CN109342765A - Vehicle collision detection method and device - Google Patents
Vehicle collision detection method and device Download PDFInfo
- Publication number
- CN109342765A CN109342765A CN201811188917.3A CN201811188917A CN109342765A CN 109342765 A CN109342765 A CN 109342765A CN 201811188917 A CN201811188917 A CN 201811188917A CN 109342765 A CN109342765 A CN 109342765A
- Authority
- CN
- China
- Prior art keywords
- timeslice
- collision
- vehicle
- collision detection
- data
- 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
- G01—MEASURING; TESTING
- G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
- G01P15/00—Measuring acceleration; Measuring deceleration; Measuring shock, i.e. sudden change of acceleration
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/01—Satellite radio beacon positioning systems transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/13—Receivers
- G01S19/14—Receivers specially adapted for specific applications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Computer Networks & Wireless Communication (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to Collision Detection fields, provide a kind of vehicle collision detection method and device.Wherein, vehicle collision detection method includes: the interception time piece from the time series data that the sensor installed on vehicle acquires；Judge whether vehicle parking occurred in timeslice；If there is parking, the cut-off time in timeslice is determined；Interception includes the new timeslice of the data of cut-off time and the multiple moment adjacent with before and after the cut-off time again from time series data；Judge whether to need to carry out collision detection in new timeslice；If desired it detects, by the data vector in new timeslice, obtains this detection data；This detection data is input to collision detection model, obtains the prediction of collision result of collision detection model output.Above-mentioned collision checking method effectively reduces the data volume in collision detection process, improves the operation efficiency of collision detection, improves the precision of collision detection.
Description
Technical field
The present invention relates to Collision Detection fields, in particular to a kind of vehicle collision detection method and device.
Background technique
With the rapid development of society, the use of automobile greatly facilitates daily life, however as possessing
Amount is continuously increased, and the traffic accident frequently occurred brought by him also seriously threatens the life of people and the safety of property.
The emergency relief overwhelming majority of traffic accident be by dial the police emergency number or traffic police patrol etc. manual types could find thing
Therefore in some solutions, it can be believed when detecting collision from transmissions such as trend traffic police departments by Collision Detection
Breath realizes the function of similar alarm.
However, the vehicle collision based on sensing data detects the difference merely by 3-axis acceleration in the prior art
(difference of the 3-axis acceleration of adjacent moment) or threshold value (value range of 3-axis acceleration) judge whether vehicle is collided,
Its Detection accuracy is not high.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of vehicle collision detection method and device, to improve vehicle collision inspection
The accuracy rate of survey.
To achieve the above object, the invention provides the following technical scheme:
In a first aspect, the embodiment of the present invention provides a kind of vehicle collision detection method, comprising:
The interception time piece from the time series data that the sensor installed on vehicle acquires includes multiple moment in timeslice
Data, the data at each moment include the data of multiple fields；
Judge whether vehicle parking occurred in timeslice；
If there is parking, the cut-off time in timeslice is determined；
Interception includes the number of cut-off time and the multiple moment adjacent with before and after the cut-off time again from time series data
According to new timeslice；
Judge whether to need to carry out collision detection in new timeslice；
If desired it detects, by the data vector in new timeslice, obtains this detection data；
This detection data is input to collision detection model, obtains the prediction of collision result of collision detection model output.
Second aspect, the embodiment of the present invention provide a kind of collision detecting apparatus for vehicle, comprising:
Timeslice division module, for the interception time piece from the time series data that the sensor installed on vehicle acquires, when
Between include in piece multiple moment data, the data at each moment include the data of multiple fields；
Stop judgment module, for judging whether vehicle parking occurred in timeslice；
Cut-off time determining module, if determining the cut-off time in timeslice for there is parking；
New timeslice interception module, for from time series data again interception include the cut-off time and with before the cut-off time
The new timeslice of the data at adjacent multiple moment afterwards；
Collision detection judgment module needs to carry out collision detection in new timeslice for judging whether；
Vectorization module, by the data vector in new timeslice, obtains this detection data for if desired detecting；
It is defeated to obtain collision detection model for this detection data to be input to collision detection model for prediction of collision module
Prediction of collision result out.
The third aspect, the embodiment of the present invention provide a kind of computer readable storage medium, on computer readable storage medium
Computer program instructions are stored with, when computer program instructions are read out by the processor and run, the embodiment of the present invention is executed and provides
Method the step of.
Fourth aspect, the embodiment of the present invention provide a kind of electronic equipment, including memory and processor, deposit in memory
Computer program instructions are contained, when computer program instructions are read out by the processor and run, execute first aspect or first aspect
Any one possible implementation provide method the step of.
Technical solution provided by the invention includes at least following the utility model has the advantages that vehicle collision inspection provided in an embodiment of the present invention
It surveys method and device interception time piece from the time series data that sensor acquires to be handled, subsequent calculating step is in the time
It is carried out in piece, therefore can carry out and be calculated, made full use of computing resource, improve the efficiency of collision detection.Meanwhile in benefit
With collision detection model prediction collide result before, can be constructed first according to the parking behavior of vehicle in timeslice there may be
The new timeslice of collision only carries out actual collision detection to these new timeslices, significantly reduces the number of collision detection
According to amount, the efficiency of collision detection is further improved.In addition, the above method and device are fully considered when carrying out collision detection
The timing behavioural characteristic of vehicle in the process of moving, compared to the prior art in only consider the single moment vehicle-state side
Method, detection accuracy are higher.
To enable above-mentioned purpose of the invention, technical scheme and beneficial effects to be clearer and more comprehensible, special embodiment below, and
Cooperate appended attached drawing, is described in detail below.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of structural block diagram that can be applied to the electronic equipment in the embodiment of the present invention；
Fig. 2 shows the flow charts of vehicle collision detection method provided in an embodiment of the present invention；
Fig. 3 shows the process of the step S01 to step S03 of vehicle collision detection method provided in an embodiment of the present invention
Figure；
Fig. 4 shows the process of the step S20 to step S24 of vehicle collision detection method provided in an embodiment of the present invention
Figure；
Fig. 5 shows the functional block diagram of collision detecting apparatus for vehicle provided in an embodiment of the present invention.
Specific embodiment
Below in conjunction with attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Usually exist
The component of the embodiment of the present invention described and illustrated in attached drawing can be arranged and be designed with a variety of different configurations herein.Cause
This, is not intended to limit claimed invention to the detailed description of the embodiment of the present invention provided in the accompanying drawings below
Range, but it is merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall within the protection scope of the present invention.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.Meanwhile of the invention
In description, term " first ", " second " etc. are only used for distinguishing one entity or operation from another entity or operation,
It is not understood to indicate or imply relative importance, can not be understood as require that or imply and be deposited between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Fig. 1 shows a kind of structural block diagram of electronic equipment 100 that can be applied in the embodiment of the present invention.Referring to Fig.1, electric
Sub- equipment 100 includes one or more processors 102, one or more storage devices 104, input unit 106 and output dress
108 are set, these components pass through the interconnection of bindiny mechanism's (not shown) of bus system 112 and/or other forms.
Processor 102 can be central processing unit (CPU) or have data-handling capacity and/or instruction execution capability
Other forms processing unit, and can control the other assemblies in electronic equipment 100 to execute desired function.
Storage device 104 can with various forms of computer readable storage mediums, such as volatile memory and/or it is non-easily
The property lost memory.Volatile memory for example may include random access memory (RAM) and/or cache memory
(cache) etc..Nonvolatile memory for example may include read-only memory (ROM), hard disk, flash memory etc..Computer-readable
Can store one or more computer program instructions on storage medium, processor 102 can run computer program instructions, with
Realize vehicle collision detection method provided in an embodiment of the present invention and/or other desired functions hereafter.In computer
Various application programs and various data can also be stored in readable storage medium storing program for executing, such as application program is used and/or generated each
Kind data etc..
Input unit 106 can be the device that user is used to input instruction, and may include keyboard, mouse, microphone
One or more of with touch screen etc..
Output device 108 can export various information (for example, image or sound) to external (for example, user), and can
To include one or more of display, loudspeaker etc..
It is appreciated that structure shown in FIG. 1 is only to illustrate, electronic equipment 100 may also include it is more than shown in Fig. 1 or
Less component, or with the configuration different from shown in Fig. 1.Each component shown in Fig. 1 can using hardware, software or its
Combination is realized.In the embodiment of the present invention, electronic equipment 100 can be server, personal computer, mobile device, intelligence and wear
The equipment such as equipment, mobile unit are worn, physical equipment can also be not limited to, such as can be virtual machine, Cloud Server etc..
Fig. 2 shows the flow charts of vehicle collision detection method provided in an embodiment of the present invention.Below, with the party
Method is applied to be illustrated for the processor 102 of electronic equipment 100, i.e. the equal processor 102 of the executing subject of method and step.Ginseng
According to Fig. 2, vehicle collision detection method includes:
Step S10: the interception time piece from the time series data that the sensor installed on vehicle acquires.
Vehicle in step S10 refers to the vehicle for needing to carry out collision detection, specifically can be motor vehicle be also possible to it is non-
Motor vehicle.Sensor, such as 3-axis acceleration sensor, GPS sensor are installed on vehicle.Mistake of the sensor in vehicle driving
The data such as data, such as 3-axis acceleration, location information are constantly acquired according to certain frequency in journey.Since these data exist
It is continuous when on the time, because of referred to herein as time series data.Particularly, it is contemplated that concrete application scene of the invention, it is designated herein
Time series data refers to data relevant to collision detection, and in practice, sensor may also collect unrelated with collision detection
Data do not discuss these data below, or can consider that these are unrelated with collision detection before step S10
Data filtering fall.
Time series data has following structure:
[{t1:[C1,2,C3,…,Cm]},
{t2:[C1,C2,C3,…,Cm]},
…
{tn:[C1,C2,C3,…,Cm]}]
Wherein, [t1,t2,t3,…,n] at the time of be n acquisition data, quarter each time acquires [C1,C2,C3,…,Cm] altogether
The data of m field, these fields can be speed, acceleration, longitude, latitude etc..Included by the data of different moments
Field be it is identical, only value can be different, convenient to illustrate, can be by tnThe value of m-th of field at moment is denoted as Cnm。
Timeslice refers to one in the time series data data slot with certain time length, i.e. when including in a timeslice
The data at ordinal number multiple continuous moment in.To ensure that all time series datas are all detected, in step slo it is generally necessary to cut
Take multiple nonseptate timeslices between each other (can have between each timeslice overlapping or non-overlapping).For example, can adopt
It is intercepted with fixed window and fixed step-length:
Timeslice 1
[{t1:[C1,C2,C3,…,Cm]},
{t2:[C1,C2,C3,…,Cm]},
…
{tN:[C1,C2,C3,…,Cm]}]
Timeslice 2
[{tM+1:[C1,C2,C3,…,Cm]},
{tM+2:[C1,C2,C3,…,Cm]},
…
{tM+N:[C1,C2,C3,…,Cm]}]
…
Timeslice w
[{tn-N+1:[C1,C2,C3,…,Cm]},
{tn-N+2:[C1,C2,C3,…,Cm]},
…
{tn:[C1,C2,C3,…,Cm]}]
Window size is N in the mode of interception time piece above, i.e., each timeslice includes N number of moment, and step-length is
M is spaced M moment between that is, adjacent company timeslice.It should be understood that above are only example, different cut can also be taken
Take the mode of timeslice.
Subsequent most of steps, are independently handled both for each timeslice, due to the processing of each timeslice
It is identical when mode, therefore when illustrating subsequent step, it is believed that it is to be illustrated for one of timeslice.It needs
It is to be noted, that specific implementation is, step S10 has different executive modes, for example, can intercept out a timeslice handles one
Timeslice is handled again after can also intercepting out all timeslices.Meanwhile when handling each timeslice, string can be taken
The mode of row processing or parallel processing, specifically without limitation, wherein can make full use of calculating money by the way of parallel processing
Source accelerates arithmetic speed, greatly improves the efficiency of collision detection.
Step S11: judge whether vehicle parking occurred in timeslice.
Under normal condition, vehicle can all cause to stop after colliding, therefore the necessary condition occurred for collision of stopping.Also
If being to say in some timeslice do not occurred parking, the timeslice is also without necessity of progress collision detection.Detect some time
Whether stop, can use in piece, but mode not limited to the following:
Three axis that the value of the 3-axis acceleration field at each moment being primarily based in timeslice calculates each moment accelerate
The signal phasor amplitude (Signal Vector Magnitude, abbreviation SVM) of degree.It might as well assume C3、C4、C5For 3-axis acceleration
Field, then tiThe signal phasor amplitude at momentAfter each moment is calculated, obtain more
A signal phasor amplitude [SVM1,SVM2,SVM3,…,SVMN].Then the maximum value a in multiple signal phasor amplitudes is determinedpAnd
Minimum value av.Finally calculate variable quantity (Change of Signal Vector Magnitude, the abbreviation of signal phasor amplitude
CSVM), CSVM=ap-av, and judge CSVM whether less than the first preset threshold.If difference determines vehicle less than the first preset threshold
Occurred parking in timeslice, and executed step S12, and if difference is not less than the first preset threshold, determined vehicle in timeslice
In do not occurred parking, do not continue to handle the timeslice, handle next timeslice.
Step S12: the cut-off time in timeslice is determined.
It is likely to occur multiple parking in one timeslice, but in step s 12 only it needs to be determined that a cut-off time, example
Such as, first moment that the value of Speed fields in the timeslice is 0 can be determined as the cut-off time.
Step S13: from time series data again interception include the cut-off time and with the cut-off time before and after it is adjacent multiple
The new timeslice of the data at moment.
It should be understood that being not that each parking all corresponds to collision, in sufficiently investigation cut-off time nearby period
The behavior of vehicle, there are some problems for the timeslice intercepted before continuing to use, for example, first moment of timeslice is to stop
The vehicle moment can not investigate the behavior of the vehicle in front at the moment, therefore need to intercept new timeslice in step s 13.
New timeslice can also be intercepted from time series data using fixed window, its certain window size can and step
Difference in S10, new timeslice should include the data at several moment of the front and back of cut-off time.For example, the cut-off time is ts
Moment, then new timeslice can be with are as follows:
[{ts-N1:[C1,C2,C3,…,Cm]},
…
{ts-2:[C1,C2,C3,…,Cm]},
{ts-1:[C1,C2,C3,…,Cm]},
{ts:[C1,C2,C3,…,Cm]}
{ts+1:[C1,C2,C3,…,Cm]},
{ts+2:[C1,C2,C3,…,Cm]},
…
{ts+N2:[C1,C2,C3,…,Cm]}]
Window size is N1+N2+1, tsN1 moment has been intercepted before moment, has intercepted N2 moment later.
Step S14: judge whether to need to carry out collision detection in new timeslice.
Inventor is studied for a long period of time discovery, in certain period of time, is persistently stopped if vehicle is changed by lasting traveling
Vehicle, it may be possible to which (collision parking) caused by collision needs further detection collision or vehicle to be changed by lasting parking and continue
Traveling, it is also possible to (start after collision) caused by collision, needs further to detect collision, and vehicle is repeatedly travelling and is stopping
Switching state between vehicle then not usually caused by collision, does not need further to detect collision, for example, vehicle is when waiting red lights
Short time parking after start or reversing process in repeatedly parking starting etc..Based on above-mentioned discovery, following method can be taken to sentence
It is disconnected whether to need to carry out collision detection in new timeslice:
The dead ship condition at detection acquisition vehicle each moment in new timeslice first.Dead ship condition takes there are two types of possible
Value, traveling and parking, therefore can indicate to travel with 0,1 indicates parking, naturally it is also possible to be indicated with other numerical value or symbol.Inspection
Surveying can be 0 using the value with same method in step S12, i.e., Speed fields sometime, then dead ship condition is 1, no
Then dead ship condition is 0.
Then the changed number of dead ship condition in statistical time piece.For example, will own after the completion of detection above
The dead ship condition at moment is arranged together according to timing, be may be constructed the sequence that one only includes 0 and 1, is counted 01 in the sequence
Or 10 occur the changed number of number, that is, dead ship condition.
Finally, whether the number that judgement counts is equal to 1, if being equal to 1, according to mentioning before the study found that needing
Collision detection is carried out in new timeslice, executes step S15, if being not equal to 1, according to mentioning before the study found that not needing
Collision detection is carried out in new timeslice, does not continue to handle the new timeslice, handles next timeslice.
Step S15: by the data vector in new timeslice, this detection data is obtained.
Vectorization refers to the form for converting the data in new timeslice to vector, because of the collision detection in step S16
Model is usually required that using vector as input.For example, the form for the new timeslice enumerated in step S13, after vectorization
It can be with are as follows:
[C(s-N1)1,C(s-2)1,…,C(s+N2-1)1,C(s+N2)1,
C(s-N1)2,C(s-2)2,…,C(s+N2-1)2,C(s+N2)2,
…
C(s-N1)m,C(s-2)m,…,C(s+N2-1)m,C(s+N2)m]
Step S16: being input to collision detection model for this detection data, and the collision for obtaining the output of collision detection model is pre-
Survey result.
There are two possible as a result, being collision or non-collision for prediction of collision result.Collision detection model usually can use
The historical data sample precondition of a large amount of vehicles obtains, and historical data sample should include that positive sample (is verified the sample for collision
This) and negative sample (being verified the sample for non-collision), whole samples are divided into training set and test set, F1 value can be set
Be set to the optimization aim of model, in the training process using cross validation, grid seek ginseng, it is random seek ginseng and combine continuous iteration,
The F1 value of lift scheme, so that it is determined that the parameters of model are finally completed the training of model, the model of acquisition can use survey
The effect of examination collection verifying prediction.
Do not limit which kind of collision detection model prediction of collision is carried out using in step S16, for example, can be using random gloomy
Woods model (Random Forest).The model can input detection data in bulk, and in bulk export prediction of collision as a result,
Therefore the multinomial detection data of same vehicle can be inputted into the model simultaneously in vector form in the specific implementation, it can also be with
The multinomial detection data of different vehicle is inputted into the model simultaneously in vector form.
After obtaining prediction of collision result, which can be stored, display, or send it to preset system, example
It is such as sent to traffic police department, realizes the function of collision automatic alarm.
In conclusion vehicle collision detection method provided in an embodiment of the present invention is cut from the time series data that sensor acquires
Timeslice is taken to be handled, subsequent calculating step carries out in timeslice, therefore can carry out and be calculated, sufficiently benefit
With computing resource, the efficiency of collision detection is improved.Meanwhile before using collision detection model prediction collision result, first can
The new timeslice there may be collision is constructed according to the parking behavior of vehicle in timeslice, only these new timeslices are carried out
Actual collision detection significantly reduces the data volume of collision detection, further improves the efficiency of collision detection.In addition,
The above method has fully considered the timing behavioural characteristic of vehicle in the process of moving when carrying out collision detection, compared to existing skill
Only consider that the method for the vehicle-state at single moment, detection accuracy are higher in art.
In one embodiment of the invention, before step S10 execution, it can first pass through and execute step S01 to step
S03 detects and handles the abnormal data in time series data, these abnormal datas is avoided to influence the result of collision detection.Referring to Fig. 3:
Step S01: judge the value of the either field of any moment in timeslice with the presence or absence of abnormal.
Following judgment mode can be taken: first, judge the value of the field at the moment whether more than the second preset threshold,
It may be noted that designated herein is more than the range for referring to and limiting more than the second preset threshold, it is default that second might not be greater than
Threshold value is also possible to less than the second preset threshold.
Second, judging the difference of the mean value of the value of the field at all moment in the value and timeslice of the field at the moment
Whether value is more than third predetermined threshold value, for example, third predetermined threshold value can be all in timeslice in common 3 σ method
3 times of the standard deviation of the value of the field at moment.
If the result of above-mentioned two judgement be it is no, determine the value of the field at the moment there is no abnormal, without carrying out
Any processing, if the results of two judgements be not it is no, it is abnormal to determine that the value of the field at the moment exists, executes step S02.
It should be understood that in some embodiments, a judgement in both the above judgement can only be taken to carry out rejecting outliers,
Or other judgment modes can be taken to carry out rejecting outliers.
Step S02: missing values are set by the value of the field at the moment.
Missing values can be null value, be also possible to a preset particular value.The presence of missing values is temporary, purpose
It is intended merely to mark exceptional value, missing values can be filled in step S03.Certainly, in some embodiments, if
In the value for detecting certain field sometime there are after exception, it is replaced with another value immediately, can also be not provided with
Missing values.
Step S03: missing values are filled using prediction model.
Prediction model exports a predicted value after receiving certain input, can use predicted value filling missing values.
It is illustrated so that prediction model is condition random field (Conditional Random Field) model as an example below:
Value training condition first with the field in other times piece different from the timeslice in time series data is random
Field model.By taking the timeslice division mode in step S10 as an example, it is assumed that the C of timeslice 112Value exist it is abnormal, when can pass through
Between C of the piece 2 into timeslice w2The value training condition random field models of field.Usually, it shall be guaranteed that trained using normal value
Model.
Then the value of the field at other moment different from the moment in the timeslice is input to condition according to timing
Random field models obtain the predicted value of the value of the field to the moment of random field models output.Example immediately above, can
With by [C22,C32,C42,…,C(N)2] it is input to conditional random field models.
It is finally filled out using the predicted value that model exports and states missing values.Using handled the time series data of abnormal data into
Row collision detection is conducive to the precision for improving collision detection.
In one embodiment of the invention, before step S15 execution, it can use wavelet transformation in new timeslice
Data carry out noise reduction and to improve the quality of data improve the precision of subsequent collision detection.Wherein, the small echo of use can be, but
It is not limited to DB4 small echo.
In one embodiment of the invention, it after step S16 execution, can will be obtained by step S20 to step S24
The prediction of collision result combination history collision sample obtained further analyzes the reliability of prediction of collision result, and it is pre- to reduce collision as far as possible
The influence of accidental sexual factor during survey obtains the collision detection result for having more reference value.Referring to Fig. 4:
Step S20: judge whether prediction of collision result is to be predicted as colliding.
Step S21: it if being predicted as colliding, calculates between this detection data and at least one history collision sample of vehicle
Similarity, obtain at least one similarity altogether.
Note that this detection data and history collision sample in step S20 should be the sample of same vehicle, otherwise calculate
Similarity has little significance.Wherein, history collision sample refers to by the history number for the vehicle that collision detection model prediction is collision
According to sample.
The definition mode of similarity is not construed as limiting, such as can collide sample pair using this detection data of calculating and history
The cosine value (because two samples be all vector form) for the angle answered, naturally it is also possible to using Euclidean distance, manhatton distance,
The modes such as Hamming distance, related coefficient define similarity.It will be calculated between each history collision sample and this detection data
One similarity.
If may be noted that the vehicle, there is no history to collide sample, it is clear that similarity can not be calculated, will directly be touched at this time
Prediction result is hit as final collision detection result.In addition, if this is predicted as non-collision, do not need further yet
It veritifies, final collision detection result directly can be determined as non-collision.
Step S22: the maximum similarity and corresponding most like history collision sample at least one similarity are determined.
If similarity is defined as vectorial angle cosine value, maximum similarity just refers to maximum cosine value.In view of phase
Like different definition modes is spent, maximum similarity is not necessarily maximum value, is also possible to minimum value, for example, utilizing vector angle
Radian value define similarity, the value of similarity is smaller to show that two vectors are more similar.
Step S23: judge that maximum similarity whether more than the 4th preset threshold, and judges most like history collision sample
Whether it is verified as collision.
In view of the different definition mode of similarity, designated herein is more than the model for referring to and limiting more than the 4th preset threshold
It encloses, the 4th preset threshold might not be greater than, be also possible to less than the 4th preset threshold.
By taking the cosine value of angle as an example, the 4th preset threshold can be defined as to 0.8, cosine value is greater than 0.8 and shows this
Detection data is similar with history collision sample, shows that this detection data and history collision sample are dissimilar less than or equal to 0.8.It goes through
History collision sample be also typically used as train collision detection model sample, therefore its whether be implicitly present in collision be it is known, can
To be verified.
Step S24: determine collision detection result for collision.
If this detection data is similar with history collision sample, and history collision sample is verified to collide, and shows to go through
History result is positive reference, this prediction result is more reliably, final collision detection result can be determined for collision.If
This detection data and history collision sample are dissimilar, and history collision sample is verified as non-collision, shows with historical results
For backward reference, this prediction result is more reliably, can equally to determine final collision detection result for collision.
In the above-described embodiments, it after step S16 obtains prediction of collision result, can first be cached, for calculating in S24
Collision detection as a result, using final collision detection result as the output of vehicle collision detection process.Due to considering vehicle
Historical behavior, therefore the collision detection result accuracy obtained is higher.
The embodiment of the present invention also provides a kind of collision detecting apparatus for vehicle, and Fig. 5 shows collision detecting apparatus for vehicle 200
Functional block diagram.Referring to Fig. 5, which includes:
Timeslice division module 210, for the interception time piece from the time series data that the sensor installed on vehicle acquires,
It include the data at multiple moment in timeslice, the data at each moment include the data of multiple fields；
Stop judgment module 220, for judging whether vehicle parking occurred in timeslice；
Cut-off time determining module 230, if determining the cut-off time in timeslice for there is parking；
New timeslice interception module 240, for when interception is including the cut-off time and with parking again from time series data
Carve the new timeslice of the data at front and back adjacent multiple moment；
Collision detection judgment module 250 needs to carry out collision detection in new timeslice for judging whether；
Vectorization module 260, by the data vector in new timeslice, obtains this testing number for if desired detecting
According to；
Prediction of collision module 270 obtains collision detection model for this detection data to be input to collision detection model
The prediction of collision result of output.
The technical effect of collision detecting apparatus for vehicle 200 provided in an embodiment of the present invention, realization principle and generation is preceding
By the agency of is stated in embodiment of the method, to briefly describe, Installation practice part does not refer to that place, the method for can refer to apply phase in example
Answer content.
The embodiment of the present invention also provides a kind of computer readable storage medium, and meter is stored on computer readable storage medium
Calculation machine program instruction when computer program instructions are read out by the processor and run, executes vehicle provided in an embodiment of the present invention and touches
The step of hitting detection method.The computer readable storage medium can be implemented as, but be not limited to storage device 104 shown in fig. 1.
The embodiment of the present invention also provides a kind of electronic equipment, including memory and processor, is stored with meter in memory
Calculation machine program instruction when computer program instructions are read out by the processor and run, executes vehicle provided in an embodiment of the present invention and touches
The step of hitting detection method.The electronic equipment can be implemented as, but be not limited to electronic equipment 100 shown in fig. 1.
It should be noted that all the embodiments in this specification are described in a progressive manner, each embodiment weight
Point explanation is the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.
For device class embodiment, since it is basically similar to the method embodiment, so being described relatively simple, related place ginseng
See the part explanation of embodiment of the method.
In several embodiments provided herein, it should be understood that disclosed device and method can also pass through it
His mode is realized.The apparatus embodiments described above are merely exemplary, for example, the flow chart and block diagram in attached drawing are aobvious
The device of multiple embodiments according to the present invention, architectural framework in the cards, the function of method and computer program product are shown
It can and operate.In this regard, each box in flowchart or block diagram can represent one of a module, section or code
Point, a part of the module, section or code includes one or more for implementing the specified logical function executable
Instruction.It should also be noted that function marked in the box can also be attached to be different from some implementations as replacement
The sequence marked in figure occurs.For example, two continuous boxes can actually be basically executed in parallel, they sometimes may be used
To execute in the opposite order, this depends on the function involved.It is also noted that each of block diagram and or flow chart
The combination of box in box and block diagram and or flow chart can be based on the defined function of execution or the dedicated of movement
The system of hardware is realized, or can be realized using a combination of dedicated hardware and computer instructions.
In addition, each functional module in each embodiment of the present invention can integrate one independent portion of formation together
Point, it is also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in computer-readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words to existing
Having the part for the part or the technical solution that technology contributes can be embodied in the form of software products, the computer
Software product is stored in a storage medium, including some instructions are used so that computer equipment executes each embodiment institute of the present invention
State all or part of the steps of method.Computer equipment above-mentioned includes: personal computer, server, mobile device, intelligently wears
The various equipment with execution program code ability such as equipment, the network equipment, virtual unit are worn, storage medium above-mentioned includes: U
Disk, mobile hard disk, read-only memory, random access memory, magnetic disk, tape or CD etc. are various to can store program code
Medium.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (11)
1. a kind of vehicle collision detection method characterized by comprising
The interception time piece from the time series data that the sensor installed on vehicle acquires includes multiple moment in the timeslice
Data, the data at each moment include the data of multiple fields；
Judge whether the vehicle described in the timeslice parking occurred；
If there is parking, the cut-off time in the timeslice is determined；
From the time series data again interception include the cut-off time and with it is adjacent multiple before and after cut-off time
The new timeslice of the data at moment；
Judge whether to need to carry out collision detection in the new timeslice；
If desired it detects, by the data vector in the new timeslice, obtains this detection data；
This described detection data is input to collision detection model, obtains the prediction of collision knot of the collision detection model output
Fruit.
2. vehicle collision detection method according to claim 1, which is characterized in that judgement institute in the timeslice
State whether vehicle parking occurred, comprising:
The value of 3-axis acceleration field based on each moment in the timeslice calculates the 3-axis acceleration at each moment
Signal phasor amplitude obtains multiple signal phasor amplitudes altogether；
Determine the maximum value and minimum value in the multiple signal phasor amplitude；
The difference for calculating the maximum value Yu the minimum value judges the difference whether less than the first preset threshold, if described
Difference is less than first preset threshold, determines that the vehicle parking occurred in the timeslice, if the difference is not small
In first preset threshold, determine that the vehicle did not occurred parking in the timeslice.
3. vehicle collision detection method according to claim 1, which is characterized in that stop in the determination timeslice
The vehicle moment, comprising:
First moment that the value of Speed fields in the timeslice is 0 is determined as the cut-off time.
4. vehicle collision detection method according to claim 1, which is characterized in that described to judge whether to need described new
Collision detection is carried out in timeslice, comprising:
Detection obtains the dead ship condition at vehicle each moment in the new timeslice；
Count the changed number of dead ship condition described in the new timeslice；
Judge whether the number is equal to 1, if being equal to 1, determines and need to carry out collision detection in the new timeslice, if differing
In 1, determination does not need to carry out collision detection in the new timeslice.
5. vehicle collision detection method according to claim 1, which is characterized in that the collision detection model is to instruct in advance
The Random Forest model perfected.
6. vehicle collision detection method according to claim 5, which is characterized in that it is described will be in the new timeslice
Before data vector, the method also includes:
Noise reduction process is carried out to the data in the new timeslice using wavelet transformation.
7. vehicle collision detection method according to claim 1, which is characterized in that in the sensing installed from vehicle
In the time series data of device acquisition after interception time piece, and whether occur in judgement vehicle described in the timeslice
It crosses before parking, the method also includes:
Judge the value of the either field of any moment in the timeslice with the presence or absence of abnormal；
It is abnormal if it exists, missing values are set by the value of the field at the moment；
The missing values are filled using prediction model.
8. vehicle collision detection method according to claim 7, which is characterized in that appointing in the judgement timeslice
The value of the either field at one moment is with the presence or absence of abnormal, comprising:
Judge that the value of the field at the moment whether more than the second preset threshold, and judges the field at the moment
Value and the timeslice in the difference of mean value of value of the field at all moment whether be more than third predetermined threshold value, if
The results of two judgements be it is no, determine the value of the field at the moment there is no abnormal, if the results of two judgements are not
Be it is no, it is abnormal to determine that the value of the field at the moment exists.
9. vehicle collision detection method according to claim 8, which is characterized in that the prediction model is condition random field
Model, it is described that the missing values are filled using prediction model, comprising:
Utilize the value training conditional random field models of the field in the other times piece different from the timeslice；
The value of the field at other moment different from the moment in the timeslice is input to the item according to timing
Part random field models obtain the predicted value of the value of the field to the moment of the random field models output；
The missing values are filled using the predicted value.
10. vehicle collision detection method according to claim 1 to 9, which is characterized in that in the acquisition institute
After the prediction of collision result for stating the output of collision detection model, the method also includes:
Judge whether the prediction of collision result is to be predicted as colliding；
If being predicted as colliding, the phase between this described detection data and at least one history collision sample of the vehicle is calculated
Like degree, at least one similarity is obtained altogether, wherein it is collision that the history collision sample, which is by the collision detection model prediction,
The vehicle historical data sample；
Determine that the maximum similarity at least one described similarity and corresponding most like history collide sample；
Judge whether the maximum similarity whether more than the 4th preset threshold, and judges the most like history collision sample
It is verified as collision；
If the maximum similarity is more than the 4th preset threshold, and the most like history collision sample is verified in fact to touch
It hits, determines collision detection result for collision；
If the maximum similarity is no more than the 4th preset threshold, and most like history collision sample be verified in fact for
Non-collision determines the collision detection result for collision.
11. a kind of collision detecting apparatus for vehicle characterized by comprising
Timeslice interception module, for the interception time piece from the time series data that the sensor installed on vehicle acquires, when described
Between include in piece multiple moment data, the data at each moment include the data of multiple fields；
Stop judgment module, for judging whether the vehicle described in the timeslice parking occurred；
Cut-off time determining module, if determining the cut-off time in the timeslice for there is parking；
New timeslice interception module, for interception to include the cut-off time and stops with described again from the time series data
The new timeslice of the data at adjacent multiple moment before and after the vehicle moment；
Collision detection judgment module needs to carry out collision detection in the new timeslice for judging whether；
Vectorization module, by the data vector in the new timeslice, obtains this detection data for if desired detecting；
Prediction of collision module obtains the collision detection mould for this described detection data to be input to collision detection model
The prediction of collision result of type output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811188917.3A CN109342765B (en) | 2018-10-12 | 2018-10-12 | Vehicle collision detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811188917.3A CN109342765B (en) | 2018-10-12 | 2018-10-12 | Vehicle collision detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109342765A true CN109342765A (en) | 2019-02-15 |
| CN109342765B CN109342765B (en) | 2020-11-10 |
Family
ID=65309337
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811188917.3A Active CN109342765B (en) | 2018-10-12 | 2018-10-12 | Vehicle collision detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109342765B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111638032A (en) * | 2020-05-27 | 2020-09-08 | 中国汽车技术研究中心有限公司 | Correction method of acceleration input waveform of trolley crash test and trolley crash test |
| CN113348369A (en) * | 2019-04-19 | 2021-09-03 | 日商爱和谊日生同和保险公司 | Collision determination method, collision determination system, and computer program |
| CN114333318A (en) * | 2021-12-31 | 2022-04-12 | 成都路行通信息技术有限公司 | Motorcycle collision detection method based on sensor space angle |
| CN114565107A (en) * | 2022-04-29 | 2022-05-31 | 阿里巴巴达摩院(杭州)科技有限公司 | Collision detection method, device and equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101168358A (en) * | 2006-10-25 | 2008-04-30 | 厦门雅迅网络股份有限公司 | Method and device used for vehicle collision/side overturn detecting and alarming |
| JP2009154770A (en) * | 2007-12-27 | 2009-07-16 | Nissan Diesel Motor Co Ltd | Collision damage reducing device |
| CN103502068A (en) * | 2011-04-28 | 2014-01-08 | 日产自动车株式会社 | Vehicle control device and vehicle control method |
| CN105069999A (en) * | 2015-08-13 | 2015-11-18 | 成都路行通信息技术有限公司 | Vehicle collision alarm detection method and processing system |
| CN105761323A (en) * | 2016-01-29 | 2016-07-13 | 北京荣之联科技股份有限公司 | Method and device for recognizing collision event based on on-board data |
| CN105976449A (en) * | 2016-05-27 | 2016-09-28 | 大连楼兰科技股份有限公司 | Remote automatic damage assessment and collision detection method and system for vehicle |
| CN107284387A (en) * | 2017-06-13 | 2017-10-24 | 得道车联网络科技（上海）有限公司 | A kind of car crass detection method based on Active Learning |
| CN107463907A (en) * | 2017-08-08 | 2017-12-12 | 东软集团股份有限公司 | Vehicle collision detection method, device, electronic equipment and vehicle |
-
2018
- 2018-10-12 CN CN201811188917.3A patent/CN109342765B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101168358A (en) * | 2006-10-25 | 2008-04-30 | 厦门雅迅网络股份有限公司 | Method and device used for vehicle collision/side overturn detecting and alarming |
| JP2009154770A (en) * | 2007-12-27 | 2009-07-16 | Nissan Diesel Motor Co Ltd | Collision damage reducing device |
| CN103502068A (en) * | 2011-04-28 | 2014-01-08 | 日产自动车株式会社 | Vehicle control device and vehicle control method |
| CN105069999A (en) * | 2015-08-13 | 2015-11-18 | 成都路行通信息技术有限公司 | Vehicle collision alarm detection method and processing system |
| CN105761323A (en) * | 2016-01-29 | 2016-07-13 | 北京荣之联科技股份有限公司 | Method and device for recognizing collision event based on on-board data |
| CN105976449A (en) * | 2016-05-27 | 2016-09-28 | 大连楼兰科技股份有限公司 | Remote automatic damage assessment and collision detection method and system for vehicle |
| CN107284387A (en) * | 2017-06-13 | 2017-10-24 | 得道车联网络科技（上海）有限公司 | A kind of car crass detection method based on Active Learning |
| CN107463907A (en) * | 2017-08-08 | 2017-12-12 | 东软集团股份有限公司 | Vehicle collision detection method, device, electronic equipment and vehicle |
Non-Patent Citations (3)
| Title |
|---|
| HIRONORI SUZUKI, ET AL: "Mitigation of rear-end collision risk based on intent inference of preceding car"s deceleration behavior", 《2017 2ND IEEE INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION ENGINEERING》 * |
| 邹铁方 等: "基于Pc-Crash的多车碰撞事故再现仿真分步方法", 《中国安全科学学报》 * |
| 陈伟: "车辆碰撞预报警***的研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113348369A (en) * | 2019-04-19 | 2021-09-03 | 日商爱和谊日生同和保险公司 | Collision determination method, collision determination system, and computer program |
| CN111638032A (en) * | 2020-05-27 | 2020-09-08 | 中国汽车技术研究中心有限公司 | Correction method of acceleration input waveform of trolley crash test and trolley crash test |
| CN114333318A (en) * | 2021-12-31 | 2022-04-12 | 成都路行通信息技术有限公司 | Motorcycle collision detection method based on sensor space angle |
| CN114333318B (en) * | 2021-12-31 | 2023-04-28 | 成都路行通信息技术有限公司 | Motorcycle collision detection method based on sensor space angle |
| CN114565107A (en) * | 2022-04-29 | 2022-05-31 | 阿里巴巴达摩院(杭州)科技有限公司 | Collision detection method, device and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109342765B (en) | 2020-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109342765A (en) | Vehicle collision detection method and device | |
| US11363411B2 (en) | Methods for combining sensor data to determine vehicle movement information | |
| Vlahogianni et al. | Driving analytics using smartphones: Algorithms, comparisons and challenges | |
| CN104827963B (en) | A kind of method, control system and control device for collision prevention of vehicle intelligent early-warning | |
| CN109076093A (en) | The system and method for controlling sensor-based data signal sampling and processing in vehicle | |
| Xu et al. | Leveraging audio signals for early recognition of inattentive driving with smartphones | |
| CN109408583B (en) | Data processing method and device, computer readable storage medium and electronic equipment | |
| CN113593218B (en) | Method and device for detecting traffic abnormal event, electronic equipment and storage medium | |
| CN106446771A (en) | Driver attention evaluation | |
| US10740634B1 (en) | Detection of decline in concentration based on anomaly detection | |
| US10838782B2 (en) | Event abstractor | |
| US10993079B2 (en) | Motion detection method, device, and medium | |
| US11710082B2 (en) | Systems and methods for utilizing machine learning and feature selection to classify driving behavior | |
| CN104063604B (en) | A kind of motion state detection method and apparatus | |
| CN114970705B (en) | Running state analysis method, device, equipment and medium based on multi-sensing data | |
| CN105868266A (en) | Clustering model based high-dimensional data stream outlier detection method | |
| CN110164181A (en) | A kind of car accident detection, method for early warning, equipment and system | |
| WO2022153125A3 (en) | System and method for processing vehicle event data for analysis of road segments and turn ratios | |
| CN104964736B (en) | Optical fiber invasion vibration source identification method based on time-frequency characteristic maximum expected classification | |
| CN110196440A (en) | It is overlapped recognition methods and device, electronic equipment, the storage medium of track | |
| CN111709665A (en) | Vehicle safety evaluation method and device | |
| CN113806413B (en) | Trajectory screening and classifying method and device, computer equipment and storage medium | |
| CN109919293A (en) | A kind of dangerous driving determination method and device | |
| Zhang et al. | Driving data generation using affinity propagation, data augmentation, and convolutional neural network in communication system | |
| CN112820111B (en) | Abnormal vehicle identification method and device and electronic equipment |
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 |