time window is greater than the preset contour change threshold. The longitudinal displacement trend of the centroid is upward within the preset time window and the displacement distance exceeds the preset displacement threshold. To obtain the first flame suspicion monitoring result corresponding to the flame candidate region.

7. The workshop fire safety monitoring method according to claim 1, characterized in that, After obtaining the first suspected flame detection result, the following is also included: Based on the temperature monitoring data and smoke monitoring data of the workshop monitoring area, combined with the temperature change and smoke concentration change within the preset time window, the first flame suspect monitoring result is jointly verified by multiple sensors to obtain the multi-sensor verification result. Based on the multi-sensor verification results and the process linkage information related to the operating status of workshop process equipment and / or the process operation time period, the multi-sensor verification results are used to determine the operating condition constraints and generate the final fire alarm result for the workshop monitoring area.

8. The workshop fire safety monitoring method according to claim 7, characterized in that, The process involves using temperature and smoke monitoring data from the workshop monitoring area, combined with temperature and smoke concentration changes within a preset time window, to perform multi-sensor joint verification on the first flame suspicion monitoring result, yielding multi-sensor verification results, including: Within the monitoring sub-region corresponding to the flame candidate region, the baseline values ​​of ambient temperature and smoke concentration are acquired. Within a preset time window, determine the amount of temperature change between the actual temperature value and the ambient temperature baseline value, and the amount of concentration change between the actual smoke concentration value and the smoke concentration baseline value; When the temperature change exceeds a preset temperature rise threshold and the concentration change exceeds a preset smoke rise threshold, the first flame suspicion monitoring result is confirmed as a flame suspicion event verified by multiple sensors. If the judgment condition is not met, the first flame suspicion monitoring result is confirmed as an optical anomaly event that has failed the multi-sensor joint verification.

9. The workshop fire safety monitoring method according to claim 8, characterized in that, The acquisition of the ambient temperature baseline value and the smoke concentration baseline value includes: During the time period when no flame candidate region is detected, the temperature monitoring data and smoke monitoring data of the target monitoring sub-region are averaged over a sliding time window to obtain the corresponding temperature baseline value and smoke concentration baseline value, and are periodically updated.

10. The workshop fire safety monitoring method according to claim 7, characterized in that, The step of performing condition constraint judgment on the multi-sensor verification results based on the multi-sensor verification results and process linkage information related to the operating status of workshop process equipment and / or process operation time periods includes: Obtain the process equipment operation status information and / or process operation plan information corresponding to the flame candidate area. The operation status information includes at least the start and stop status of the welding equipment, and the process operation plan information includes at least the allowed time period for welding operations. When the welding equipment at the corresponding workstation is in operation and / or within the permitted welding operation time period, the suspected flame event from the flame candidate area, verified by multiple sensors, is determined to be a low-level fire event constrained by process interference, and the corresponding alarm trigger threshold is raised and / or the alarm duration determination time is shortened. When the welding equipment at the corresponding workstation is stopped and / or not within the permitted welding operation time period, the suspected flame events from the flame candidate area that have been jointly verified by multiple sensors are judged as high-priority suspected fire events, and the corresponding alarm trigger threshold is lowered and / or the alarm duration judgment time is extended.

11. The workshop fire safety monitoring method according to claim 10, characterized in that, The acquisition of the process equipment operating status information corresponding to the flame candidate region includes: Receive process linkage signals representing welding start and welding end commands from the production control system, manufacturing execution system and / or welding equipment controller, and determine the operating status of the welding equipment at the corresponding workstation based on the process linkage signals.

12. The workshop fire safety monitoring method according to claim 7, characterized in that, The generation of the final fire alarm result for the workshop monitoring area includes: When the multi-sensor verification result indicates that there is no suspected fire event, a non-fire event marker is generated, the fire alarm is not triggered, and only a log is recorded or a prompt message is generated. When the multi-sensor verification results indicate that a suspected fire event has been constituted and the event is determined to be a target priority fire event by the operating condition constraints, a fire alarm signal is generated and output to the workshop fire alarm device and / or linkage control device. When the multi-sensor verification results indicate a suspected flame event and the operating condition constraints determine it to be a low-level fire event constrained by process interference, a low-level alarm signal is generated for manual confirmation or as a basis for trend monitoring.

13. A workshop fire safety monitoring system, characterized in that, include: The acquisition module is used to acquire optical monitoring data for the workshop monitoring area; The optical signal intensity of at least two different spectral bands is obtained from the optical monitoring data, wherein one spectral band is the first flame characteristic band and the other spectral band is the second reference band; The determination module is used to determine the spectral ratio parameter and / or spectral combination characteristic parameter based on the optical signal intensity of the first flame characteristic band and the optical signal intensity of the second reference band, so as to obtain the spectral characteristic data corresponding to the workshop monitoring area; The analysis module is used to perform image analysis and / or spot analysis on the optical monitoring data, extract candidate brightness regions, and obtain motion feature data of the morphology of the candidate brightness regions. The determination module is used to determine the flame suspicion of the candidate brightness region based on the spectral feature data and the motion feature data, and obtain the first flame suspicion monitoring result.

14. An electronic device, characterized in that, include: processor; as well as, Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1 to 12 by executing the executable instructions.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 12.

Images