CN116090333A - Urban public space disaster modeling and preventing system based on perception blind area estimation - Google Patents
Urban public space disaster modeling and preventing system based on perception blind area estimation Download PDFInfo
- Publication number
- CN116090333A CN116090333A CN202211609912.XA CN202211609912A CN116090333A CN 116090333 A CN116090333 A CN 116090333A CN 202211609912 A CN202211609912 A CN 202211609912A CN 116090333 A CN116090333 A CN 116090333A
- Authority
- CN
- China
- Prior art keywords
- track
- pedestrian
- segment
- public space
- urban public
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008447 perception Effects 0.000 title claims abstract description 48
- 230000004927 fusion Effects 0.000 claims abstract description 33
- 238000001514 detection method Methods 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000002159 abnormal effect Effects 0.000 claims abstract description 24
- 230000002265 prevention Effects 0.000 claims abstract description 14
- 238000011084 recovery Methods 0.000 claims abstract description 7
- 238000012512 characterization method Methods 0.000 claims abstract description 6
- 238000012544 monitoring process Methods 0.000 claims description 38
- 206010000117 Abnormal behaviour Diseases 0.000 claims description 23
- 239000012634 fragment Substances 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000004088 simulation Methods 0.000 claims description 18
- 238000000034 method Methods 0.000 claims description 17
- 239000000779 smoke Substances 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000012546 transfer Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 9
- 230000007704 transition Effects 0.000 claims description 9
- 230000005856 abnormality Effects 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 230000006399 behavior Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 4
- 239000000463 material Substances 0.000 claims description 4
- 230000002238 attenuated effect Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000013434 data augmentation Methods 0.000 claims description 3
- 238000004821 distillation Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims 1
- 238000007500 overflow downdraw method Methods 0.000 abstract description 2
- 238000010276 construction Methods 0.000 abstract 1
- 208000027744 congestion Diseases 0.000 description 30
- 238000005516 engineering process Methods 0.000 description 6
- 241000700605 Viruses Species 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000000969 carrier Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 230000005180 public health Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/80—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for detecting, monitoring or modelling epidemics or pandemics, e.g. flu
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Medical Informatics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Educational Administration (AREA)
- Computer Security & Cryptography (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- Alarm Systems (AREA)
Abstract
The invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, which comprises six parts including local information characterization, social force model simulator construction, local information fusion, global track recovery and multi-task combined abnormal event prediction. According to the invention, the segment track is obtained based on local information characterization in the sensing range of the sensor obtained by multi-source heterogeneous data and combined with the social force model simulator, so that a track fusion method suitable for multiple sensors in the urban public space is provided, a fusion result is obtained by solving an optimal matching problem based on two factors of direction and time, and the complete sensing in the urban public space is recovered based on the social force model simulator, so that the problem that the sensing blind area information is difficult to obtain is solved; and finally, based on the correlation of occurrence relations among a plurality of different abnormal events, carrying out joint training on a plurality of abnormal event detection or prediction tasks, and realizing disaster modeling and prevention based on blind area perception.
Description
Technical Field
The invention relates to a disaster prevention system for urban public space, and belongs to the technical field of artificial intelligence.
Background
With the development of modern society and the gradual improvement of economic substance conditions, sensing devices such as: temperature sensor, humidity sensor and CO 2 The sensor, the monitoring camera and the like are widely applied to management of modern cities, especially public places, so that the urban operation efficiency is improved, the public place safety is ensured, and the living happiness of citizens is improved. Existing sensing technologies, such as: crowd tracking has evolved relatively well for crowd estimation within the perception range of a single perceptron.
However, in practical public places, a single sensor has great limitations on the spatial coverage of the public places, mainly in three aspects:
1) The perception range of the sensor is limited, taking a camera as an example, the effective perception area is limited, and the effective perception area is very limited by the installation position and angle;
2) The complexity of the environment in the public field limits the perception range of the sensor to a great extent, and the complexity of public facilities such as shielding of obstacles, stairs, elevators and the like, so that the perception range and the accuracy of the camera and the sensor are greatly influenced;
3) The diversity of crowd behaviors and the variability of crowd density in public areas can also affect the perception range of the perceptrons.
Therefore, it is difficult for a single sensor to fully cover a public environment, and even if a sensing network is formed by a plurality of sensors, a sensing blind area is unavoidable. However, the existing sensing technology, whether based on a single sensor or a sensing network composed of multiple sensors, is limited to the effective sensing range of the existing sensor, and ignores the sensing blind area. Taking crowd tracking as an example, the crowd tracking technology of a single camera focuses on crowd tracking on video data extracted by the single camera, and the problem of track fusion among cameras is considered in multi-camera tracking, but the existing method mainly deals with outdoor scenes with sparse crowd, the used track fusion method mainly takes greedy strategies and less global optimization strategies into consideration, and meanwhile, the methods only pay attention to pedestrian tracks in a perceivable area, but do not pay attention to pedestrian tracks in a perceivable blind area. However, in the public field, these blind areas are also important parts of the public space, especially in places where the sensors are very sparsely distributed.
Disclosure of Invention
The invention aims to solve the technical problems that: in order to solve the defects in the prior art, the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation.
In order to solve the technical problems, the invention adopts the following technical scheme:
the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, which comprises the following steps:
step S1, obtaining local track information characterization in a sensing range of a sparse sensor to obtain a pedestrian track, namely a fragment track set, in the sensing range of each sensor;
s2, drawing an environment plan, and calculating the nearest point of each position to each obstacle according to the plan information for inquiring in the simulation process; constructing a social force model simulator, calculating social force born by each pedestrian, and carrying out path inquiry and intermediate target reselection by constructing a communication network diagram when the social force model fails; outputting a state sequence of each pedestrian, and processing to obtain a pedestrian track in a simulation time period; finally, according to the local track information in the step S1, parameters of the social force model simulator are adjusted;
step S3, segment track fusion: the segment track fusion problem is designed into a task allocation problem through a transfer matrix; then solving the maximum matching optimization problem by using a Hungary algorithm to obtain a fragment track fusion sequence;
s4, restoring the pedestrian track according to the fragment track fusion sequence obtained in the step S3; deducing the blind area track by using a social force model simulator, and generating a complete simulated pedestrian track according to the output of the simulator;
s5, detecting the crowd density of the local area in a specific period by using a crowd density detection model, and correcting the prediction result of the crowd movement track segment in the step S4; and counting the pedestrian density in the range of the perception dead zone according to the obtained pedestrian track, and realizing modeling and prediction of various urban public space disasters.
Furthermore, the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, and the step S1 is specifically as follows:
step 101, constructing a training specific data set and manually marking; the data in the constructed dataset includes: the subway station comprises a camera monitoring video, a gate shooting record, millimeter wave sensor data and temperature and humidity sensor data;
step 102, selecting a single-sensor multi-target tracking frame, loading pre-trained parameters, and migrating to a marked training data set for training;
and 103, using the trained model for implementing a track recovery data set to acquire the track of the pedestrian in the sensing range of each sensor, namely a fragment track set.
Furthermore, the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, and the step S2 is specifically as follows:
step 201, drawing an environment plan, rasterizing and storing the environment plan as a dictionary, wherein keywords are represented by int-type digital symbols, and K= (x-x) min )+(y-y min ) H, a recorded value of 0 indicates that the location is passable, -1 indicates that the location is occupied by an obstacle or boundary.
Step 202, calculating the nearest point of each position to each obstacle according to the geometric plan information, and storing the nearest point as a dictionary for inquiring in the simulation process;
step 203, constructing a connected network graph, wherein the vertexes of the network graph represent each grid in the position, no non-passable area exists on the grids, and if two vertexes are directly reachable, namely one grid is in the adjacent eight directions of the other grid, an edge exists between the two vertexes, and the connected graph is used for path inquiry and intermediate target reselection when the social force model fails;
step 204, firstly adding pedestrians in a new entering environment at each moment, then judging whether a blocked pedestrian exists, if the moving range of a certain pedestrian in the previous 5 steps is concentrated in a grid range, indicating that the pedestrian cannot normally travel based on the social force calculated by the current target, searching a shortest path, setting a new target, searching a coarse-precision shortest path by using dijkstra shortest path algorithm based on the established communication diagram, and then inserting a sampling position point in the shortest path into the new intermediate target;
step 205, calculating social force suffered by each pedestrian in the following way:
m i for the mass of the pedestrian i,
acceleration for pedestrian i->
Three social forces to which the pedestrian i is subjected are respectively represented: driving force, obstacle materials and interaction force; the specific calculation is as follows: />
1) Driving force:
wherein τ i To indicate the time required for a pedestrian to transition to an ideal speed for relaxationTime of (2);
2) Obstacle material resource Obstacle force:
d io (t) represents the distance from the pedestrian i to the obstacle o at time t;
3) Interaction force Interactive force:
r is the collision radius of the pedestrian.
The acceleration of each pedestrian is obtained through the calculation, so that the travelling speed of the current time step is obtained; before updating the position, judging whether the pedestrian enters an obstacle in the next step, if so, reducing the travelling speed according to the set attenuation degree, and if the pedestrian still cannot travel according to the attenuated speed after repeating the step for three times, setting the current speed to be 0;
step 206, repeating the steps 204 and 205 until the time steps are completed, finally outputting the state sequence of each pedestrian, and processing to obtain the pedestrian track in the simulation time period;
step 207, according to the local track information obtained in the step S1, adjusting parameters of the social force model simulator; the main parameters were set as follows: desired speed v 0 ＝1.5，A o ＝1.0，A i ＝2.7。
Furthermore, the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, and the step S3 is specifically as follows:
step 301: the segment track fusion problem is designed into a task allocation problem by designing a transfer matrix, and the transfer matrix is as follows:
wherein C is 1 Representing the transfer matrix between the two segment tracks,
C 2 a transition matrix representing a segment of the track as a track termination segment,
C 3 a transition matrix representing a segment of the track as an initial segment of the track,
step 302: calculating a transfer matrix for C 1 A section for calculating a similarity between the two segment trajectories for each element, the similarity being made up of two sections, one based on time and the other based on movement direction; for C 2 And C 3 Based on the prior information of the segment track, if the segment track k is the initial track segment, C (N+k)k = + infinity of the two points, if the segment track k is the initial track segment, then C k(N+k) = + infinity of the two points, N is the number of segment tracks;
step 303: and solving the maximum matching optimization problem by using the Hungary algorithm, and obtaining the track fragment fusion sequence result.
Furthermore, the invention provides a city public space disaster modeling and preventing system based on perception blind area estimation, and the step S4 is specifically as follows:
step 401, an input file required for constructing a social force model comprises pedestrian entering conditions at all times, and an intermediate target point set of each pedestrian: obtaining all local known fragment track sequences of each pedestrian according to the track fusion result obtained in the step S3; setting some middle position points of the track fragment sequences as middle targets of pedestrians in the simulator, and guiding the simulation intelligent body to walk as far as possible according to the existing route, so as to truly restore the tracks of the pedestrians;
and step 402, executing a social force model simulator, and generating a simulated pedestrian track according to the output of the simulator to obtain a final recovered complete track.
Furthermore, the urban public space disaster modeling and preventing system based on the perception blind area estimation provided by the invention realizes modeling and predicting of various urban public space disasters in step S5, and specifically comprises the following steps:
step 501, aiming at the characteristics of dense smoke and open fire accompanied by fire, acquiring an image data set containing the dense smoke and the open fire, expanding the fire data set by using a data augmentation method comprising Rotation, randomCrop, and constructing an image two-class model containing the open fire or smoke and not containing the open fire or smoke by using a deep convolution network so as to model the fire to obtain a fire detection model, thereby realizing accurate identification and alarm of images shot in a fire scene;
step 502, classifying the abnormal behaviors of pedestrians in the monitoring video into two types of abnormal objects and abnormal actions, collecting monitoring video data containing the abnormal behaviors of the pedestrians, taking the restored input video segments as targets to train an automatic encoder model, modeling the abnormal behaviors of the pedestrians in the monitoring video, and considering the abnormal behaviors of the pedestrians to occur when the automatic encoder model cannot accurately restore the input video segments; aiming at the characteristics of a part of scenes occurring in a fire scene, the abnormal degree of a video fragment is marked in an auxiliary mode by using a fire detection model, and the training of a pedestrian abnormal behavior detection model is guided by using a training framework of distillation learning, so that accurate time positioning and alarming of a pedestrian abnormal behavior event are realized;
step 503, aiming at the correlation between the crowd congestion anomaly and the crowd density in the urban public space, collecting monitoring video data containing pedestrians and constructing a crowd density monitoring model in the urban public space, and when the crowd density exceeds a specific threshold value, regarding the occurrence of congestion so as to model the congestion and trampling disasters; aiming at the characteristic that the extreme change of crowd density often accompanies abnormal pedestrian behaviors, a multi-task learning combined training pedestrian abnormal behavior detection model and a crowd density detection model are used; the multi-scale feature extraction and fusion crowd density estimation algorithm architecture of the encoder-decoder structure is used in combination with the cavity convolution method, and the multi-scale feature extraction and fusion crowd density estimation algorithm architecture consists of a feature extraction module based on multi-column convolution, a self-supervision training module based on the encoder-decoder and a density regression module based on the multi-task architecture, so that the crowd congestion abnormality in the monitoring video image is accurately detected together;
step 504: detecting the crowd density of the local area in a specific period by using the crowd density detection model obtained in the step 503, and correcting the prediction result of the crowd movement track segment in the step S4;
step 505: and (3) counting urban public space environments into grids of 2m multiplied by 2m, counting pedestrian densities in the environments, especially in the range of perception dead zones, according to the pedestrian tracks obtained in the step S4, monitoring the congestion, prompting early warning information for the grids with crowd densities larger than delta, and timely taking intervention measures to dredge the congestion.
Compared with the prior art, the invention has the following technical effects:
(1) Aiming at the situation that the current research lacks in the research of the public place perception blind area, the invention provides a blind area track recovery scheme, and the complete track of pedestrians in the concerned space can be recovered through sparse perception data by applying a social force model simulator.
(2) Aiming at the tracking result of a single sensor, the invention provides a new scheme for fusing the segment tracks, which is suitable for the scene, not only considers time factors, but also uses spatial information as far as possible, considers the conversion cost of the moving direction, and can better fuse the segment tracks and recover the people stream condition.
(3) And designing a disaster prevention combined training scheme, and predicting pedestrian abnormality and urban public space abnormality through the restored track so as to take preventive measures in time.
(4) Congestion prediction is implemented, the congestion situation which is about to occur in the environment is predicted in advance, and corresponding measures are convenient for management personnel to take.
(5) The epidemic contactor query based on local perception is realized, and is used for epidemic contactor tracking in epidemic propagation management, so that the management efficiency is improved, and the management cost is reduced.
(6) And modeling and visualizing urban public space disasters are realized.
Drawings
FIG. 1 is a diagram of a blind spot deduction architecture.
FIG. 2 is a disaster modeling task graph.
FIG. 3 is a flow chart of social networking model operation.
Fig. 4-7 are examples of track restoration results, where (a) is the true track pairing, (b) is the track fusion result, and (c) is the comparison of the restored complete track with the true track.
Fig. 8 is an example of trace restoration failure, where (a) is the true trace pairing, (b) is the trace fusion result, and (c) is the comparison of the restored complete trace with the true trace.
Fig. 9-12 are examples of congestion monitoring results, where (a) is true crowd density and (b) is recovered crowd density.
Fig. 13-15 are congestion monitoring failure examples, where (a) is true crowd density and (b) is recovered crowd density.
Fig. 16 is an example of an epidemic contactor query Truepositive.
Fig. 17 is an example of an epidemic contactor query false.
Fig. 18 is an example of an epidemic contactor query false.
Fig. 19 is an example of congestion prediction and prevention for a crowd motion simulation model in which (a) is true crowd density and (b) is predicted crowd density.
Fig. 20 is a schematic diagram of an intra-station emergency early warning module according to an embodiment of the present invention, where (a) is a schematic diagram for early warning of crowd congestion events that may occur in the future in the station; (b) Is a schematic diagram for giving an alarm to open fire and dense smoke in the station.
Detailed Description
The technical scheme of the invention is further described with reference to the accompanying drawings and the embodiments. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The invention firstly obtains the local information characterization of pedestrians in the perception range based on a sparse sensor, extracts the segment track set of the pedestrians, combines a social force model simulator to fuse the segment tracks so as to obtain all the information of the pedestrians in the known perception area, and designs a segment track fusion mode based on the maximum matching problem. And then, estimating the complete track based on the fusion result of the segment tracks, and particularly acquiring the track of the pedestrian in the blind area to obtain crowd activity information of the whole area. And finally, modeling various disaster events, performing joint training on detection or prediction tasks of a plurality of abnormal events based on the correlation of occurrence relations among the plurality of different abnormal events, and implementing epidemic contact tracking, so that the management and control cost is reduced, the management efficiency is improved, meanwhile, congestion prediction is implemented, the position of the subway station to be congested is predicted in advance, and management staff can conveniently take intervention measures even.
As shown in FIG. 1, the invention provides a technical scheme for acquiring global tracks by estimating a perception blind area through local perception of converged multi-source data. The local perception information such as gate shooting information, monitoring camera video, millimeter wave recording, temperature and humidity sensor recording and the like is used for the perception inference of the blind area, so that the complete track of pedestrians, namely global crowd perception, is obtained. The obtained global crowd perception is used for multi-task combined training, and is used for monitoring and predicting congestion, monitoring abnormal behaviors, monitoring disasters such as smoke and fire and the like, and performing model training on the tasks with the existing global information to complete corresponding tasks.
As shown in fig. 2, the invention constructs a disaster modeling system in the urban public space, and based on a cooperative interoperation mechanism and data aggregation and optimization, the invention fuses multi-source heterogeneous data to implement related detection and prediction tasks of disasters and anomalies in the urban public space. The disaster modeling system in the urban public space constructed by the invention comprises pedestrian abnormal behavior detection, local congestion monitoring and prediction, fire disaster and hazard gas detection, and realizes comprehensive and systematic management and control on the series of abnormalities, thereby assisting regional management staff to discover and take corresponding intervention measures in time.
According to the invention, modeling is carried out on disasters in the urban public space, a corresponding detection system is constructed, and local information characterization is obtained based on sparse deployment sensors, so that dead zones are inferred, and the urban public space disaster prevention system based on sparse sensor perception is realized. The specific implementation steps of the invention comprise:
step 1: the method comprises the following steps of obtaining local information representation in a perception range of a sparse sensor:
step 1-1: constructing a training specific data set, wherein the data set is constructed by a camera monitoring video, a gate shooting record, millimeter wave sensor data, temperature and humidity sensor data and the like in a subway station, and is manually marked;
step 1-2: selecting an existing single-sensor multi-target tracking frame, loading pre-trained parameters, and migrating to a marked training data set for training;
step 1-3: and using the trained model for implementing the track recovery data set to acquire the pedestrian track in each camera, namely a fragment track set, wherein N is the fragment track number.
Step 2: the social force model crowd simulator is constructed according to the following steps, and the flow chart of the steps is shown in fig. 3:
step 2-1: drawing an environment plan, rasterizing the environment plan into a rough map of 10cm multiplied by 10cm for improving the running speed of the simulator, storing the rough map as a dictionary, and representing keywords by int-type digital symbols for saving memory, wherein K= (x-x) min )+(y-y min ) H, a recorded value of 0 indicates that the location is passable, -1 indicates that the location is occupied by an obstacle or boundary;
step 2-2: according to the geometric plan information, calculating the nearest point of each position to each obstacle, and storing the nearest point as a dictionary for inquiring in the simulation process, wherein the calculation accuracy is still 10cm multiplied by 10cm;
step 2-3: constructing a connected network graph with the precision of 1m multiplied by 1m, wherein the vertexes of the network graph represent each 1m multiplied by 1m grid on the position, no non-passable area is arranged on each grid (so as to ensure the accessibility of a path), and if two vertexes are directly accessible, namely one grid is in eight adjacent directions (east, southeast, south, southwest, west, northwest, north and northeast) of the other grid, an edge exists between the two vertexes, and the connected graph is used for path inquiry and intermediate target reselection when a social force model fails;
step 2-4: at each moment, firstly adding a pedestrian newly entering an environment, then judging whether a blocked pedestrian exists, if the moving range of a certain pedestrian in the previous 5 steps is concentrated in a range of 10cm multiplied by 10cm, indicating that the pedestrian cannot normally travel based on the social force calculated by the current target, searching a shortest path, setting a new target, searching a coarse-precision shortest path by using dijkstra shortest path algorithm based on the established communication diagram, and then inserting a sampling position point into the new intermediate target on the shortest path;
step 2-5: the social force suffered by each pedestrian is calculated as follows:
m i for the mass of the pedestrian i,
acceleration for pedestrian i->
Three social forces to which the pedestrian i is subjected are respectively represented, and are specifically calculated as follows:
1)Driven force：
wherein τ i For relaxation time, which is the time required for the pedestrian to transition to an ideal speed, it is understood that the pedestrian's movement is continuous and smooth, typically with τ set i ＝2。
2)Obstacle force：
d io (t) recording the distance from the pedestrian i to the obstacle o at the moment t.
3)Interactive force：
r is the collision radius of the pedestrian and is typically 0.2m.
The acceleration of each pedestrian is obtained through the calculation, so that the travelling speed of the current time step is obtained; before updating the position, judging whether the pedestrian enters an obstacle in the next step, if so, reducing the travelling speed according to the attenuation degree of 0.8, and if the pedestrian still cannot travel according to the attenuated speed after repeating the step for three times, setting the current speed to be 0;
step 2-6: repeating the steps 2-4 and 2-5 until the time steps are completed, finally outputting the state sequence of each pedestrian, and processing to obtain the pedestrian track in the simulation time period.
Step 2-7: and (3) adjusting parameters of the social force model simulator according to the local track information obtained in the step (1), wherein the main parameters are set as follows: desired speed v 0 ＝1.5，A o ＝1.0，A i ＝2.7。
Step 3: the segment tracks are fused into a segment track sequence according to the following steps, and track fusion results are shown in the (a) and (b) diagrams of fig. 4-8, and are real track segment pairing and track segment fusion results respectively. Fig. 4, 5, 6, and 7 are examples of track restoration results, and fig. 8 is an example of track restoration failure.
Step 3-1: the segment track fusion problem is designed into a task allocation problem by designing a transfer matrix, and the transfer matrix is as follows:
wherein C is 1 Representing the transfer matrix between the two segment tracks,
C 2 a transition matrix representing a segment of the track as a track termination segment,
C 3 a transition matrix representing a segment of the track as an initial segment of the track,
step 3-2: calculating a transfer matrix for C 1 In the parts, each element calculates the similarity between every two segment tracks, and the similarity consists of two parts, one is based on time and the other is based on the moving direction. For C 2 And C 3 Based on the prior information of the segment track, if the segment track k is the initial track segment, C (N+k)k = + infinity of the two points, if the segment track k is the initial track segment, then C k(N+k) ＝+∞。
Step 3-3: and solving the maximum matching optimization problem by using the Hungary algorithm, and obtaining the track fragment fusion sequence result.
Step 4: and deducing the blind area track by using a social force model simulator, wherein an example of a track recovery result is shown in (c) diagrams of fig. 4-8, and a broken line is a real track and is realized as a recovered track.
Step 4-1: an input file required for constructing the social force model comprises pedestrian entering conditions at all times, and an intermediate target point set of each pedestrian: obtaining all local known fragment track sequences of each pedestrian according to the track fusion result obtained in the step 3; some middle position points of the track fragment sequences are set as middle targets of pedestrians in the simulator and used for guiding the simulation intelligent body to walk according to the existing route as far as possible, so that the tracks of the pedestrians are truly restored.
Step 4-2: executing a social force model simulator, and generating a simulated pedestrian track according to the output of the simulator to obtain a final recovered complete track.
Step 5: disaster modeling and prediction based on multi-task combined training are carried out according to the following steps:
step 5-1: aiming at the characteristic that the fire is accompanied by the dense smoke and the open fire, the image data set containing the dense smoke and the open fire is collected, and the fire is considered to happen when the dense smoke and the open fire appear in the image. The fire data set is expanded by using a Rotation, randomCrop data augmentation method, and an image two-classification model containing open fire or smoke and not containing open fire or smoke is constructed by using a deep convolution network so as to model the fire and realize accurate identification and alarm of images shot on a fire scene.
Step 5-2: the pedestrian abnormal behavior in the monitoring video is divided into two types of abnormal objects and abnormal actions, monitoring video data containing the pedestrian abnormal behavior are collected, an automatic encoder model is trained by taking the input video segment as a target to be recovered, the pedestrian abnormal behavior in the monitoring video is modeled, and the pedestrian abnormal behavior is considered to occur when the automatic encoder model cannot accurately recover the input video segment. Aiming at the characteristics of the scene of the fire disaster, the abnormal degree of the video clips is marked with the aid of the fire disaster detection model, the training of the pedestrian abnormal behavior detection model is guided by the training framework of distillation learning, and accurate time positioning and alarming of the pedestrian abnormal behavior event are realized.
Step 5-3: aiming at the correlation between crowd congestion abnormality and crowd density in urban public space, collecting monitoring video data containing pedestrians and constructing a crowd density monitoring model in urban public space, and when the crowd density exceeds a specific threshold value, the crowd is regarded as congestion occurrence so as to model congestion and trampling disasters. Aiming at the characteristic that the extreme change of crowd density is often accompanied by abnormal pedestrian behaviors, a multi-task learning combined training pedestrian abnormal behavior detection model and a crowd density detection model are used. The multi-scale feature extraction and fusion crowd density estimation algorithm architecture of the encoder-decoder structure is used in combination with the cavity convolution method, and the multi-scale feature extraction and fusion crowd density estimation algorithm architecture consists of a feature extraction module based on multi-column convolution, a self-supervision training module based on the encoder-decoder and a density regression module based on the multi-task architecture, so that accurate detection of crowd congestion abnormality in a monitoring video image is achieved together, and intervention measures are taken in time. Fig. 9, 10, 11, 12 are examples of congestion monitoring results, where (a) is a true congestion situation and (b) is a recovered congestion situation. Fig. 13, 14, and 15 show examples of congestion monitoring failure, and the recovered congestion situation differs from the actual situation.
Step 5-4: and (3) detecting the crowd density of the local area in a specific period by using the crowd density detection model obtained in the step (5-3), and correcting the prediction result of the crowd movement track segment in the step (4).
Step 5-5: counting the environment in the urban public space as a grid of 2m multiplied by 2m, counting the pedestrian density in the environment, especially in the range of a perception dead zone, according to the pedestrian track obtained in the step 4, monitoring the congestion, prompting the early warning information for the grid with the crowd density larger than delta, and timely taking intervention measures to dredge the congestion.
Embodiment one: crowd congestion prediction and prevention
With the development of society, large-scale activities are increased, and the risk of crowded trampling accidents caused by mass crowds is also improved obviously. The disaster is characterized by sudden occurrence, serious influence once occurrence and difficult rescue. Aiming at how to predict crowd congestion disasters and avoid trampling accidents, the disaster modeling and prediction technology based on the multi-task combined training can assist management staff in predicting time and place where the congestion disasters possibly happen in advance, and people can be arranged in advance to guide crowd movement, so that the risk of congestion trampling accidents caused by crowd gathering in a large amount is reduced.
As shown in fig. 2, crowd density distribution in a future time period is predicted by constructing a crowd motion simulation model through a travel track and a map, crowd density information of a local area is acquired through a monitoring camera and a millimeter wave sensor, crowd density distribution prediction results of the crowd motion simulation model are corrected in real time, and an alarm is sent to a manager when crowd congestion occurs in the map of the crowd motion simulation model. The manager can send out staff in advance to guide the crowd to move, so that crowd congestion and trampling risks are avoided.
Fig. 19 is an example of congestion prediction and prevention for a crowd motion simulation model. In the Unity map, the model predicts the time and the position of crowd congestion, so that the robot is arranged in advance to guide the crowd to move, and the crowd congestion accident in a period of time in the future of congestion is avoided.
Therefore, by executing the method of the invention, the congestion prediction is implemented in the subway station, and the position of the subway station where the congestion is about to occur is predicted in advance, so that the management personnel can conveniently take intervention measures.
Embodiment two: epidemic contact tracking, prevention and control
In the field of public health management, space-time associated personnel tracking and tracing of epidemic virus carriers are one of the key points and difficulties in implementing epidemic management. In order to reduce the management and control cost and improve the epidemic contactor screening efficiency, the invention is based on the disaster modeling and prevention technology based on sparse perception, can assist management staff to accurately screen space-time crossing people of epidemic virus carriers, and can implement accurate screening and improve the management efficiency.
As shown in fig. 1, the gate, the temperature and humidity sensor, the monitoring camera, the millimeter wave and the like provide local perception for pedestrians in the urban public space, and the complete track of the pedestrians in the whole environment is finally obtained by combining the inference of the perception blind areas.
Based on the complete trajectory of the pedestrian, an "epidemic contactor" is defined, and a query operation is performed. "epidemic contactor" means: in the same space, a time range t and a space distance range s are given, and for an abnormal pedestrian a, such as a confirmed epidemic virus carrier, other pedestrians meeting the time interval t and the space distance s on the passing path are called epidemic contactors of a, namely important attention crowd in epidemic prevention and control.
Fig. 16 is a true example of an epidemic contactor query Truepositive, i.e., referring to groudtluth information, and recovering the queried epidemic contactor through a trajectory, which is actually a pedestrian of a real epidemic contactor.
Fig. 17 is an example of an epidemic contactor query, i.e., referring to groudtluth information, by trace restoration of pedestrians among the queried epidemic contactors, who are not actually epidemic contactors.
Fig. 18 is an example of an epidemic contactor query, i.e., referring to groudtluth information, actually a real epidemic contactor, but pedestrians that are not queried by trajectory restoration.
The method and the device use track recovery for epidemic disease contact person tracking to search space-time intersection crowd of epidemic disease transmitters so as to realize fine and accurate epidemic disease contact person tracking, reduce management and control cost and improve management efficiency.
Embodiment III: urban public space disaster modeling and visualization
Driven by technologies such as AI, cloud computing, big data analysis, the intelligent security industry is becoming more and more open, and single video monitoring has not been able to meet diversified security service requirements. Aiming at the characteristics of various abnormal events in a subway station, a robust detection prediction model is designed based on fusion data to realize a sensing 'fine-fast-wide-fine' emergency abnormal condition monitoring and response mechanism, a visual interface is provided to assist staff in decision making, and harm caused by the abnormal events is reduced.
As shown in fig. 20, the in-station emergency early warning module carries out early warning prompt on abnormal events such as fire, harmful gas leakage, crowd behavior abnormality, local congestion, congestion prediction and the like. The emergency column displays the emergency occurrence time, the type of the abnormal event and the type of the sensor for identifying the abnormal event so as to allow staff to process the emergency in time.
In fig. 20 (a), a crowd congestion event that may occur in the future in the station is pre-warned, and the time and place at which the congestion time may occur, and the traffic data of the gate in the station according to the time and place are listed, so as to assist the staff in judging the crowd congestion risk in the station.
In fig. 20 (b), an alarm is given to open fire and dense smoke in the station, and specific occurrence places and detection time of the open fire and the dense smoke, and numbers of the cameras and the gas sensors are listed, so that staff can be assisted in timely finding and processing the fire in the station, and loss caused by the fire is reduced.
The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereto, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the present invention.
Claims (10)
1. The urban public space disaster modeling and preventing system based on the perception blind area estimation is characterized by comprising the following steps of:
step S1, obtaining local track information characterization in a sensing range of a sparse sensor to obtain a pedestrian track, namely a fragment track set, in the sensing range of each sensor;
s2, drawing an environment plan, and calculating the nearest point of each position to each obstacle according to the plan information for inquiring in the simulation process; constructing a social force model simulator, calculating social force born by each pedestrian, and carrying out path inquiry and intermediate target reselection by constructing a communication network diagram when the social force model fails; outputting a state sequence of each pedestrian, and processing to obtain a pedestrian track in a simulation time period; finally, according to the local track information in the step S1, parameters of the social force model simulator are adjusted;
step S3, segment track fusion: the segment track fusion problem is designed into a task allocation problem through a transfer matrix; then solving the maximum matching optimization problem by using a Hungary algorithm to obtain a fragment track fusion sequence;
s4, restoring the pedestrian track according to the fragment track fusion sequence obtained in the step S3; deducing the blind area track by using a social force model simulator, and generating a complete simulated pedestrian track according to the output of the simulator;
s5, detecting the crowd density of the local area in a specific period by using a crowd density detection model, and correcting the prediction result of the crowd movement track segment in the step S4; and counting the pedestrian density in the range of the perception dead zone according to the obtained pedestrian track, and realizing modeling and prediction of various urban public space disasters.
2. The urban public space disaster modeling and preventing system based on the perception blind area estimation according to claim 1, wherein the step S1 is specifically as follows:
step 101, constructing a training specific data set and manually marking;
step 102, selecting a single-sensor multi-target tracking frame, loading pre-trained parameters, and migrating to a marked training data set for training;
and 103, using the trained model for implementing a track recovery data set to acquire the track of the pedestrian in the sensing range of each sensor, namely a fragment track set.
3. The urban public space disaster modeling and prevention system based on the blind spot estimation according to claim 1, wherein in step 101, the data in the constructed data set comprises: the subway station comprises a camera monitoring video, a gate shooting record, millimeter wave sensor data and temperature and humidity sensor data.
4. The urban public space disaster modeling and preventing system based on the perception blind area estimation according to claim 1, wherein the step S2 is specifically as follows:
step 201, drawing an environment plan, rasterizing the environment plan and storing the environment plan as a dictionary;
step 202, calculating the nearest point of each position to each obstacle according to the geometric plan information, and storing the nearest point as a dictionary for inquiring in the simulation process;
step 203, constructing a connected network graph, wherein the vertexes of the network graph represent each grid in the position, no non-passable area exists on the grids, and if two vertexes are directly reachable, namely one grid is in the adjacent eight directions of the other grid, an edge exists between the two vertexes, and the connected graph is used for path inquiry and intermediate target reselection when the social force model fails;
step 204, firstly adding pedestrians in a new entering environment at each moment, then judging whether a blocked pedestrian exists, if the moving range of a certain pedestrian in the previous 5 steps is concentrated in a grid range, indicating that the pedestrian cannot normally travel based on the social force calculated by the current target, searching a shortest path, setting a new target, searching a coarse-precision shortest path by using dijkstra shortest path algorithm based on the established communication diagram, and then inserting a sampling position point in the shortest path into the new intermediate target;
step 205, calculating social force suffered by each pedestrian in the following way:
m i for the mass of the pedestrian i,
acceleration for pedestrian i->
Three social forces to which the pedestrian i is subjected are respectively represented: driving force, obstacle materials and interaction force;
the acceleration of each pedestrian is obtained through the calculation, so that the travelling speed of the current time step is obtained; before updating the position, judging whether the pedestrian enters an obstacle in the next step, if so, reducing the travelling speed according to the set attenuation degree, and if the pedestrian still cannot travel according to the attenuated speed after repeating the step for three times, setting the current speed to be 0;
step 206, repeating the steps 204 and 205 until the time steps are completed, finally outputting the state sequence of each pedestrian, and processing to obtain the pedestrian track in the simulation time period;
step 207, adjusting parameters of the social force model simulator according to the local track information acquired in the step S1.
5. The urban public space disaster modeling and prevention system based on the perception blind area estimation according to claim 1, wherein the three social forces in step 205 are specifically calculated as follows:
1) Driving force:
wherein τ i To be a relaxation time, the time required for the pedestrian to transition to an ideal speed is expressed;
2) Obstacle material resource Obstacle force:
d io (t) represents the distance from the pedestrian i to the obstacle o at time t;
3) Interaction force Interactive force:
r is the collision radius of the pedestrian.
6. The urban public space disaster modeling and preventing system based on the perception blind area estimation according to claim 1, wherein the step S3 is specifically as follows:
step 301: the segment track fusion problem is designed into a task allocation problem by designing a transfer matrix, and the transfer matrix is as follows:
wherein C is 1 Representing the transfer matrix between the two segment tracks,
C 2 a transition matrix representing a segment of the track as a track termination segment,
C 3 a transition matrix representing a segment of the track as an initial segment of the track,
step 302: calculating a transfer matrix for C 1 A section for calculating a similarity between the two segment trajectories for each element, the similarity being made up of two sections, one based on time and the other based on movement direction; for C 2 And C 3 Based on the prior information of the segment track, if the segment track k is the initial track segment, C (N+k)k = + infinity of the two points, if the segment track k is the initial track segment, then C k(N+k) = + infinity of the two points, N is the number of segment tracks;
step 303: and solving the maximum matching optimization problem by using the Hungary algorithm, and obtaining the track fragment fusion sequence result.
7. The urban public space disaster modeling and preventing system based on the perception blind area estimation according to claim 1, wherein the step S4 is specifically as follows:
step 401, an input file required for constructing a social force model comprises pedestrian entering conditions at all times, and an intermediate target point set of each pedestrian: obtaining all local known fragment track sequences of each pedestrian according to the track fusion result obtained in the step S3; setting some middle position points of the track fragment sequences as middle targets of pedestrians in the simulator, and guiding the simulation intelligent body to walk as far as possible according to the existing route, so as to truly restore the tracks of the pedestrians;
and step 402, executing a social force model simulator, and generating a simulated pedestrian track according to the output of the simulator to obtain a final recovered complete track.
8. The urban public space disaster modeling and preventing system based on the perception blind area estimation according to claim 1, wherein in step S5, modeling and predicting of various urban public space disasters is implemented, and the system specifically includes:
step 501, aiming at the characteristics of dense smoke and open fire accompanied by fire, acquiring an image data set containing the dense smoke and the open fire, expanding the fire data set by using a data augmentation method comprising Rotation, randomCrop, and constructing an image two-class model containing the open fire or smoke and not containing the open fire or smoke by using a deep convolution network so as to model the fire to obtain a fire detection model, thereby realizing accurate identification and alarm of images shot in a fire scene;
step 502, classifying the abnormal behaviors of pedestrians in the monitoring video into two types of abnormal objects and abnormal actions, collecting monitoring video data containing the abnormal behaviors of the pedestrians, taking the restored input video segments as targets to train an automatic encoder model, modeling the abnormal behaviors of the pedestrians in the monitoring video, and considering the abnormal behaviors of the pedestrians to occur when the automatic encoder model cannot accurately restore the input video segments; aiming at the characteristics of a part of scenes occurring in a fire scene, the abnormal degree of a video fragment is marked in an auxiliary mode by using a fire detection model, and the training of a pedestrian abnormal behavior detection model is guided by using a training framework of distillation learning, so that accurate time positioning and alarming of a pedestrian abnormal behavior event are realized;
step 503, aiming at the correlation between the crowd congestion anomaly and the crowd density in the urban public space, collecting monitoring video data containing pedestrians and constructing a crowd density monitoring model in the urban public space, and when the crowd density exceeds a specific threshold value, regarding the occurrence of congestion so as to model the congestion and trampling disasters; aiming at the characteristic that the extreme change of crowd density often accompanies abnormal pedestrian behaviors, a multi-task learning combined training pedestrian abnormal behavior detection model and a crowd density detection model are used; the multi-scale feature extraction and fusion crowd density estimation algorithm architecture of the encoder-decoder structure is used in combination with the cavity convolution method, and the multi-scale feature extraction and fusion crowd density estimation algorithm architecture consists of a feature extraction module based on multi-column convolution, a self-supervision training module based on the encoder-decoder and a density regression module based on the multi-task architecture, so that the crowd congestion abnormality in the monitoring video image is accurately detected together;
step 504: detecting the crowd density of the local area in a specific period by using the crowd density detection model obtained in the step 503, and correcting the prediction result of the crowd movement track segment in the step S4;
step 505: and (3) counting urban public space environments into grids of 2m multiplied by 2m, counting pedestrian densities in the environments, especially in the range of perception dead zones, according to the pedestrian tracks obtained in the step S4, monitoring the congestion, prompting early warning information for the grids with crowd densities larger than delta, and timely taking intervention measures to dredge the congestion.
9. The urban public space disaster modeling and prevention system based on the perception blind area estimation according to claim 4, wherein in step 201, the environmental plan is rasterized and stored as a dictionary, the keywords are represented by int-type digital symbols, k=x-x min +y-y min * h, a recorded value of 0 indicates that the location is passable, -1 indicates that the location is passableThe location is occupied by an obstacle or boundary.
10. The urban public space disaster modeling and prevention system based on the perception blind area estimation according to claim 4, wherein in step 207, parameters of the social force model simulator are adjusted, and the parameters are set as follows: desired speed v 0 ＝1.5，A o ＝1.0，A i ＝2.7。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609912.XA CN116090333A (en) | 2022-12-14 | 2022-12-14 | Urban public space disaster modeling and preventing system based on perception blind area estimation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609912.XA CN116090333A (en) | 2022-12-14 | 2022-12-14 | Urban public space disaster modeling and preventing system based on perception blind area estimation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116090333A true CN116090333A (en) | 2023-05-09 |
Family
ID=86200215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211609912.XA Pending CN116090333A (en) | 2022-12-14 | 2022-12-14 | Urban public space disaster modeling and preventing system based on perception blind area estimation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116090333A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116577671A (en) * | 2023-07-12 | 2023-08-11 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
| CN117074627A (en) * | 2023-10-16 | 2023-11-17 | 三科智能(山东)集团有限公司 | Medical laboratory air quality monitoring system based on artificial intelligence |
| CN117292548A (en) * | 2023-11-10 | 2023-12-26 | 腾讯科技（深圳）有限公司 | Traffic simulation method, device, equipment and storage medium |
-
2022
- 2022-12-14 CN CN202211609912.XA patent/CN116090333A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116577671A (en) * | 2023-07-12 | 2023-08-11 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
| CN116577671B (en) * | 2023-07-12 | 2023-09-29 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
| CN117074627A (en) * | 2023-10-16 | 2023-11-17 | 三科智能(山东)集团有限公司 | Medical laboratory air quality monitoring system based on artificial intelligence |
| CN117074627B (en) * | 2023-10-16 | 2024-01-09 | 三科智能(山东)集团有限公司 | Medical laboratory air quality monitoring system based on artificial intelligence |
| CN117292548A (en) * | 2023-11-10 | 2023-12-26 | 腾讯科技（深圳）有限公司 | Traffic simulation method, device, equipment and storage medium |
| CN117292548B (en) * | 2023-11-10 | 2024-02-09 | 腾讯科技（深圳）有限公司 | Traffic simulation method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022126669A1 (en) | Subway pedestrian flow network fusion method based on video pedestrian recognition, and pedestrian flow prediction method | |
| CN112868022B (en) | Driving scenario for an autonomous vehicle | |
| CN116090333A (en) | Urban public space disaster modeling and preventing system based on perception blind area estimation | |
| KR101995107B1 (en) | Method and system for artificial intelligence based video surveillance using deep learning | |
| CN114970321A (en) | Scene flow digital twinning method and system based on dynamic trajectory flow | |
| Fernando et al. | Deep inverse reinforcement learning for behavior prediction in autonomous driving: Accurate forecasts of vehicle motion | |
| Nasernejad et al. | Modeling pedestrian behavior in pedestrian-vehicle near misses: A continuous Gaussian Process Inverse Reinforcement Learning (GP-IRL) approach | |
| CN110008978A (en) | Classification of risks training method, classification of risks method, auxiliary or Automated Vehicle Operation system | |
| Wang et al. | STMAG: A spatial-temporal mixed attention graph-based convolution model for multi-data flow safety prediction | |
| CN108898520B (en) | Student safety monitoring method and system based on trajectory data | |
| CN112767644A (en) | Method and device for early warning of fire in highway tunnel based on video identification | |
| CN111079722A (en) | Hoisting process personnel safety monitoring method and system | |
| CN114372503A (en) | Cluster vehicle motion trail prediction method | |
| Basalamah et al. | Deep learning framework for congestion detection at public places via learning from synthetic data | |
| CN114913447A (en) | Police intelligent command room system and method based on scene recognition | |
| CN114494998A (en) | Intelligent analysis method and system for vehicle data | |
| Katariya et al. | A pov-based highway vehicle trajectory dataset and prediction architecture | |
| Thakur et al. | Graph (Graph): A Nested Graph-Based Framework for Early Accident Anticipation | |
| Ali et al. | Real-time safety monitoring vision system for linemen in buckets using spatio-temporal inference | |
| KR102614856B1 (en) | System and method for predicting risk of crowd turbulence | |
| Brax et al. | Finding behavioural anomalies in public areas using video surveillance data | |
| Qu et al. | Towards efficient traffic crash detection based on macro and micro data fusion on expressways: A digital twin framework | |
| CN114925994A (en) | Urban village risk assessment and risk factor positioning method based on deep learning | |
| Prezioso et al. | Integrating Object Detection and Advanced Analytics for Smart City Crowd Management | |
| Jiang et al. | Deploying scalable traffic prediction models for efficient management in real-world large transportation networks during hurricane evacuations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |