CN106781458A - A kind of traffic accident monitoring method and system - Google Patents
A kind of traffic accident monitoring method and system Download PDFInfo
- Publication number
- CN106781458A CN106781458A CN201611079545.1A CN201611079545A CN106781458A CN 106781458 A CN106781458 A CN 106781458A CN 201611079545 A CN201611079545 A CN 201611079545A CN 106781458 A CN106781458 A CN 106781458A
- Authority
- CN
- China
- Prior art keywords
- video
- traffic accident
- road conditions
- characteristic
- accident
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
Abstract
This application discloses a kind of traffic accident monitoring method, including:Obtain road conditions video;Extract the characteristic information of road conditions video;Wherein, the characteristic information of road conditions video includes visual dictionary, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and the temporal characteristics of road conditions video;The traffic accident prediction model that the characteristic vector input that will be generated according to the characteristic information of road conditions video is pre-created, obtains the traffic accident probability of happening exported by traffic accident prediction model;Wherein, traffic accident prediction model is the model obtained based on machine learning algorithm;If traffic accident probability of happening is more than predetermined probabilities threshold value, corresponding incident response mechanism is triggered.The application realizes the effective monitoring to traffic accident.In addition, the application further correspondingly discloses a kind of system for monitoring traffic accident.
Description
Technical field
The present invention relates to traffic monitoring technical field, more particularly to a kind of traffic accident monitoring method and system.
Background technology
At present, as economic fast development, the car ownership in city are continuously increased, the day in present many cities is caused
Normal traffic becomes increasingly congestion, and traffic accident constantly takes place frequently, and the security of the lives and property for giving people causes extreme
Influence.
In order to be reduced as far as traffic accident quantity and reduce the life and property loss that traffic accident is triggered, having must
Monitor in real time is carried out to the traffic accident being likely to occur on road.And it is current for how realizing to the effective monitoring of traffic accident
Problem also to be solved.
The content of the invention
In view of this, it is an object of the invention to provide a kind of traffic accident monitoring method and system, realize to traffic
The effective monitoring of accident.Its concrete scheme is as follows:
A kind of traffic accident monitoring method, including:
Obtain road conditions video;
Extract the characteristic information of the road conditions video;Wherein, the characteristic information of the road conditions video is regarded including the road conditions
The visual dictionary of frequency, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and temporal characteristics;
The traffic accident that the characteristic vector input that will be generated according to the characteristic information of the road conditions video is pre-created is estimated
Model, obtains the traffic accident probability of happening exported by the traffic accident prediction model;Wherein, mould is estimated in the traffic accident
Type is the model obtained based on machine learning algorithm;
If the traffic accident probability of happening is more than predetermined probabilities threshold value, corresponding incident response mechanism is triggered.
Preferably, the process of the visual dictionary of the road conditions video is extracted, including:
Using default image characteristic point extraction algorithm, the feature in the extraction road conditions video on vehicle region
Point, obtains corresponding vehicle vision lexical set;
The visual dictionary model that each vocabulary in the vehicle vision lexical set is respectively mapped to be pre-created
In corresponding cluster centre, obtain the visual dictionary of the road conditions video.
Preferably, the establishment process of the visual dictionary model, including:
Obtain video image sample collection;Wherein, the video image sample collection include respectively in different acquisition parameters and
Under different shooting environmentals, the video image for different types of vehicle obtained after video image acquisition;Acquisition parameters bag
Shooting visual angle and shooting time section are included, shooting environmental includes foggy environment, sleet environment, fine day environment and Sand Dust Environment;
Using described image feature point extraction algorithm, each video image sample that the video image sample is concentrated is entered
The treatment of row feature point extraction, is correspondingly made available the vehicle vision lexical set corresponding to each video image sample;
Using K-means clustering algorithms, the vehicle vision lexical set corresponding to each video image sample is entered successively
Row clustering processing, obtains the visual dictionary model.
Preferably, the wagon flow feature of the road conditions video include vehicle flowrate, flow speeds, wagon flow occupation rate, lane-change frequency,
Rate of acceleration change and maximum angular rate;
Driver's state feature of the road conditions video includes tired driver degree;
The roadway characteristic of the road conditions video rubs including traffic lights setting situation, road grade, road curvature degree and road
Wipe the factor;
The environmental characteristic of the road conditions video includes road surface foreign matter situation and weather conditions;
The temporal characteristics of the road conditions video include video capture date and video capture moment.
Preferably, the process of the corresponding incident response mechanism of triggering, including:
Produce the first control instruction;
The positional information of first control instruction and accident section is sent to the unmanned plane of advance binding, to utilize
First control instruction, controls the self-contained picture harvester of the unmanned plane unlatching and flight to the accident road
The upper dummy section of section;
Obtain the accident section real-time pictures collected by the picture harvester that the unmanned plane sends.
Preferably, the process of the corresponding incident response mechanism of triggering, including:
Produce the second control instruction;
Second control instruction is sent to the warning device for being set on accident section in advance, to utilize described second
Control instruction, controls the warning device to send corresponding warning information.
Preferably, the establishment process of the traffic accident prediction model, including:
Obtain training set;Wherein, the training set includes positive sample training set and negative sample training set;The positive sample instruction
Practicing collection includes N number of positive sample video, and the negative sample training set includes M negative sample video, and N and M are positive integer;Positive sample
Video is the video comprising traffic accident picture, and negative sample video is the video not comprising traffic accident picture;
The characteristic information of each video in the training set, and the feature letter according to each video respectively are extracted respectively
The corresponding characteristic vector of breath generation, obtains corresponding set of eigenvectors;
Using the machine learning algorithm, learning training is carried out to each characteristic vector that the characteristic vector is concentrated, obtained
To the traffic accident prediction model.
The invention also discloses a kind of system for monitoring traffic accident, including:
Video acquisition device, for obtaining road conditions video;
Characteristic information extracting module, the characteristic information for extracting the road conditions video;Wherein, the spy of the road conditions video
Reference breath include the visual dictionary of the road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and when
Between feature;
Feature vector generation module, for the characteristic information according to the road conditions video, generates corresponding characteristic vector;
Model creation module, traffic accident prediction model is created for being in advance based on machine learning algorithm;
Accident estimates module, and the characteristic vector for the feature vector generation module to be generated is input into the traffic accident
Prediction model, obtains the traffic accident probability of happening exported by the traffic accident prediction model;Wherein, the traffic accident is pre-
It is the model obtained based on machine learning algorithm to estimate model;
Incident response module, for being more than predetermined probabilities threshold value when the traffic accident probability of happening, then triggers corresponding
Incident response mechanism.
Preferably, the incident response module, including:
First instruction generation unit, for producing the first control instruction;
Control information transmitting element, for the positional information of first control instruction and accident section to be sent to pre-
The unmanned plane first bound, to utilize first control instruction, controls the unmanned plane to open self-contained picture collection dress
Put and fly to the upper dummy section of the accident section;
Image information acquiring unit, for obtaining the thing collected by the picture harvester that the unmanned plane sends
Therefore section real-time pictures.
Preferably, the model creation module, including:
Training set acquisition submodule, for obtaining training set;Wherein, the training set includes positive sample training set and negative sample
This training set;The positive sample training set includes N number of positive sample video, and the negative sample training set includes M negative sample video,
N and M are positive integer;Positive sample video is the video comprising traffic accident picture, and negative sample video is not comprising traffic accident
The video of picture;
Video processing submodule, the characteristic information for extracting each video in the training set respectively, and root respectively
Corresponding characteristic vector is generated according to the characteristic information of each video, corresponding set of eigenvectors is obtained;
Training submodule, for utilizing the machine learning algorithm, each characteristic vector concentrated to the characteristic vector
Learning training is carried out, the traffic accident prediction model is obtained.
In the present invention, traffic accident monitoring method, including:Obtain road conditions video;Extract the characteristic information of road conditions video;Its
In, the characteristic information of road conditions video includes visual dictionary, wagon flow feature, driver's state feature, roadway characteristic, the ring of road conditions video
Border feature and temporal characteristics;The traffic accident that the characteristic vector input that will be generated according to the characteristic information of road conditions video is pre-created
Prediction model, obtains the traffic accident probability of happening exported by traffic accident prediction model;Wherein, traffic accident prediction model is
Based on the model that machine learning algorithm is obtained;If traffic accident probability of happening is more than predetermined probabilities threshold value, corresponding thing is triggered
Therefore response mechanism.
It can be seen that, the present invention is in advance based on machine learning algorithm and creates traffic accident prediction model, is regarded road conditions are got
After frequency, it will to the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and time
, then be input into the characteristic vector generated according to these characteristic informations to above-mentioned traffic accident prediction model, so as to obtain by feature
Corresponding traffic accident probability of happening, based on the traffic accident probability of happening, it may be determined that whether have needs to start corresponding accident
Response mechanism.Due to the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and when
Between feature can objective, comprehensively reflect the situation of current road conditions traffic, may be such that and estimated eventually through above-mentioned traffic accident
Traffic accident probability of happening obtained by model can relatively accurately reflect under current road conditions the possibility of traffic accident
Property.As can be seen here, the present invention realizes the effective monitoring to traffic accident.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Inventive embodiment, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
The accompanying drawing of offer obtains other accompanying drawings.
Fig. 1 is a kind of traffic accident monitoring method flow chart disclosed in the embodiment of the present invention;
Fig. 2 is visual dictionary model creation method flow diagram disclosed in the embodiment of the present invention;
Fig. 3 is traffic accident prediction model creation method flow chart disclosed in the embodiment of the present invention;
Fig. 4 is a kind of system for monitoring traffic accident structural representation disclosed in the embodiment of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
The embodiment of the invention discloses a kind of traffic accident monitoring method, shown in Figure 1, the method includes:
Step S11:Obtain road conditions video.
Specifically, the present embodiment can carry out reality using the video camera by road to the road conditions on present road
When video pictures collection, obtain above-mentioned road conditions video.
Step S12:Extract the characteristic information of road conditions video;Wherein, the characteristic information of road conditions video includes road conditions video
Visual dictionary, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and temporal characteristics.
It is understood that before the characteristic information of above-mentioned road conditions video is extracted, can also enter to above-mentioned road conditions video
Row pretreatment, the preprocessing process includes but is not limited to the treatment of video pictures scaling, video pictures noise reduction process and video image quality
Enhancing is processed.
In addition, it is necessary to explanation, in the present embodiment, the visual vocabulary in above-mentioned visual dictionary is specially in road conditions video
Vehicle characteristics point information on vehicle region.
Step S13:The traffic accident that the characteristic vector input that will be generated according to the characteristic information of road conditions video is pre-created
Prediction model, obtains the traffic accident probability of happening exported by traffic accident prediction model;Wherein, traffic accident prediction model is
Based on the model that machine learning algorithm is obtained.
That is, current embodiment require that using visual dictionary, wagon flow feature, the driver's state extracted in above-mentioned steps S12
Feature, roadway characteristic, environmental characteristic and temporal characteristics generate corresponding characteristic vector, are then input into this feature vector
To being in advance based on the traffic accident prediction model that machine learning algorithm is obtained.
In the present embodiment, the machine learning algorithm in above-mentioned steps S13 preferentially uses SVM algorithm (SVM, i.e. Support
Vector Machine)。
Step S14:If traffic accident probability of happening is more than predetermined probabilities threshold value, corresponding incident response mechanism is triggered.
In the present embodiment, above-mentioned predetermined probabilities threshold value is specifically as follows the first probability threshold value, when gained in above-mentioned steps S13
The traffic accident probability of happening for arriving is more than above-mentioned first probability threshold value, then mean that the traffic on present road compares evil
It is bad, if not carrying out traffic intervention, it is likely that traffic accident occurs, can now trigger accident early warning response mechanism.
Certainly, above-mentioned predetermined probabilities threshold value is specifically as follows the second probability threshold value, when friendship resulting in above-mentioned steps S13
Logical contingency occurrence probability is more than above-mentioned second probability threshold value, then it can be assumed that having occurred in that traffic accident on present road, this
When can trigger accident post processing response mechanism.
It is understood that above-mentioned second probability threshold value is more than above-mentioned first probability threshold value, the two probability threshold values are specific
Can be set according to actual conditions needs, it not limited specifically herein.
It can be seen that, the embodiment of the present invention is in advance based on machine learning algorithm and creates traffic accident prediction model, is getting
After road conditions video, it will to the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic
And temporal characteristics, then the characteristic vector generated according to these characteristic informations is input into above-mentioned traffic accident prediction model, from
And corresponding traffic accident probability of happening is obtained, based on the traffic accident probability of happening, it may be determined that go out whether to need to start accordingly
Incident response mechanism.Because the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environment are special
Temporal characteristics of seeking peace can objective, comprehensively reflect the situation of current road conditions traffic, may be such that eventually through above-mentioned traffic thing
Therefore the traffic accident probability of happening obtained by prediction model can relatively accurately reflect that traffic accident occurs under current road conditions
Possibility.As can be seen here, the embodiment of the present invention realizes the effective monitoring to traffic accident.
The embodiment of the invention discloses a kind of specific traffic accident monitoring method, relative to a upper embodiment, this implementation
Example has made further instruction and optimization to technical scheme.Specifically:
, it is necessary to be extracted to the characteristic information in road conditions video in upper embodiment step S12.Specifically, this implementation
In example, the process of the visual dictionary of said extracted road conditions video specifically includes below step S121 and S122:
Step S121:Using default image characteristic point extraction algorithm, in extraction road conditions video on vehicle region
Characteristic point, obtains corresponding vehicle vision lexical set.
It is understood that the characteristic point in above-mentioned road conditions video on vehicle region constitutes regarding for the road conditions video
Feel vocabulary, all characteristic points in the road conditions video on vehicle region constitute the corresponding vehicle vision of above-mentioned road conditions video
Lexical set.
Step S122:Each vocabulary in vehicle vision lexical set is respectively mapped to the visual dictionary being pre-created
Corresponding cluster centre in model, obtains the visual dictionary of road conditions video.
That is, according to cluster centre corresponding in above-mentioned visual dictionary model, in above-mentioned vehicle vision lexical set
Each vocabulary carry out classification treatment, the vocabulary frequency in then counting per class vocabulary, so as to obtain and above-mentioned road conditions video pair
The visual dictionary answered.It is understood that above-mentioned visual dictionary reflects the characteristic information of vehicle in road conditions video.
In addition, shown in Figure 2, the establishment process of above-mentioned visual dictionary model, specifically includes:
Step S21:Obtain video image sample collection;Wherein, video image sample collection includes joining in different shootings respectively
Under number and different shooting environmentals, the video image for different types of vehicle obtained after video image acquisition;Shoot ginseng
Number includes shooting visual angle and shooting time section, and shooting environmental includes foggy environment, sleet environment, fine day environment and Sand Dust Environment;
Step S22:Using image characteristic point extraction algorithm, each video image sample that video image sample is concentrated is entered
The treatment of row feature point extraction, is correspondingly made available the vehicle vision lexical set corresponding to each video image sample;
Step S23:Using K-means clustering algorithms, successively to the vehicle vision word corresponding to each video image sample
Collecting conjunction carries out clustering processing, obtains visual dictionary model.
In the present embodiment, above-mentioned image characteristic point extraction algorithm is specifically as follows SIFT algorithms (SIFT, i.e. Scale-
Invariant Feature Transform, Scale invariant features transform) or SURF algorithm (SURF, i.e. Speeded Up
Robust Features, accelerate robust feature).
In upper embodiment step S12, the visual dictionary of the characteristic information including road conditions video of the road conditions video for being extracted,
Wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and temporal characteristics.
Specifically, the wagon flow feature of above-mentioned road conditions video include vehicle flowrate, flow speeds, wagon flow occupation rate, lane-change frequency,
Rate of acceleration change and maximum angular rate.
Wherein, above-mentioned vehicle flowrate refers to the driving vehicle number for detecting section by certain in a certain special time period;On
It refers to each arithmetic mean of instantaneous value of vehicle speed in road conditions video to state flow speeds;Above-mentioned rate of acceleration change refers in a certain spy
The variance yields of vehicle acceleration in section of fixing time;Above-mentioned maximum angular rate refers to pass through road most in a certain special time period
Big angular speed;Above-mentioned wagon flow occupation rate refers to the ratio that vehicle occupies length summation and road surface length, it is contemplated that cannot be in section
It is upper using direct method come the summation of measuring vehicle length, the embodiment of the present invention specifically can indirectly be calculated using time quantum
Above-mentioned wagon flow occupation rate, specific computing formula is:
In formula, FoccupyRepresent wagon flow occupation rate, ToccupyRepresent the time that road is occupied by vehicle, TallRepresent total observation
Time.
In addition, above-mentioned lane-change frequency refers to the average lane-change number of times of the vehicle in a certain special time period, it is corresponding to calculate
Formula is:
In formula, FchannelRepresent lane-change frequency, nchannelRepresent lane-change total degree, FnumRepresent that the vehicle in road conditions video is total
Number.
In the present embodiment, driver's state feature of above-mentioned road conditions video includes tired driver degree.Specifically, can be using department
The blink frequency of machine is used as the parameter for weighing tired driver degree.
In the present embodiment, the roadway characteristic of above-mentioned road conditions video includes traffic lights setting situation, road grade, road curvature
Degree and the road friction factor.Specifically, the traffic lights that the present embodiment can weigh road by determining whether traffic lights sets
Put situation;Above-mentioned road grade can be obtained by using the height of road divided by the horizontal range of road;Above-mentioned road bend
Curvature is embodied as on road direction the bending action corresponding to every meter of length;The above-mentioned road friction factor is specially tire
Coefficient of friction between road surface.
In the present embodiment, the environmental characteristic of above-mentioned road conditions video includes road surface foreign matter situation and weather conditions.Wherein, it is above-mentioned
The rubbish that road surface foreign matter situation can specifically be expressed as road occupies the ratio of length and road total length;Above-mentioned weather conditions bag
Include greasy weather, sleety weather, fine day and dust and sand weather etc..
In the present embodiment, the temporal characteristics of above-mentioned road conditions video include video capture date and video capture moment.
In upper embodiment step S14, in the case where traffic accident probability of happening is more than predetermined probabilities threshold value, can trigger
Corresponding incident response mechanism.Specifically, in the present embodiment, the process of the above-mentioned corresponding incident response mechanism of triggering, including under
Face step S141 to S143:
Step S141:Produce the first control instruction;
Step S142:The positional information of the first control instruction and accident section is sent to the unmanned plane of advance binding,
To utilize the first control instruction, it is upper to accident section that control unmanned plane opens self-contained picture harvester and flight
Dummy section;
Step S143:Obtain the accident section real-time pictures collected by picture harvester that unmanned plane sends.
It is understood that incident response mechanism corresponding in above-mentioned steps S141 to S143 can be specifically to judge road
Road has occurred what is launched after the situation of traffic accident, it is of course also possible to be after determining and traffic accident occurring
Launch.
It is further noted that after the situation that traffic accident has occurred in road is judged, it is possible to use above-mentioned
The accident section real-time pictures that above-mentioned unmanned plane sends are analyzed by manual identified method or deep learning algorithm, are worked as
Traffic accident grade on preceding road.
In the present embodiment, traffic accident grade can be divided into:Ordinary accident, it is meant that vehicle has slight damage;Compared with
Major break down, it is meant that vehicle deformation is more serious, the personnel in vehicle can normally take action;Major accident, it is meant that in vehicle
Personnel it is injured, and bleeding is serious;Special major accident, it is meant that the personnel in vehicle are stranded and in-car.
Further, the process of the corresponding incident response mechanism of above-mentioned triggering, can also include:The second control instruction is produced,
Then the second control instruction is sent to the warning device for being set on accident section in advance, to utilize the second control instruction, control
Warning device processed sends corresponding warning information.
It should be noted that above-mentioned warning device can be pre-set on traffic signals bar, it is also possible to be arranged on accident
The both sides in section, for example, be arranged on the street lamp post of road both sides.The quantity of above-mentioned warning device can be one, or
More than one.The type of above-mentioned warning device can be specifically audible-visual annunciator, it is of course also possible to use phonetic alarm.
, it is necessary to determine current friendship using the traffic accident prediction model being pre-created in upper embodiment step S13
Logical contingency occurrence probability.Shown in Figure 3, the establishment process of above-mentioned traffic accident prediction model can specifically include:
Step S131:Obtain training set;Wherein, training set includes positive sample training set and negative sample training set;Positive sample
Training set includes N number of positive sample video, and negative sample training set includes M negative sample video, and N and M are positive integer;Positive sample is regarded
Frequency is the video comprising traffic accident picture, and negative sample video is the video not comprising traffic accident picture.
It should be noted that in order to the phenomenon of data skew occur, above-mentioned positive sample can be ensured as far as possible in the present embodiment
Sample size in training set and negative sample training set is same or about, that is, ensureing that above-mentioned N values are identical with M values as far as possible
Or it is close.
Step S132:The characteristic information of each video in training set is extracted respectively, and respectively according to each video
Characteristic information generates corresponding characteristic vector, obtains corresponding set of eigenvectors.
Wherein, above-mentioned steps S132 detailed processes of the characteristic information of each video in training set is extracted are implemented with upper one
The detailed process of the characteristic information of the extraction road conditions video in example step S12 is similar, is no longer repeated herein.
Step S133:Using machine learning algorithm, learning training is carried out to each characteristic vector that characteristic vector is concentrated, obtained
To traffic accident prediction model.
Wherein, above-mentioned machine learning algorithm is specifically as follows SVM algorithm.
Accordingly, it is shown in Figure 4 the embodiment of the invention also discloses a kind of system for monitoring traffic accident, the system bag
Include:
Video acquisition device 11, for obtaining road conditions video;
Characteristic information extracting module 12, the characteristic information for extracting road conditions video;Wherein, the characteristic information of road conditions video
Visual dictionary including road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and temporal characteristics;
Feature vector generation module 13, for the characteristic information according to road conditions video, generates corresponding characteristic vector;
Model creation module 14, traffic accident prediction model is created for being in advance based on machine learning algorithm;
Accident estimates module 15, and the characteristic vector input traffic accident for feature vector generation module 13 to be generated is estimated
Model, obtains the traffic accident probability of happening exported by traffic accident prediction model;Wherein, traffic accident prediction model be based on
The model that machine learning algorithm is obtained;
Incident response module 16, for being more than predetermined probabilities threshold value when traffic accident probability of happening, then triggers corresponding thing
Therefore response mechanism.
It can be seen that, the embodiment of the present invention is in advance based on machine learning algorithm and creates traffic accident prediction model, is getting
After road conditions video, it will to the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic
And temporal characteristics, then the characteristic vector generated according to these characteristic informations is input into above-mentioned traffic accident prediction model, from
And corresponding traffic accident probability of happening is obtained, based on the traffic accident probability of happening, it may be determined that go out whether to need to start accordingly
Incident response mechanism.Because the visual dictionary in road conditions video, wagon flow feature, driver's state feature, roadway characteristic, environment are special
Temporal characteristics of seeking peace can objective, comprehensively reflect the situation of current road conditions traffic, may be such that eventually through above-mentioned traffic thing
Therefore the traffic accident probability of happening obtained by prediction model can relatively accurately reflect that traffic accident occurs under current road conditions
Possibility.As can be seen here, the embodiment of the present invention realizes the effective monitoring to traffic accident.
In the present embodiment, features described above information extraction modules can specifically include visual vocabulary extracting sub-module, dictionary mould
Type creates submodule and visual dictionary generation submodule;Wherein,
Visual vocabulary extracting sub-module, for utilizing default image characteristic point extraction algorithm, extracts car in road conditions video
Characteristic point on region, obtains corresponding vehicle vision lexical set;
Dictionary model creates submodule, for being pre-created visual dictionary model;
Visual dictionary generates submodule, above-mentioned for each vocabulary in vehicle vision lexical set to be respectively mapped to
Corresponding cluster centre in visual dictionary model, obtains the visual dictionary of road conditions video.
In the present embodiment, above-mentioned dictionary model creates submodule and specifically includes sample set acquiring unit, feature point extraction list
Unit and words clustering unit;Wherein,
Sample set acquiring unit, for obtaining video image sample collection;Wherein, video image sample collection is included respectively not
Under same acquisition parameters and different shooting environmentals, the video figure for different types of vehicle obtained after video image acquisition
Picture;Acquisition parameters include shooting visual angle and shooting time section, shooting environmental include foggy environment, sleet environment, fine day environment and
Sand Dust Environment;
Feature point extraction unit, for utilizing image characteristic point extraction algorithm, each the regarding to video image sample concentration
Frequency image pattern carries out feature point extraction treatment, is correspondingly made available the vehicle vision word finder corresponding to each video image sample
Close;
Words clustering unit, for utilizing K-means clustering algorithms, successively to the car corresponding to each video image sample
Visual vocabulary set carries out clustering processing, obtains visual dictionary model.
In the present embodiment, above-mentioned image characteristic point extraction algorithm is specifically as follows SIFT algorithms or SURF algorithm.
Further, above-mentioned incident response module can specifically include that the first instruction generation unit, control information send list
Unit and image information acquiring unit;Wherein,
First instruction generation unit, for producing the first control instruction;
Control information transmitting element, for the positional information of the first control instruction and accident section to be sent to tying up in advance
Fixed unmanned plane, to utilize the first control instruction, the self-contained picture harvester of control unmanned plane unlatching and flight are extremely
The upper dummy section of accident section;
Image information acquiring unit, the accident section reality collected by picture harvester for obtaining unmanned plane transmission
When picture.
It is understood that above-mentioned obtain single by the first instruction generation unit, control information transmitting element and image information
The incident response mechanism that is triggered of unit can specifically judge that launching after the situation of traffic accident has occurred in road, when
So, or determine traffic accident will occurs after launch.
It is further noted that after the situation that traffic accident has occurred in road is judged, thing can also be included
Therefore grade acquisition module, for by above-mentioned manual identified method or deep learning algorithm, the accident sent to above-mentioned unmanned plane
Section real-time pictures are analyzed, and obtain the traffic accident grade on present road.
In the present embodiment, traffic accident grade can be divided into:Ordinary accident, it is meant that vehicle has slight damage;Compared with
Major break down, it is meant that vehicle deformation is more serious, the personnel in vehicle can normally take action;Major accident, it is meant that in vehicle
Personnel it is injured, and bleeding is serious;Special major accident, it is meant that the personnel in vehicle are stranded and in-car.
Certainly, above-mentioned incident response module can also specifically include the second instruction generation unit, for producing second to control
Instruction, then sends to the warning device for being set on accident section in advance the second control instruction, refers to using the second control
Order, control warning device sends corresponding warning information.
It should be noted that above-mentioned warning device can be pre-set on traffic signals bar, it is also possible to be arranged on accident
The both sides in section, for example, be arranged on the street lamp post of road both sides.The quantity of above-mentioned warning device can be one, or
More than one.The type of above-mentioned warning device can be specifically audible-visual annunciator, it is of course also possible to use phonetic alarm.
In addition, in the present embodiment, above-mentioned model creation module specifically includes training set acquisition submodule, Video processing
Module and training submodule;Wherein,
Training set acquisition submodule, for obtaining training set;Wherein, training set includes that positive sample training set and negative sample are instructed
Practice collection;Positive sample training set includes N number of positive sample video, and negative sample training set includes M negative sample video, and N and M is just whole
Number;Positive sample video is the video comprising traffic accident picture, and negative sample video is the video not comprising traffic accident picture;
Video processing submodule, the characteristic information for extracting each video in training set respectively, and respectively according to every
One characteristic information of video generates corresponding characteristic vector, obtains corresponding set of eigenvectors;
Training submodule, for utilizing machine learning algorithm, learns to each characteristic vector that characteristic vector is concentrated
Training, obtains traffic accident prediction model.
It should be noted that in order to the phenomenon of data skew occur, above-mentioned positive sample can be ensured as far as possible in the present embodiment
Sample size in training set and negative sample training set is same or about, that is, ensureing that above-mentioned N values are identical with M values as far as possible
Or it is close.
Wherein, above-mentioned machine learning algorithm is specifically as follows SVM algorithm.
Finally, in addition it is also necessary to explanation, herein, such as first and second or the like relational terms be used merely to by
One entity or operation make a distinction with another entity or operation, and not necessarily require or imply these entities or operation
Between there is any this actual relation or order.And, term " including ", "comprising" or its any other variant meaning
Covering including for nonexcludability, so that process, method, article or equipment including a series of key elements not only include that
A little key elements, but also other key elements including being not expressly set out, or also include for this process, method, article or
The intrinsic key element of equipment.In the absence of more restrictions, the key element limited by sentence "including a ...", does not arrange
Except also there is other identical element in the process including the key element, method, article or equipment.
A kind of traffic accident monitoring method provided by the present invention and system are described in detail above, herein should
Principle of the invention and implementation method are set forth with specific case, the explanation of above example is only intended to help and manages
The solution method of the present invention and its core concept;Simultaneously for those of ordinary skill in the art, according to thought of the invention,
Be will change in specific embodiment and range of application, in sum, this specification content should not be construed as to this hair
Bright limitation.
Claims (10)
1. a kind of traffic accident monitoring method, it is characterised in that including:
Obtain road conditions video;
Extract the characteristic information of the road conditions video;Wherein, the characteristic information of the road conditions video includes the road conditions video
Visual dictionary, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and temporal characteristics;
The traffic accident prediction model that the characteristic vector input that will be generated according to the characteristic information of the road conditions video is pre-created,
Obtain the traffic accident probability of happening exported by the traffic accident prediction model;Wherein, the traffic accident prediction model is
Based on the model that machine learning algorithm is obtained;
If the traffic accident probability of happening is more than predetermined probabilities threshold value, corresponding incident response mechanism is triggered.
2. traffic accident monitoring method according to claim 1, it is characterised in that extract the visual word of the road conditions video
The process of allusion quotation, including:
Using default image characteristic point extraction algorithm, the characteristic point in the extraction road conditions video on vehicle region is obtained
To corresponding vehicle vision lexical set;
Each vocabulary in the vehicle vision lexical set is respectively mapped to phase in the visual dictionary model being pre-created
Corresponding cluster centre, obtains the visual dictionary of the road conditions video.
3. traffic accident monitoring method according to claim 2, it is characterised in that the establishment of the visual dictionary model
Journey, including:
Obtain video image sample collection;Wherein, the video image sample collection is included respectively in different acquisition parameters and difference
Shooting environmental under, the video image for different types of vehicle obtained after video image acquisition;Acquisition parameters include clapping
Visual angle and shooting time section are taken the photograph, shooting environmental includes foggy environment, sleet environment, fine day environment and Sand Dust Environment;
Using described image feature point extraction algorithm, spy is carried out to each video image sample that the video image sample is concentrated
An extraction process is levied, the vehicle vision lexical set corresponding to each video image sample is correspondingly made available;
Using K-means clustering algorithms, the vehicle vision lexical set corresponding to each video image sample is gathered successively
Class treatment, obtains the visual dictionary model.
4. traffic accident monitoring method according to claim 1, it is characterised in that
The wagon flow feature of the road conditions video includes vehicle flowrate, flow speeds, wagon flow occupation rate, lane-change frequency, acceleration change
Rate and maximum angular rate;
Driver's state feature of the road conditions video includes tired driver degree;
The roadway characteristic of the road conditions video include traffic lights setting situation, road grade, road curvature degree and road friction because
Son;
The environmental characteristic of the road conditions video includes road surface foreign matter situation and weather conditions;
The temporal characteristics of the road conditions video include video capture date and video capture moment.
5. traffic accident monitoring method according to claim 1, it is characterised in that the corresponding incident response machine of triggering
The process of system, including:
Produce the first control instruction;
The positional information of first control instruction and accident section is sent to the unmanned plane of advance binding, with using described
First control instruction, controls the self-contained picture harvester of the unmanned plane unlatching and flight to the accident section
Upper dummy section;
Obtain the accident section real-time pictures collected by the picture harvester that the unmanned plane sends.
6. traffic accident monitoring method according to claim 1, it is characterised in that the corresponding incident response machine of triggering
The process of system, including:
Produce the second control instruction;
Second control instruction is sent to the warning device for being set on accident section in advance, is controlled with using described second
Instruction, controls the warning device to send corresponding warning information.
7. the traffic accident monitoring method according to any one of claim 1 to 6, it is characterised in that the traffic accident is pre-
Estimate the establishment process of model, including:
Obtain training set;Wherein, the training set includes positive sample training set and negative sample training set;The positive sample training set
Including N number of positive sample video, the negative sample training set includes M negative sample video, and N and M are positive integer;Positive sample video
It is the video comprising traffic accident picture, negative sample video is the video not comprising traffic accident picture;
The characteristic information of each video in the training set, and the characteristic information life according to each video respectively are extracted respectively
Into corresponding characteristic vector, corresponding set of eigenvectors is obtained;
Using the machine learning algorithm, learning training is carried out to each characteristic vector that the characteristic vector is concentrated, obtain institute
State traffic accident prediction model.
8. a kind of system for monitoring traffic accident, it is characterised in that including:
Video acquisition device, for obtaining road conditions video;
Characteristic information extracting module, the characteristic information for extracting the road conditions video;Wherein, the feature letter of the road conditions video
Breath includes that visual dictionary, wagon flow feature, driver's state feature, roadway characteristic, environmental characteristic and the time of the road conditions video are special
Levy;
Feature vector generation module, for the characteristic information according to the road conditions video, generates corresponding characteristic vector;
Model creation module, traffic accident prediction model is created for being in advance based on machine learning algorithm;
Accident estimates module, and the characteristic vector for the feature vector generation module to be generated is input into the traffic accident and estimates
Model, obtains the traffic accident probability of happening exported by the traffic accident prediction model;Wherein, mould is estimated in the traffic accident
Type is the model obtained based on machine learning algorithm;
Incident response module, for being more than predetermined probabilities threshold value when the traffic accident probability of happening, then triggers corresponding accident
Response mechanism.
9. system for monitoring traffic accident according to claim 8, it is characterised in that the incident response module, including:
First instruction generation unit, for producing the first control instruction;
Control information transmitting element, for the positional information of first control instruction and accident section to be sent to tying up in advance
Fixed unmanned plane, to utilize first control instruction, control the unmanned plane open self-contained picture harvester with
And fly to the upper dummy section of the accident section;
Image information acquiring unit, for obtaining the accident road collected by the picture harvester that the unmanned plane sends
Section real-time pictures.
10. system for monitoring traffic accident according to claim 8 or claim 9, it is characterised in that the model creation module, bag
Include:
Training set acquisition submodule, for obtaining training set;Wherein, the training set includes that positive sample training set and negative sample are instructed
Practice collection;The positive sample training set includes N number of positive sample video, and the negative sample training set includes M negative sample video, N and M
It is positive integer;Positive sample video is the video comprising traffic accident picture, and negative sample video is not comprising traffic accident picture
Video;
Video processing submodule, the characteristic information for extracting each video in the training set respectively, and respectively according to every
One characteristic information of video generates corresponding characteristic vector, obtains corresponding set of eigenvectors;
Training submodule, for utilizing the machine learning algorithm, is carried out to each characteristic vector that the characteristic vector is concentrated
Learning training, obtains the traffic accident prediction model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611079545.1A CN106781458B (en) | 2016-11-30 | 2016-11-30 | A kind of traffic accident monitoring method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611079545.1A CN106781458B (en) | 2016-11-30 | 2016-11-30 | A kind of traffic accident monitoring method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106781458A true CN106781458A (en) | 2017-05-31 |
CN106781458B CN106781458B (en) | 2019-10-18 |
Family
ID=58898959
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611079545.1A Active CN106781458B (en) | 2016-11-30 | 2016-11-30 | A kind of traffic accident monitoring method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106781458B (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107945558A (en) * | 2017-12-21 | 2018-04-20 | 路斌 | It is a kind of that path method and system are seen based on Big Dipper location-based service |
CN107977638A (en) * | 2017-12-11 | 2018-05-01 | 智美达(江苏)数字技术有限公司 | Video monitoring alarm method, device, computer equipment and storage medium |
CN108256447A (en) * | 2017-12-29 | 2018-07-06 | 广州海昇计算机科技有限公司 | A kind of unmanned plane video analysis method based on deep neural network |
CN108921028A (en) * | 2018-06-01 | 2018-11-30 | 上海博泰悦臻电子设备制造有限公司 | The scene of a traffic accident regards acquisition method and system |
CN109410582A (en) * | 2018-11-27 | 2019-03-01 | 易念科技(深圳)有限公司 | Traffic condition analysis method and terminal device |
CN109671006A (en) * | 2018-11-22 | 2019-04-23 | 斑马网络技术有限公司 | Traffic accident treatment method, apparatus and storage medium |
CN109841061A (en) * | 2019-01-24 | 2019-06-04 | 浙江大华技术股份有限公司 | A kind of traffic accident treatment method, apparatus, system and storage medium |
CN110246331A (en) * | 2019-05-30 | 2019-09-17 | 武汉智云集思技术有限公司 | Road condition analyzing method, equipment and readable storage medium storing program for executing based on achievement data |
CN110264711A (en) * | 2019-05-29 | 2019-09-20 | 北京世纪高通科技有限公司 | A kind of traffic accident method of determining probability and device |
WO2019242286A1 (en) * | 2018-06-19 | 2019-12-26 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for allocating service requests |
CN111144485A (en) * | 2019-12-26 | 2020-05-12 | 北京中交兴路车联网科技有限公司 | Vehicle accident judgment method and system based on xgboost classification algorithm |
CN111612200A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Order security prediction method, device, server and storage medium |
CN111859291A (en) * | 2020-06-23 | 2020-10-30 | 北京百度网讯科技有限公司 | Traffic accident recognition method, device, equipment and computer storage medium |
CN112102640A (en) * | 2020-08-11 | 2020-12-18 | 东莞理工学院 | Weather traffic correlation analysis system |
CN112634614A (en) * | 2020-12-16 | 2021-04-09 | 安徽百诚慧通科技有限公司 | Long downhill traffic incident real-time detection method, device and storage medium |
CN114519917A (en) * | 2018-10-29 | 2022-05-20 | 赫克斯冈技术中心 | Mobile monitoring system |
US20230334506A1 (en) * | 2018-09-25 | 2023-10-19 | Capital One Services, Llc | Machine learning-driven servicing interface |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101105892A (en) * | 2007-07-30 | 2008-01-16 | 深圳市融合视讯科技有限公司 | Vehicle traffic accident automatic detection method |
CN101976258A (en) * | 2010-11-03 | 2011-02-16 | 上海交通大学 | Video semantic extraction method by combining object segmentation and feature weighing |
US8368559B2 (en) * | 2009-08-26 | 2013-02-05 | Raytheon Company | Network of traffic behavior-monitoring unattended ground sensors (NeTBUGS) |
US20130287261A1 (en) * | 2012-04-25 | 2013-10-31 | Hon Hai Precision Industry Co., Ltd. | Electronic device and method for managing traffic flow |
CN102654940B (en) * | 2012-05-23 | 2014-05-14 | 上海交通大学 | Processing method of traffic information acquisition system based on unmanned aerial vehicle and |
CN204166693U (en) * | 2014-11-13 | 2015-02-18 | 深圳大学 | A kind of road traffic cruising inspection system based on SUAV (small unmanned aerial vehicle) |
CN106097714A (en) * | 2016-07-27 | 2016-11-09 | 北京信路威科技股份有限公司 | A kind of through street event detection device based on unmanned plane and detection method |
US20160327645A1 (en) * | 2013-12-20 | 2016-11-10 | Perkons S.A. | Traffic monitoring and surveillance system and method, and corresponding traffic infraction recording system and unmanned air vehicle |
-
2016
- 2016-11-30 CN CN201611079545.1A patent/CN106781458B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101105892A (en) * | 2007-07-30 | 2008-01-16 | 深圳市融合视讯科技有限公司 | Vehicle traffic accident automatic detection method |
US8368559B2 (en) * | 2009-08-26 | 2013-02-05 | Raytheon Company | Network of traffic behavior-monitoring unattended ground sensors (NeTBUGS) |
CN101976258A (en) * | 2010-11-03 | 2011-02-16 | 上海交通大学 | Video semantic extraction method by combining object segmentation and feature weighing |
US20130287261A1 (en) * | 2012-04-25 | 2013-10-31 | Hon Hai Precision Industry Co., Ltd. | Electronic device and method for managing traffic flow |
CN102654940B (en) * | 2012-05-23 | 2014-05-14 | 上海交通大学 | Processing method of traffic information acquisition system based on unmanned aerial vehicle and |
US20160327645A1 (en) * | 2013-12-20 | 2016-11-10 | Perkons S.A. | Traffic monitoring and surveillance system and method, and corresponding traffic infraction recording system and unmanned air vehicle |
CN204166693U (en) * | 2014-11-13 | 2015-02-18 | 深圳大学 | A kind of road traffic cruising inspection system based on SUAV (small unmanned aerial vehicle) |
CN106097714A (en) * | 2016-07-27 | 2016-11-09 | 北京信路威科技股份有限公司 | A kind of through street event detection device based on unmanned plane and detection method |
Cited By (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107977638B (en) * | 2017-12-11 | 2020-05-26 | 智美达(江苏)数字技术有限公司 | Video monitoring alarm method, device, computer equipment and storage medium |
CN107977638A (en) * | 2017-12-11 | 2018-05-01 | 智美达(江苏)数字技术有限公司 | Video monitoring alarm method, device, computer equipment and storage medium |
CN107945558A (en) * | 2017-12-21 | 2018-04-20 | 路斌 | It is a kind of that path method and system are seen based on Big Dipper location-based service |
CN108256447A (en) * | 2017-12-29 | 2018-07-06 | 广州海昇计算机科技有限公司 | A kind of unmanned plane video analysis method based on deep neural network |
CN108921028A (en) * | 2018-06-01 | 2018-11-30 | 上海博泰悦臻电子设备制造有限公司 | The scene of a traffic accident regards acquisition method and system |
WO2019242286A1 (en) * | 2018-06-19 | 2019-12-26 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for allocating service requests |
CN110839346A (en) * | 2018-06-19 | 2020-02-25 | 北京嘀嘀无限科技发展有限公司 | System and method for distributing service requests |
US20230334506A1 (en) * | 2018-09-25 | 2023-10-19 | Capital One Services, Llc | Machine learning-driven servicing interface |
CN114519917A (en) * | 2018-10-29 | 2022-05-20 | 赫克斯冈技术中心 | Mobile monitoring system |
CN109671006A (en) * | 2018-11-22 | 2019-04-23 | 斑马网络技术有限公司 | Traffic accident treatment method, apparatus and storage medium |
CN109671006B (en) * | 2018-11-22 | 2021-03-02 | 斑马网络技术有限公司 | Traffic accident handling method, device and storage medium |
CN109410582B (en) * | 2018-11-27 | 2021-11-16 | 易念科技(深圳)有限公司 | Traffic condition analysis method and terminal equipment |
CN109410582A (en) * | 2018-11-27 | 2019-03-01 | 易念科技(深圳)有限公司 | Traffic condition analysis method and terminal device |
CN109841061A (en) * | 2019-01-24 | 2019-06-04 | 浙江大华技术股份有限公司 | A kind of traffic accident treatment method, apparatus, system and storage medium |
CN111612200A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Order security prediction method, device, server and storage medium |
CN111612200B (en) * | 2019-02-25 | 2023-07-14 | 北京嘀嘀无限科技发展有限公司 | Order security prediction method, device, server and storage medium |
CN110264711A (en) * | 2019-05-29 | 2019-09-20 | 北京世纪高通科技有限公司 | A kind of traffic accident method of determining probability and device |
CN110246331A (en) * | 2019-05-30 | 2019-09-17 | 武汉智云集思技术有限公司 | Road condition analyzing method, equipment and readable storage medium storing program for executing based on achievement data |
CN111144485A (en) * | 2019-12-26 | 2020-05-12 | 北京中交兴路车联网科技有限公司 | Vehicle accident judgment method and system based on xgboost classification algorithm |
US11328600B2 (en) | 2020-06-23 | 2022-05-10 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for identifying traffic accident, device and computer storage medium |
CN111859291B (en) * | 2020-06-23 | 2022-02-25 | 北京百度网讯科技有限公司 | Traffic accident recognition method, device, equipment and computer storage medium |
CN111859291A (en) * | 2020-06-23 | 2020-10-30 | 北京百度网讯科技有限公司 | Traffic accident recognition method, device, equipment and computer storage medium |
CN112102640A (en) * | 2020-08-11 | 2020-12-18 | 东莞理工学院 | Weather traffic correlation analysis system |
CN112634614B (en) * | 2020-12-16 | 2022-05-06 | 安徽百诚慧通科技股份有限公司 | Long downhill traffic incident real-time detection method, device and storage medium |
CN112634614A (en) * | 2020-12-16 | 2021-04-09 | 安徽百诚慧通科技有限公司 | Long downhill traffic incident real-time detection method, device and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106781458B (en) | 2019-10-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106781458B (en) | A kind of traffic accident monitoring method and system | |
Jiansheng | Vision-based real-time traffic accident detection | |
CN109460699B (en) | Driver safety belt wearing identification method based on deep learning | |
CN102693432B (en) | Use reliable partial model more to newly arrive and regulate clear path to detect | |
WO2020042984A1 (en) | Vehicle behavior detection method and apparatus | |
US11380105B2 (en) | Identification and classification of traffic conflicts | |
US11315026B2 (en) | Systems and methods for classifying driver behavior | |
CN111800507A (en) | Traffic monitoring method and traffic monitoring system | |
CN110619747A (en) | Intelligent monitoring method and system for highway road | |
CN105788269A (en) | Unmanned aerial vehicle-based abnormal traffic identification method | |
CN102682301B (en) | Adaptation for clear path detection with additional classifiers | |
CN110222596B (en) | Driver behavior analysis anti-cheating method based on vision | |
CN111091110B (en) | Reflection vest wearing recognition method based on artificial intelligence | |
CN105956568A (en) | Abnormal behavior detecting and early warning method based on monitored object identification | |
CN112289037B (en) | Motor vehicle illegal parking detection method and system based on high visual angle under complex environment | |
CN115546742A (en) | Rail foreign matter identification method and system based on monocular thermal infrared camera | |
Sridhar et al. | Helmet violation detection using YOLO v2 deep learning framework | |
CN110097571B (en) | Quick high-precision vehicle collision prediction method | |
Wang et al. | Vision-based highway traffic accident detection | |
CN113706901A (en) | Intelligent prevention, control and early warning system for accidents at entrance section of highway tunnel | |
Ravish et al. | Intelligent traffic violation detection | |
Athree et al. | Vision-based automatic warning system to prevent dangerous and illegal vehicle overtaking | |
Singh et al. | Detection of vacant parking spaces through the use of convolutional neural network | |
Song et al. | Method of Vehicle Behavior Analysis for Real-Time Video Streaming Based on Mobilenet-YOLOV4 and ERFNET | |
KR102340902B1 (en) | Apparatus and method for monitoring school zone |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |