Smoke alarm method based on smoke particle motion trail

Abstract

The invention relates to the technical field of fire detection, in particular to a smog alarm method based on a smog particle motion trail, which comprises the following steps: constructing a three-dimensional flow field distribution diagram of smog particles, and aiming at the sectional area change and curvature change of a current channel at a branch path, taking the predicted smog particle position as a center to obtain a smog particle motion trail distribution diagram; associating the flow velocity data of each branch path to obtain a flow velocity gradient sequence of each branch path; on the basis of the flow velocity gradient sequence of each branch path, when the flow velocity gradient exceeds a flow velocity threshold value, the smoke particle position in each branch path is updated; performing trajectory fitting on the smoke particles, and configuring concentration intensity indexes and updating time of the smoke particles in different branch paths according to concentration gradients corresponding to the motion trajectory sequences on the branch paths; and configuring a classification label of each branch path, and carrying out graded early warning according to flow velocity differentiation. And the accuracy and timeliness of smoke alarm are improved.

Description

Technical Field

[0001] This invention relates to the field of fire detection technology, specifically a smoke alarm method based on the trajectory of smoke particles. Background Technology

[0002] Existing smoke sensors typically use single-point sensors to collect concentration or flow rate data and identify the distribution of smoke based on optical detection. They do not construct a global three-dimensional flow field model that matches the channel structure, resulting in the collected smoke concentration being mostly local discrete data. This makes it difficult to reflect the global flow distribution under smoke diversion and concentration dilution. After a fire occurs, it is impossible to complete the smoke alarm processing based on the distribution of smoke flow.

[0003] For example, Chinese Patent Publication No. CN120126271A discloses a maze optimization design for a smoke sensor and a method for resisting dust pollution. The method includes: detecting fused smoke and dust particles and water vapor particles in the air detection area using a signal transmitting device; filtering out water vapor particles to reduce false smoke alarms caused by them; using differential signal processing technology to distinguish and identify fused particles; introducing the fused particles into the maze measurement cavity; measuring smoke and dust concentrations within the maze measurement cavity to enhance detection stability; automatically adjusting the width of the maze channel within the maze measurement cavity according to the smoke concentration to spatially segment smoke and dust particles, increasing the detector's adaptability to complex environments; judging fire early warning based on the smoke concentration within the maze measurement cavity; and activating a cleaning device added to the maze measurement cavity for cleaning based on the smoke and dust concentrations.

[0004] For example, Chinese Patent Publication No. CN117789394A discloses an early fire smoke detection method based on motion history images, belonging to the field of fire monitoring and alarm technology. To address the deficiencies in accuracy, reliability, and real-time performance of fire early warning systems, this method improves the accuracy of smoke-based fire prediction through smoke feature extraction and smoke motion analysis. By analyzing motion patterns and smoke characteristics, it can accurately identify fire smoke. The real-time performance of the early warning is improved by acquiring image sequences; continuous acquisition and real-time analysis of image sequences enables timely detection of fire smoke with high reliability. Analysis of historical smoke motion images avoids false alarms and missed alarms from sensors. Furthermore, it has a wide range of applications, as smoke motion analysis and fire smoke detection can be combined with different influencing factors for different monitoring scenarios.

[0005] Existing technologies use optical refraction to determine the reflection intensity under the combination of humidity, smoke particles, and dust particles, thereby determining the smoke particles and smoke concentration in maze-like scenes; and use pixel aggregation of open flame images to determine the smoke accumulation situation when a fire occurs. However, these processing methods ignore the flow field during smoke diffusion, resulting in low smoke distribution recognition accuracy and disordered particle tracking trajectories in some locations in maze-like scenes due to the different distribution of gradients in multi-branch processes, thus reducing the accuracy of smoke alarms. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a smoke alarm method based on the movement trajectory of smoke particles, comprising: constructing a three-dimensional flow field distribution map of smoke particles based on the installation coordinates of each sensor in the current channel structure.

[0007] Based on the changes in cross-sectional area and curvature of the current channel at the branch paths, and taking the predicted smoke particle positions as the center, the velocity data of each branch path are correlated to obtain the velocity gradient sequence of each branch path.

[0008] Based on the velocity gradient sequence of each branch path, when the velocity gradient exceeds the velocity threshold, the position of smoke particles in each branch path is updated, and the positional change of smoke particles on the branch path at each time point is output.

[0009] Trajectory fitting is performed on smoke particles to form a sequence of motion trajectories corresponding to the smoke particles. Based on the concentration gradient corresponding to the motion trajectory sequence on each branch path, the concentration intensity index and update time of smoke particles on different branch paths are configured.

[0010] Based on the update time and concentration intensity indicators of different branch paths, classification labels are configured for each branch path, and graded early warnings are carried out according to the differences in flow velocity to obtain the monitoring results of each branch path.

[0011] The beneficial effects of this invention are as follows: First, this invention acquires basic flow field data such as channel type, initial position of smoke particles, air velocity, cross-sectional area, and curvature of each branch path, and then synchronizes the data to the sensor coordinates. A three-dimensional flow field distribution map is generated through multi-view reconstruction. Then, the particle position deviation is verified and determined using turbulence effects. Using the branch cross-sectional area and curvature as input geometric parameters, the correlation between the position deviation and the particle position is calculated using the Pearson correlation coefficient. After matching the optimal geometric parameters, the position deviation is synchronized to the geometric parameters through random particle placement and concentrated point superposition processing. Finally, the flow velocity at each position is fitted to construct a velocity gradient sequence. By clarifying the geometric structure and coordinates of smoke particles under each branch, and correcting the position deviation caused by turbulence effects through correlation matching and particle superposition processing, the velocity gradient sequence better reflects the dynamic distribution of flow velocity in each branch. This allows the constructed three-dimensional flow field to serve as the data benchmark for current smoke tracking, thereby improving the accuracy of velocity recognition in multiple scenarios.

[0012] Second, this invention extracts vorticity and strain rate as flow velocity features, segments the flow velocity rise / fall dual thresholds, and then captures the cross-sectional anomalies of particles based on the branch geometry. Combined with the flow velocity gradient change to trigger position update, the coordinate tracking of smoke particles is completed. At the same time, the flow velocity change is introduced to avoid drift and lag in smoke positioning, and to provide a basis for position update for the subsequent correlation between concentration and trajectory.

[0013] Third, this invention aligns the trajectory points of the motion trajectory sequence with concentration values ​​using timestamps, calculates the concentration gradient at each trajectory point, and determines the local concentration gradient of each branch. The ratio of the maximum local concentration gradient to the warning concentration gradient is used as the concentration intensity index. This is then correlated with particle update time to determine the data update process in the global scenario. Semantic classification labels are then generated based on the update time and concentration intensity index. These labels are updated in real time, and changes are assessed to finally obtain the output monitoring results. This achieves location-guided flow velocity tracing of each branch, improving the accuracy and timeliness of smoke alarms. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1 This is a flowchart illustrating a smoke alarm method based on the trajectory of smoke particles.

[0016] Figure 2 This is a flowchart illustrating step S2 of a smoke alarm method based on the trajectory of smoke particles.

[0017] Figure 3 This is a flowchart illustrating step S3 of a smoke alarm method based on the trajectory of smoke particles.

[0018] Figure 4 This is a flowchart illustrating step S4 of a smoke alarm method based on the trajectory of smoke particles.

[0019] Figure 5 This is a flowchart illustrating step S5 of a smoke alarm method based on the trajectory of smoke particles. Detailed Implementation

[0020] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0021] See Figure 1 A smoke alarm method based on the motion trajectory of smoke particles includes: S1, constructing a three-dimensional flow field distribution map of smoke particles based on the installation coordinates of each sensor within the current channel structure.

[0022] S2, taking the predicted smoke particle positions as the center, and associating the velocity data of each branch path with the cross-sectional area and curvature changes of the current channel at the branch path, to obtain the velocity gradient sequence of each branch path.

[0023] S3, based on the velocity gradient sequence of each branch path, updates the position of smoke particles in each branch path when the velocity gradient exceeds the velocity threshold, and outputs the positional change of smoke particles on the branch path at each time.

[0024] S4. Perform trajectory fitting on the smoke particles to form a sequence of motion trajectories corresponding to the smoke particles, and configure the concentration intensity index and update time of the smoke particles on different branch paths according to the concentration gradient corresponding to the motion trajectory sequence on each branch path.

[0025] S5 configures classification labels for each branch path based on the update time and concentration intensity index of different branch paths, and performs graded early warning according to flow velocity differences to obtain the monitoring results of each branch path.

[0026] The current solution involves monitoring smoke in maze-like or spiral channels to determine the smoke diffusion pattern in each branch path or spiral channel. This data is then synchronized to the three-dimensional flow field distribution, ultimately enabling the differentiation of different areas along the smoke spread path and improving the location accuracy of smoke alarms.

[0027] A three-dimensional PIV system can be used. A three-dimensional PIV system is a fluid dynamics measurement instrument based on the principle of binocular vision. Building upon a two-dimensional PIV system, it uses two sets of digital devices to simultaneously capture images of the experimental area, synthesizing a three-dimensional velocity vector field. The system features high-precision algorithm modules, an ultra-high spatial resolution camera, and high-speed digital image storage, and is widely used in low- to high-speed velocity field measurement, industrial fluid dynamics analysis, and microscopic flow field research. Furthermore, compared to a two-dimensional system, this system adds the ability to measure velocity components in the depth direction, thus assisting in identifying the flow patterns of smoke particles in multiple directions.

[0028] The selected apparatus includes a laser light source, a high-speed camera array, a tracer particle generator, and a synchronization controller to construct the three-dimensional flow field. The laser light source illuminates the smoke particles, forming a captureable optical signal. The high-speed camera array uses multiple cameras to achieve multi-view three-dimensional particle position capture. The tracer particle generator can generate 1-5μm oil mist particles. The density of oil mist particles is close to that of air, and they have good tracking ability. They are released synchronously with the smoke to assist in the accurate tracking of smoke particles. Through smoke alarm processing based on multiple flow field calculations, subsequent rapid tracking of smoke can be achieved based on the smoke concentration at multiple locations, thereby completing the tracking processing of smoke diffusion. The synchronization controller is used to synchronize the timestamps of all data.

[0029] Then, by acquiring light direction and position information through a single shot, the particle displacement in continuous image frames from multiple perspectives is calculated, and the three-dimensional velocity of the particles in the axial, radial and circumferential directions is measured. Transmission laser concentration sensors are evenly distributed at each branch port, pitch change point and inside and outside the pipe cross section to simultaneously measure the smoke concentration, so as to realize the construction of the three-dimensional flow field of the smoke.

[0030] By releasing tracer particles, the original data of the three-dimensional flow field (particle position, flow velocity) and the concentration data at each point are recorded simultaneously, and a three-dimensional flow field distribution map with timestamps is retained.

[0031] The implementation of step S1 includes: S11, obtaining the channel type corresponding to each branch path within the current channel structure, and setting the initial position coordinates, air velocity, cross-sectional area, and curvature of the smoke particles in each branch path to obtain the basic flow field data for each branch path. This is achieved by uniformly distributing a particle set along the corresponding branch path, with each particle representing a smoke particle, and marking it as the initial position set, thereby obtaining the initial position coordinates; the air velocity is obtained by measuring different branch paths.

[0032] In the current scenario, the channel types mainly refer to maze-like channels and spiral channels. Maze-like channels refer to tree-like or grid-like topological channels composed of multiple straight pipe sections, bends, branch nodes, confluence nodes, and dead ends, with a large number of optional branch paths. Spiral channels refer to coiled channels whose channel axis extends as a spiral with equal or variable curvature. In the current scheme, spiral channels belong to a type of complex structure channel. The influence of its axial curvature change on smoke movement is similar to that of multi-branch paths. At this time, for each branch location, the processing method corresponding to the current channel will be explained according to the different forms of structure of the pipe, such as straight line, straight pipe, and grid.

[0033] As for the velocity and direction of the smoke particles, they will be obtained by continuously acquiring images, decomposing them into three directions: axial, radial, and circumferential, and obtaining the velocity values ​​in the corresponding directions.

[0034] S12 synchronizes the basic flow field data to the coordinates of the corresponding sensor, and performs multi-view reconstruction using the sensor coordinates to obtain a three-dimensional flow field distribution map.

[0035] When synchronizing the basic process data, the currently collected data is synchronized to the coordinates at the corresponding viewpoint of the sensor based on the three-dimensional coordinates configured by the sensor, thereby obtaining a three-dimensional process distribution map.

[0036] In one embodiment of the present invention, in step S2, the real-time position of smoke particles is tracked according to the particle filtering algorithm, and the velocity gradient in the corresponding branch channel is determined according to the turbulence effect during air flow, thereby capturing the correlation between particle motion and velocity.

[0037] As for the predicted smoke particle positions, the initial position coordinates of the smoke particles are updated by particle filtering, and the particle positions are iterated by the posterior probability estimation of particle filtering to obtain the predicted smoke particle positions. Then, the current observed position is subtracted from the predicted position to obtain the position deviation generated in different branch paths.

[0038] like Figure 2 As shown, the implementation of step S2 includes: S21, verifying the turbulence effect of each branch path and determining the positional deviation of smoke particles in different branch paths.

[0039] S22, synchronize the positional deviation to the cross-sectional area and curvature values ​​of the branch path, and calculate the flow velocity corresponding to each smoke particle according to the axial, radial and circumferential angles of the smoke particles.

[0040] S23, fit the flow velocity of smoke particles at each location to construct the flow velocity gradient sequence of each branch path.

[0041] The purpose of verifying the turbulence effect on the branch path is to ensure that the particles are placed in the region with the most dynamic characteristics of the flow field. That is, after calculating the turbulence intensity value through the turbulence effect, the value of the turbulence intensity is used to determine whether the currently tracked smoke particles are located in the high turbulence region. A turbulence intensity threshold can be configured, such as 0.15, or the average value of the turbulence intensity measured at the pitch change of the spiral channel, the labyrinth branch, etc. The branch path with a turbulence intensity greater than the threshold is used as the main part to be verified, and the relevant positional deviation is determined.

[0042] Highly turbulent regions typically dominate the evolution of the overall flow field, and verifying these regions can improve prediction accuracy. Then, we track the smoke particles located in the highly turbulent regions and synchronize this positional deviation to the path attributes of the corresponding branch paths, such as cross-sectional area and curvature values. Then, according to the corresponding axial, radial, and circumferential angles in the three-dimensional direction, we determine the flow velocity in the three-dimensional scene. After fitting multiple positions, we can know the flow velocity gradient sequence corresponding to the smoke particles in each branch region.

[0043] In step S22, the position deviation is synchronized to the cross-sectional area and curvature value of the branch path. Since the position deviation is the basis for the current flow velocity calculation, it is necessary to correct or check the position deviation based on the changes in cross-sectional area and curvature of the spiral / branch channels in the spiral channel and the labyrinth channel, thereby completing the correction of the branch flow velocity and the setting of the gradient sequence.

[0044] The implementation of step S22 also includes: S221, using the cross-sectional area and curvature value of the branch path as input to determine the differences between different branch paths.

[0045] S222, calculate the correlation between the differences in the branch paths and the positional deviation of the smoke particles, and use the correlation to match the cross-sectional area and curvature values ​​of the branch paths.

[0046] When matching the cross-sectional area and curvature values ​​of branch paths, the implementation also includes: extracting the main branch and sub-branch of the current branch path, and forming a connected branch where the main branch and sub-branch are connected.

[0047] For the cross-sectional area and curvature values ​​of each branch path in the connected branches, a characteristic value corresponding to the cross-sectional area and curvature values ​​is set based on the difference relative to the overall mean of the connected branches. Here, the difference represents the ratio of the difference between the normalized cross-sectional area and curvature values ​​and the overall mean, thereby illustrating the influence of different shapes in multiple connected branches on the transmission of smoke particles.

[0048] The positional deviation and eigenvalues ​​of each branch path in the connected branches are input, and the Pearson correlation coefficient is used for calculation. The set of cross-sectional area and curvature values ​​with the highest Pearson correlation coefficient is taken as the output matching result. During the correlation calculation, since there are multiple sets of positional deviations of smoke particles in a single branch path, multiple sets of values ​​are extracted for different positions of a single branch path. If there are multiple smoke particles at a certain cross-section, the average value of the smoke particles is selected to complete the data setting. As for the eigenvalues ​​of cross-sectional area and curvature, they are set according to the sum of the corresponding proportional values ​​of cross-sectional area and curvature or the ratio to the total eigenvalues.

[0049] If the Pearson correlation coefficient between the displacement deviation and the cross-sectional area and curvature is large, it indicates that the branch geometry parameters have a significant impact on the displacement deviation of smoke particles; otherwise, it indicates that the branch structure is simple and has little impact on the movement of smoke particles.

[0050] S223, after completing the matching, randomly deploy smoke particles at the corresponding branch path, and concentrate the positions of the randomly deployed smoke particles to determine the concentration point of the smoke particles.

[0051] After random distribution, the K-means clustering algorithm will be used to cluster the distributed particles into K clusters. The multiple cluster centers formed by these clusters can reflect the distribution of smoke particles at specific locations, making it easier to find the concentration points of smoke particles during flow, thereby explaining the distribution of various branch paths such as the outer pipe wall of the spiral branch and the bend of the maze branch.

[0052] S224. Based on the center position of the smoke particle concentration point, the smoke particles are superimposed to determine the position of the corresponding smoke particles after superposition, and these smoke particles are synchronized to the cross-sectional area and curvature value of the branch path; then, the velocity gradient reflected on different branch paths can be obtained based on the position changes of these smoke particles.

[0053] During the overlay process, the overlay is performed on the clusters containing the concentration points. The contour coefficient corresponding to each smoke particle is used as the distance weight, and the ratio of the positional deviation of each smoke particle to the maximum positional deviation is used as its accuracy weight. The product of the accuracy weight and the distance weight is used as the ratio of the total accuracy weight and the distance weight to calculate the weighted average of the coordinates in the clusters containing the concentration points, resulting in the weighted average center coordinates. These center coordinates represent the concentration of smoke particles in different branches, further reflecting the geometric characteristics at different branch paths and reducing positional deviations at complex paths.

[0054] In one embodiment of the present invention, in step S3, the smoke particles set in the aforementioned part are checked and the velocity gradients corresponding to these smoke particles are recorded to prevent some velocity values ​​from being too large, which would cause the motion tracking of smoke particles on the branch path to be distorted.

[0055] like Figure 3 As shown, the implementation of step S3 also includes: S31, based on the distribution position of smoke particles in the branch path, using the velocity gradient of the smoke particles to obtain vorticity and strain rate as the velocity characteristics of the current branch path.

[0056] The vorticity and strain rate selected are values ​​obtained from the velocity gradient under second-order calculation. These values ​​do not directly depend on the position coordinates and can capture the propagation characteristics of smoke particles in the three-dimensional flow field.

[0057] Vorticity can be calculated by using the central difference method to determine the partial derivatives of the velocity field in three directions. Then, the three components are combined into a vorticity vector, and the magnitude of the vorticity vector can be taken to determine the current value of vorticity.

[0058] Strain is calculated by numerical differentiation of the partial derivatives of all velocity components with respect to the coordinates. After extracting the symmetrical parts, the average value of the principal strain rate is directly calculated. The high strain region obtained can correspond to the parts with obvious branch path movement, such as the shear layer or the flow acceleration zone. Furthermore, the deformation and diffusion of smoke particles can be shown through strain.

[0059] The Q criterion can also be used to measure the difference between the square of vorticity and the square of strain rate to identify the vortex core structure corresponding to the current smoke particles. A value greater than 0 indicates that rotation is dominant and smoke is prone to form concentrated vortices in the corresponding branch path; a value less than 0 indicates that strain is dominant and smoke is prone to stretching and diffusion; if equal to 0, it indicates that the corresponding branch path is a transition region and there is no directly dominant type of diffusion at this position.

[0060] S32, based on the flow velocity characteristics of different branch paths, divides the flow velocity threshold into a flow velocity increase threshold and a flow velocity decrease threshold.

[0061] The velocity threshold will be set according to the low-speed and high-speed zones of smoke particle movement in the branch path. For example, the velocity increase threshold is the critical value for the transition from the low-speed zone to the high-speed zone, and the velocity decrease threshold is the opposite, which is the transition from the high-speed zone to the low-speed zone; in order to determine the dynamic characteristics of the acceleration and deceleration of smoke particles.

[0062] For example, refer to the average flow velocity of the corresponding branch path in historical data as the reference flow velocity, and set the flow velocity rise threshold and flow velocity fall threshold in the form of mean ± 3 times standard deviation. Alternatively, introduce an empirical coefficient and select 2-4 times the current flow velocity standard deviation in the way of minimum error sum of squares. This multiple is determined based on the minimum error sum of squares.

[0063] S33. Based on the geometry of the branch path, determine the anomaly of the smoke particles in the cross section. When the smoke particles approach the path bifurcation point, record the position of each smoke particle. When the velocity gradient corresponding to the current smoke particle changes, update the data at the corresponding branch path and use the updated data as the output data.

[0064] At this point, based on the geometry of the branch path, the specific details of the bifurcation angle, branch length, and cross-section will be recorded, as well as the movement of the smoke particles as they reach the bifurcation point, thereby enabling data tracking of the smoke particles.

[0065] When the velocity gradient corresponding to the current smoke particles changes, the implementation method includes: S331, if the current velocity gradient exceeds the velocity increase threshold, the position of the smoke particles in each branch path is updated; the update method will adopt Kalman filtering to introduce the optical flow field scattered by the current laser to update the position of the smoke particles.

[0066] S332, if the current velocity gradient is less than the velocity decrease threshold, then stop updating the position of smoke particles in each branch path.

[0067] S333: If the current velocity gradient is between the velocity decrease threshold and the velocity increase threshold, then the smoke particles are tracked based on the previous update of the smoke particle position.

[0068] In step S3, in addition to describing its flow rate threshold, the gradient changes of smoke particles are also tracked to determine the spatial correlation between different branch paths, thereby realizing the correlation between each path in data updates.

[0069] During the update, the main focus is on correcting the velocity direction, and then updating the coordinates of the smoke particles according to the corrected velocity, thereby completing the recording of the smoke coordinates.

[0070] For example, after a particle enters another branch path, its velocity direction must be consistent with the direction of that branch path, while retaining some of the mainstream velocity component. Let's assume the corrected velocity is: ;in, This represents the velocity vector of the smoke particles after they enter another branch path. This represents the velocity vector of the smoke particles on the current branch path; The unit direction vector represents the branch path, where i represents the index number of the different branches and is used to point to other branch paths that are connected to the current branch path. Its value range is based on the number of other branch paths. For example, the value range of i can be 1 or 2. This represents the velocity projection term, indicating the velocity components projected onto other branch paths from the current velocity. This represents the direction correction factor, with a value range of 0.7-0.95, which is set based on the angle of the branch path and the inertia of the current smoke particles under the corresponding branch path; This represents the mainstream velocity vector, typically indicating the average flow velocity upstream of the current branch path; This represents the mainstream velocity retention coefficient, used to reflect the inertia when flowing through multiple branch paths. It can be set to any value between 0.05 and 0.3 and is affected by the number of smoke particles deployed on the current branch path. This represents the turbulent velocity disturbance vector, which is the product of the turbulence intensity and a unit random vector, to explain the random disturbances caused by the turbulent velocity. This represents the turbulence disturbance coefficient, which indicates the degree of influence of turbulence disturbance on particle velocity. It is generally taken as 0.05-0.2, and its value will be selected based on the turbulence intensity.

[0071] At this point, by tracking the flow at the intersection of the branch paths, the coordinate changes of the smoke particles in the three-dimensional flow field can be further verified; as for updating the coordinates, the particle update can be completed based on the velocity vector and time at this time.

[0072] If the updated coordinates exceed the 3D boundary of the branch path, then the velocity direction needs to be corrected according to the law of reflection and the coordinates need to be updated again. For example... ;in, This represents the boundary normal vector, ultimately completing the setting and tracking of coordinate changes.

[0073] In one embodiment of the present invention, in step S4, the trajectory of smoke particles under the velocity gradient and the coordinates corresponding to the smoke concentration are fitted to determine the movement of smoke particles in the corresponding branch path, forming a motion trajectory sequence including position, concentration, velocity, and time. Then, according to the concentration value, a concentration intensity index is set for the branch path. For example, the ratio of the maximum local concentration gradient at the branch path location to the warning concentration gradient is selected as the concentration intensity index for the current branch path; or the ratio of the concentration value at the branch path location to the warning concentration value is used to illustrate the concentration distribution compared to the warning point. This index reflects the local concentration gradient of smoke particles during movement, thus explaining the response required during smoke tracking. Under the currently set three-dimensional flow field, the concentration gradient better reflects the spatial diffusion of smoke, thereby explaining the situation of fire intensification or smoke spreading to high-risk areas.

[0074] like Figure 4 As shown, the implementation of step S4 includes: S41, extracting the trajectory points of the motion trajectory sequence, aligning each trajectory point with the concentration value using a timestamp, and thus obtaining the concentration value corresponding to the trajectory point of the smoke particles during their motion.

[0075] S42. Based on the concentration values ​​corresponding to the trajectory points, calculate the concentration gradient corresponding to each trajectory point, and use the concentration gradient of all trajectory points belonging to the same path as the local concentration gradient under the corresponding branch path. When belonging to the same path, it mainly describes the smoke particles fitted within a branch path to explain the local concentration at each trajectory point. At the same time, for positions such as corners of branch paths, the concentration points of smoke particles can be used based on these positions, and the concentration gradient corresponding to the concentration points can be used as the local concentration gradient at the corresponding positions.

[0076] S43, select the ratio of the maximum value of the local concentration gradient of the branch path to the warning concentration gradient as the concentration intensity index of the current branch path; as for the value of the warning concentration gradient, it will be based on the concentration gradient measured at the corresponding location when the smoke alarm issues a warning, so as to measure the part of the multi-branch path that can present a smoke alarm under smoke diffusion.

[0077] S44 records the concentration intensity index at each smoke particle position update. The concentration intensity index is correlated with the update time of each smoke particle, and the correlated data is regarded as the current output data. When recording the update time, the update of the smoke particle coordinates is mainly based on the change in the velocity gradient. If the velocity gradient change is too small, even if it passes through the branch point of the branch path, the position of other branch paths can be obtained by only correcting its velocity direction and coordinate transformation according to the current velocity gradient.

[0078] Meanwhile, the update time reflects the system's sensitivity to changes in the flow field. For example, a short update time indicates drastic changes in the flow field, while a long update time indicates that the system only responds to significant changes in velocity gradients, thus illustrating the changes in the flow field at different branch paths.

[0079] Preferably, when fitting the trajectory of smoke particles, the method of using the smoke center point in step S2 is often used to achieve the fitting process of multiple smoke particles. Alternatively, the coordinates of each smoke particle can be transformed into a maximum weight matching problem according to the Hungarian algorithm. Each smoke particle is assigned a probability value according to the posterior probability, and this probability value is used as the matching weight. Then, the positions of multiple smoke particles in the branch path are fitted to obtain a fitted motion trajectory sequence.

[0080] In one embodiment of the present invention, in step S5, each branch path is identified mainly based on the obtained update time and concentration intensity index, and the flow velocity gradient recorded on each branch path is combined with the description of each region according to a combination of multiple features to complete the classification label configuration for each branch path; then, according to the different smoke trajectories caused by the difference in flow velocity, the warning process of smoke changes from the starting position, the path it takes, the residence time, and the concentration changes is clarified, and the monitoring and processing of each branch path is completed with a warning triggered at any time.

[0081] like Figure 5 As shown, the implementation of step S5 includes: S51, using the update time and concentration intensity index of each branch path, setting semantic labels for the branch paths, and obtaining the configured classification labels in the form of a combination of semantic labels; the semantic labels corresponding to the concentration intensity index may include labels related to concentration diffusion at the corresponding branch path, such as high concentration rapid accumulation, high concentration slow accumulation, and low concentration slow diffusion; the update time will be divided into long update time and short update time to explain the sudden change in velocity gradient at the junction of different branch paths; the velocity gradient uses high velocity gradient, low velocity gradient, etc., and then describes whether each branch path is complex or its shape, thereby completing the configuration of the classification labels.

[0082] S52, traverse each branch path and update the classification label of each branch path in real time; judge the change of the classification label before and after the update. If the change occurs after the update, confirm the difference in update time, concentration intensity index and flow velocity gradient respectively, and make graded warnings based on the difference of each index.

[0083] When the classification label changes, it indicates a significant change in the current process. For example, increased flow rate causes particles to be carried away quickly, increasing the path update frequency; concentration gradients drive particles to diffuse into low-concentration areas, changing the path update direction; the combined effect of flow rate and concentration causes the path to spiral or bifurcate, resulting in significant differences in the current update description, requiring timely handling. In addition, when the classification label changes on each path, it will also correspond to abnormal events such as leakage, release, blockage, and obstruction, requiring further explanation of the specific situation of the smoke particles.

[0084] When using the differences between various indicators to implement graded early warning, the implementation method includes: determining the part belonging to the abnormality based on the difference value of the current update time, concentration intensity index and flow velocity gradient.

[0085] If a single indicator is abnormal, the output will be the difference value of the abnormality, and the coordinate corresponding to the difference value will be regarded as the output monitoring result.

[0086] If multiple indicators are abnormal, the current abnormal scenario is determined by the combination of abnormal indicators, and the abnormal scenario is used as the output monitoring result.

[0087] When a single indicator is abnormal, such as the update time, the average update time of all branch paths globally will be used as the baseline time. If the update time changes by more than 20% in any branch path, it will be considered that the update time is abnormal, indicating that there will be abnormal flow rate or concentration in the corresponding part, which needs to be checked in time.

[0088] The concentration intensity value will also set a baseline value based on the global average value. However, the concentration intensity value will identify the part where the rate of change exceeds 50% of the baseline value to determine the part where pollution accumulates rapidly, without verifying the absolute increase or decrease in concentration.

[0089] Because the flow velocity in the branch path is updated based on rising and falling thresholds, the flow velocity is determined according to the geometric characteristics of the current branch path. The average flow velocity under the corresponding geometric characteristics is set as its baseline value, and any flow exceeding the baseline value by 30% is identified to trigger an early warning, thus identifying flow field blockage or leakage. This completes the single-indicator anomaly judgment for the three indicators.

[0090] If the anomaly is attributed to multiple indicators, the differences between the three indicators before and after the update are introduced, and these differences are matched according to a pre-prepared numerical range to determine the current anomaly scenario. This anomaly scenario is then used as the monitoring result for the corresponding branch path.

[0091] S53, if no changes occur after the update, the current category label will be used as the basis for hierarchical early warning, and the data after the early warning will be regarded as the monitoring results of each branch path.

[0092] When there is no change, it indicates that the current three-dimensional flow field is in a relatively stable state. Early warning can be directly issued based on the high concentration gradient and smoke accumulation parts indicated in the current classification label to complete the early warning processing of each branch region.

[0093] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A smoke alarm method based on the trajectory of smoke particles, characterized in that, include: Based on the installation coordinates of each sensor within the current channel structure, a three-dimensional flow field distribution map of smoke particles is constructed. Based on the changes in cross-sectional area and curvature of the current channel at the branch paths, and taking the predicted smoke particle positions as the center, the velocity data of each branch path are correlated to obtain the velocity gradient sequence of each branch path. Based on the velocity gradient sequence of each branch path, when the velocity gradient exceeds the velocity threshold, the position of smoke particles in each branch path is updated, and the positional change of smoke particles on the branch path at each time point is output. Trajectory fitting is performed on smoke particles to form a sequence of motion trajectories corresponding to the smoke particles. Based on the concentration gradient corresponding to the motion trajectory sequence on each branch path, the concentration intensity index and update time of smoke particles on different branch paths are configured. Based on the update time and concentration intensity indicators of different branch paths, classification labels are configured for each branch path, and graded early warnings are carried out according to the differences in flow velocity to obtain the monitoring results of each branch path.

2. The smoke alarm method based on the movement trajectory of smoke particles according to claim 1, characterized in that, The methods for implementing three-dimensional flow field distribution maps include: Obtain the channel type corresponding to each branch path within the current channel structure, and set the initial position coordinates, air velocity, cross-sectional area, and curvature of the smoke particles in each branch path to obtain the basic flow field data for each branch path; The basic flow field data is synchronized to the coordinates of the corresponding sensor, and multi-view reconstruction is performed using the sensor coordinates to obtain a three-dimensional flow field distribution map.

3. The smoke alarm method based on the motion trajectory of smoke particles according to claim 1, characterized in that, The implementation methods of flow velocity gradient sequences include: Turbulence effects were verified in each branch path to determine the positional deviation of smoke particles in different branch paths. The positional deviation is synchronized to the cross-sectional area and curvature values ​​of the branch path, and the flow velocity corresponding to each smoke particle is calculated according to the axial, radial and circumferential angles of the smoke particles. The velocity of smoke particles at various locations is fitted to construct the velocity gradient sequence for each branch path.

4. The smoke alarm method based on the motion trajectory of smoke particles according to claim 3, characterized in that, When synchronizing the positional deviation to the cross-sectional area and curvature values ​​of the branch path, the implementation methods also include: Using the cross-sectional area and curvature values ​​of the branch paths as input, the differences between different branch paths are determined; The correlation between the differences in branch paths and the positional deviation of smoke particles is determined, and the cross-sectional area and curvature values ​​of the branch paths are matched with the correlation. After matching is completed, smoke particles are randomly deployed at the corresponding branch paths, and the locations of the randomly deployed smoke particles are concentrated to determine the concentration point of the smoke particles. Based on the center position of the smoke particle concentration point, the smoke particles are superimposed to determine the position of the corresponding smoke particles after superposition, and these smoke particles are synchronized to the cross-sectional area and curvature value of the branch path.

5. A smoke alarm method based on the trajectory of smoke particles according to claim 4, characterized in that, The implementation methods for matching the cross-sectional area and curvature values ​​of branch paths also include: Extract the main branch and sub-branch of the current branch path, and combine the branch paths into connected branches that link the main branch and sub-branch; For each branch path in the connected branches, the characteristic value corresponding to the cross-sectional area and curvature value is set based on the difference relative to the overall mean of the connected branches; The positional deviation and eigenvalue of each branch path in the connected branches are input, and the Pearson correlation coefficient is used to calculate the result. The set of cross-sectional area and curvature values ​​with the largest Pearson correlation coefficient is used as the output matching result.

6. A smoke alarm method based on the trajectory of smoke particles according to claim 1, characterized in that, The implementation of updating the smoke particle positions within each branch path also includes: Based on the distribution of smoke particles in the branch path, the vorticity and strain rate are obtained using the velocity gradient of the smoke particles, which are used as the velocity characteristics of the current branch path. Based on the flow velocity characteristics of different branch paths, the flow velocity threshold is divided into a flow velocity increase threshold and a flow velocity decrease threshold. Based on the geometry of the branch path, the anomaly of the smoke particles in the cross section is determined. When the smoke particles approach the path bifurcation point, the position of each smoke particle is recorded. When the velocity gradient corresponding to the current smoke particle changes, the data at the corresponding branch path is updated, and the updated data is used as the output data.

7. A smoke alarm method based on the trajectory of smoke particles according to claim 6, characterized in that, When the velocity gradient corresponding to the current smoke particles changes, the implementation methods include: If the current velocity gradient exceeds the velocity rise threshold, the positions of smoke particles in each branch path will be updated. The update method will use Kalman filtering to introduce the optical flow field scattered by the current laser to update the positions of smoke particles. If the current velocity gradient is less than the velocity decrease threshold, then stop updating the position of smoke particles in each branch path; If the current velocity gradient is between the velocity decrease threshold and the velocity increase threshold, the smoke particles are tracked based on the previous update of the smoke particle position.

8. A smoke alarm method based on the trajectory of smoke particles according to claim 1, characterized in that, When configuring the concentration intensity index and update time of smoke particles in different branch paths, the implementation methods also include: The trajectory points of the motion trajectory sequence are extracted, and each trajectory point is aligned with the concentration value by timestamp, so as to obtain the concentration value corresponding to the trajectory point of the smoke particles when they are moving. Based on the concentration values ​​corresponding to the trajectory points, the concentration gradient corresponding to each trajectory point is calculated, and the concentration gradient of all trajectory points belonging to the same path is regarded as the local concentration gradient under the corresponding branch path. The ratio of the maximum local concentration gradient of the branch path to the warning concentration gradient is selected as the concentration intensity index of the current branch path. Record the concentration intensity index at each smoke particle location update, correlate the concentration intensity index with the update time of each smoke particle, and use the correlated data as the current output data.

9. A smoke alarm method based on the trajectory of smoke particles according to claim 1, characterized in that, When obtaining the monitoring results for each branch path, the implementation methods include: Using the update time and concentration intensity indicators of each branch path, semantic labels are set for the branch paths, and the configured classification labels are obtained in the form of a combination of semantic labels. Traverse each branch path and update the classification labels of each branch path in real time; determine the changes in the classification labels before and after the update; if the changes occur after the update, confirm the differences in update time, concentration intensity index and flow velocity gradient respectively, and conduct graded early warning based on the differences in each index. If no changes occur after the update, the current category label will be used as the basis for the hierarchical warning, and the data after the warning will be regarded as the monitoring results of each branch path.

10. A smoke alarm method based on the trajectory of smoke particles according to claim 9, characterized in that, When implementing tiered early warning based on differences in various indicators, the methods include: The portion belonging to the anomaly is determined based on the difference between the current update time, concentration intensity index, and flow velocity gradient. If a single indicator is abnormal, the output will be the difference value of the abnormality, and the coordinate corresponding to the difference value will be regarded as the output monitoring result. If multiple indicators are abnormal, the current abnormal scenario is determined by the combination of abnormal indicators, and the abnormal scenario is used as the output monitoring result.

Images