CN109063642A - The prediction technique and system that pedestrian based on HMM algorithm goes across the road - Google Patents
The prediction technique and system that pedestrian based on HMM algorithm goes across the road Download PDFInfo
- Publication number
- CN109063642A CN109063642A CN201810861624.0A CN201810861624A CN109063642A CN 109063642 A CN109063642 A CN 109063642A CN 201810861624 A CN201810861624 A CN 201810861624A CN 109063642 A CN109063642 A CN 109063642A
- Authority
- CN
- China
- Prior art keywords
- road
- data
- pedestrian
- deep learning
- learning model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses prediction techniques and system that the pedestrian based on HMM algorithm goes across the road, which comprises receives the current data of acquisition device acquisition；HMM algorithm, which is based on, according to the current data and pre-stored historical data establishes deep learning model；Judge that front has pedestrian to go across the road by the deep learning model；Output control signals to loudspeaker；Loudspeaker is sounded an alarm according to the control signal, with alerting drivers, can not only detect that vehicle periphery has pedestrian in this way, while can also predict whether pedestrian can go across the road.In addition, road data is applied during establishing deep learning model, the judgement whether gone across the road by the way that road information is added to pedestrian, so that prediction result is more accurate.
Description
Technical field
The present invention relates to technical field of data prediction, and in particular to the prediction technique that the pedestrian based on HMM algorithm goes across the road
And system.
Background technique
Pedestrian detection technology is to judge to whether there is pedestrian in image or video sequence and give to be accurately positioned.Existing row
It is the method based on statistical learning that people, which detects most common method, i.e., constructs pedestrian detection classifier according to a large amount of sample.Sample
The feature of this extraction mainly includes the information such as gray value, frame profile, color, the histogram of gradients for extracting target.Classifier master
It to include neural network, SVM, adaboost and the deep learning for being considered as favorite by computer vision now.And a large amount of sample
It originally is obtained by the way that the camera at crossing is arranged in.If being positioned by pedestrian of the above method to vehicle periphery, or judgement
Out pedestrian close to vehicle when driver is given a warning.However, such method can only position the pedestrian of vehicle periphery, and
It detects that there is the presence of pedestrian in front, the behavior of pedestrian can not be predicted.And the behavior of pedestrian is predicted, to pedestrian
Safety, which is gone across the road, plays vital influence.Therefore, the interior urgent need of industry, which researches and develops one kind, to predict the behavior of pedestrian
Method.
Summary of the invention
The purpose of the invention is to overcome above the shortcomings of the prior art, providing one kind can be to the row of pedestrian
The prediction technique and system gone across the road for the pedestrian based on HMM algorithm predicted.
The purpose of the present invention is realized by the following technical solution:
The prediction technique that pedestrian based on HMM algorithm goes across the road, including, receive the current data of acquisition device acquisition；Root
HMM algorithm, which is based on, according to the current data and pre-stored historical data establishes deep learning model；Pass through the depth
It practises model and judges that front has pedestrian to go across the road；Output control signals to loudspeaker；Loudspeaker is issued according to the control signal
Alarm, with alerting drivers.
Preferably, described that depth is established based on HMM algorithm according to the current data and pre-stored historical data
Practising model includes: successively to establish deep learning model with HMM algorithm according to the historical data；Correct the deep learning mould
Type；The current data is synchronized to the deep learning model.
Preferably, it is described according to the historical data with HMM algorithm successively establish deep learning model include: will be described
Historical data is divided into the training sample of regular length；The historical data is subjected to feature extraction；Using training sample to point
Class device is trained；Baum-Welch algorithm is used according to given random parameter and the training sample, obtains best HMM ginseng
Number, establishes out deep learning model.
Preferably, described that the historical data is divided into before the training sample of regular length further include: to be gone through to described
History data are filtered Denoising disposal.
Preferably, described device acquisition is camera, and the current data includes: current behavior of the pedestrian before going across the road
Data and present road data, the historical data include: historical behavior data and history road number of the pedestrian before going across the road
According to.
The forecasting system that pedestrian based on HMM algorithm goes across the road, comprising: acquisition equipment, prediction meanss and loudspeaker；It is described
Equipment is acquired, crossing is mounted on, for acquiring the current data of pedestrian；And the prediction is sent by the current data and is filled
It sets；The prediction meanss, including database, for being calculated according to the current data and pre-stored historical data based on HMM
Method establishes deep learning model；Judge that front has pedestrian to go across the road by the deep learning model；And export control signal
To loudspeaker；The database；For storing historical data of the pedestrian before going across the road；The loudspeaker, loudspeaker is according to institute
It states control signal to sound an alarm, with alerting drivers.
Preferably, the prediction meanss include model foundation unit, Modifying model unit and data synchronisation unit；The mould
Type establishes unit, successively establishes deep learning model with HMM algorithm according to the historical data；The Modifying model unit,
For correcting the deep learning model；The data synchronisation unit, for the current data to be synchronized to the depth
Practise model.
Preferably, the model foundation unit is also used to for the historical data being divided into the training sample of regular length；
The historical data is subjected to feature extraction；Classifier is trained using training sample；According to given random parameter and
The training sample uses Baum-Welch algorithm, obtains best HMM parameter, establishes out deep learning model.
Preferably, the prediction meanss further include: data processing unit；The data processing unit, for being gone through to described
History data are filtered Denoising disposal.
Preferably, described device acquisition is camera, and the current data includes: current behavior of the pedestrian before going across the road
Data and present road data, the historical data include: historical behavior data and history road number of the pedestrian before going across the road
According to.
The present invention has the advantage that compared with the existing technology
The current data that this programme is acquired by receiving acquisition device, according to the current data and pre-stored history
Data application HMM algorithm establishes deep learning model；Judge whether front has pedestrian to go across the road by the deep learning model；
If so, outputing control signals to loudspeaker；Loudspeaker is sounded an alarm according to the control signal, with alerting drivers, in this way
It can not only detect that vehicle periphery has pedestrian, while can also predict whether pedestrian can go across the road.In addition, establishing deep learning mould
During type, road data is applied, the judgement whether gone across the road by the way that road information is added to pedestrian, so that prediction result
It is more accurate.
Detailed description of the invention
Fig. 1 is the flow diagram for the prediction technique that the pedestrian of the invention based on HMM algorithm goes across the road.
Fig. 2 is of the invention to be based on HMM algorithm according to the current data and pre-stored historical data and establish depth
The flow diagram of learning model.
Fig. 3 is that the process of the invention for using HMM algorithm successively to establish deep learning model according to the historical data is shown
It is intended to.
Fig. 4 is the structural schematic diagram for the forecasting system that the pedestrian of the invention based on HMM algorithm goes across the road.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples.
The prediction technique that pedestrian based on HMM algorithm as shown in Figs. 1-3 goes across the road, comprising:
S11 receives the current data of acquisition device acquisition；Described device acquisition is camera, and the camera can be set
Set the traffic camera on road or be arranged in vehicle body front side itself with camera, be not specifically limited herein.It is described
Current data includes: current behavior data and present road data of the pedestrian before going across the road.More preferably, the camera setting
In vehicle body front side.
S12 is based on HMM algorithm according to the current data and pre-stored historical data and establishes deep learning model；
Also the current data is saved in the database simultaneously, the historical data of deep learning model is established as next time.It is described to go through
History data include: historical behavior data and history road data of the pedestrian before going across the road.Wherein, deep learning model refers to:
It can be trained by large-scale data, obtain the machine learning frame containing many hidden layers of the representative feature of sample
Structure model.
Specifically, step S12 includes:
S121 successively establishes deep learning model with HMM algorithm according to the historical data；The HMM algorithm is hidden
Markov HMM algorithm.
S122 corrects the deep learning model；Specifically, the deep learning model is constantly corrected.
The current data is synchronized to the deep learning model by S123.
Wherein, step S121 includes:
The historical data is divided into the training sample of regular length by S1211；Before step S121 further include: right
The historical data is filtered Denoising disposal.
The historical data is carried out feature extraction by S1212；
S1213 is trained classifier using training sample；
S1214 uses Baum-Welch algorithm according to given random parameter and the training sample, obtains best HMM
Parameter establishes out deep learning model.
S13 judges whether front has pedestrian to go across the road by the deep learning model；If so, thening follow the steps S14；
If it is not, then terminating.Deep learning model after synchronous current data obtains result and exports.
S14 outputs control signals to loudspeaker；
S15, loudspeaker is sounded an alarm according to the control signal, with alerting drivers.
As shown in figure 4, being somebody's turn to do the pedestrian based on HMM algorithm that the prediction technique that the pedestrian based on HMM algorithm goes across the road is applicable in
The forecasting system gone across the road includes: acquisition equipment, prediction meanss and loudspeaker；The acquisition equipment, is mounted on crossing, for adopting
Collect the current data of pedestrian；And the prediction meanss are sent by the current data；The prediction meanss, including database,
Deep learning model is established for being based on HMM algorithm according to the current data and pre-stored historical data；By described
Deep learning model judges that front has pedestrian to go across the road；And output control signals to loudspeaker；The database；For storing
Historical data of the pedestrian before going across the road；The loudspeaker, loudspeaker are sounded an alarm according to the control signal, are driven with warning
Member.
In the present embodiment, the prediction meanss include model foundation unit, Modifying model unit and data synchronisation unit；Institute
Model foundation unit is stated, deep learning model is successively established with HMM algorithm according to the historical data；The Modifying model list
Member, for correcting the deep learning model；The data synchronisation unit, for the current data to be synchronized to the depth
Learning model.
In the present embodiment, the model foundation unit is also used to for the historical data being divided into the training of regular length
Sample；The historical data is subjected to feature extraction；Classifier is trained using training sample；According to given random ginseng
The several and training sample uses Baum-Welch algorithm, obtains best HMM parameter, establishes out deep learning model.
In the present embodiment, the prediction meanss further include: data processing unit；The data processing unit, for institute
It states historical data and is filtered Denoising disposal.
In the present embodiment, described device acquisition is camera, and the current data includes: pedestrian is current before going across the road
Behavioral data and present road data, the historical data include: historical behavior data and history road of the pedestrian before going across the road
Circuit-switched data.
It should be noted that being connected when the prediction meanss are assembled onboard with camera and loudspeaker.
The current data that this programme is acquired by receiving acquisition device, according to the current data and pre-stored history
Data application HMM algorithm establishes deep learning model；Judge whether front has pedestrian to go across the road by the deep learning model；
If so, outputing control signals to loudspeaker；Loudspeaker is sounded an alarm according to the control signal, with alerting drivers, in this way
It can not only detect that vehicle periphery has pedestrian, while can also predict whether pedestrian can go across the road.In addition, establishing deep learning mould
During type, road data is applied, the judgement whether gone across the road by the way that road information is added to pedestrian, so that prediction result
It is more accurate.
Above-mentioned specific embodiment is the preferred embodiment of the present invention, can not be limited the invention, and others are appointed
The change or other equivalent substitute modes what is made without departing from technical solution of the present invention, are included in protection of the invention
Within the scope of.
Claims (10)
1. the prediction technique that the pedestrian based on HMM algorithm goes across the road characterized by comprising
Receive the current data of acquisition device acquisition；
HMM algorithm, which is based on, according to the current data and pre-stored historical data establishes deep learning model；
Judge that front has pedestrian to go across the road by the deep learning model；
Output control signals to loudspeaker；
Loudspeaker is sounded an alarm according to the control signal, with alerting drivers.
2. the prediction technique that the pedestrian according to claim 1 based on HMM algorithm goes across the road, which is characterized in that described
Establishing deep learning model based on HMM algorithm according to the current data and pre-stored historical data includes:
Deep learning model is successively established with HMM algorithm according to the historical data；
Correct the deep learning model；
The current data is synchronized to the deep learning model.
3. the prediction technique that the pedestrian according to claim 2 based on HMM algorithm goes across the road, which is characterized in that described
Successively establishing deep learning model with HMM algorithm according to the historical data includes:
The historical data is divided into the training sample of regular length；
The historical data is subjected to feature extraction；
Classifier is trained using training sample；
Baum-Welch algorithm is used according to given random parameter and the training sample, best HMM parameter is obtained, establishes out
Deep learning model.
4. the prediction technique that the pedestrian according to claim 3 based on HMM algorithm goes across the road, which is characterized in that described to incite somebody to action
The historical data is divided into before the training sample of regular length further include:
Denoising disposal is filtered to the historical data.
5. the prediction technique that the pedestrian based on HMM algorithm of -4 any one goes across the road according to claim 1, feature
It is, described device acquisition is camera, and the current data includes: current behavior data of the pedestrian before going across the road and current
Road data, the historical data include: historical behavior data and history road data of the pedestrian before going across the road.
6. the forecasting system that the pedestrian based on HMM algorithm goes across the road characterized by comprising acquire equipment, prediction meanss and raise
Sound device；
The acquisition equipment, for acquiring the current data of pedestrian；And the prediction meanss are sent by the current data；
The prediction meanss, including database, for being calculated according to the current data and pre-stored historical data based on HMM
Method establishes deep learning model；Judge that front has pedestrian to go across the road by the deep learning model；And export control signal
To loudspeaker；
The database；For storing historical data of the pedestrian before going across the road；
The loudspeaker, for being sounded an alarm according to the control signal, with alerting drivers.
7. the forecasting system that the pedestrian according to claim 6 based on HMM algorithm goes across the road, which is characterized in that described pre-
Surveying device includes: model foundation unit, Modifying model unit and data synchronisation unit；
The model foundation unit, for successively establishing deep learning model with HMM algorithm according to the historical data；
The Modifying model unit, for correcting the deep learning model；
The data synchronisation unit, for the current data to be synchronized to the deep learning model.
8. the forecasting system that the pedestrian according to claim 7 based on HMM algorithm goes across the road, which is characterized in that the mould
Type establishes unit, is also used to for the historical data being divided into the training sample of regular length；
The historical data is subjected to feature extraction；
Classifier is trained using training sample；
Baum-Welch algorithm is used according to given random parameter and the training sample, best HMM parameter is obtained, establishes out
Deep learning model.
9. the forecasting system that the pedestrian according to claim 8 based on HMM algorithm goes across the road, which is characterized in that described pre-
Survey device further include: data processing unit；
The data processing unit, for being filtered Denoising disposal to the historical data.
10. the forecasting system that the pedestrian according to claim 6 based on HMM algorithm goes across the road, which is characterized in that the dress
Setting acquisition is camera, and the current data includes: current behavior data and present road data of the pedestrian before going across the road, institute
Stating historical data includes: historical behavior data and history road data of the pedestrian before going across the road.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810861624.0A CN109063642A (en) | 2018-08-01 | 2018-08-01 | The prediction technique and system that pedestrian based on HMM algorithm goes across the road |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810861624.0A CN109063642A (en) | 2018-08-01 | 2018-08-01 | The prediction technique and system that pedestrian based on HMM algorithm goes across the road |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109063642A true CN109063642A (en) | 2018-12-21 |
Family
ID=64832422
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810861624.0A Pending CN109063642A (en) | 2018-08-01 | 2018-08-01 | The prediction technique and system that pedestrian based on HMM algorithm goes across the road |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109063642A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120163206A1 (en) * | 2010-08-24 | 2012-06-28 | Kenneth Man-Kin Leung | Method and apparatus for analysis of user traffic within a predefined area |
| CN103640532A (en) * | 2013-11-29 | 2014-03-19 | 大连理工大学 | Pedestrian anti-collision early warning method based on recognition of braking and accelerating intention of driver |
| CN104802793A (en) * | 2014-01-23 | 2015-07-29 | 罗伯特·博世有限公司 | Method and device for classifying a behavior of a pedestrian when crossing a roadway of a vehicle as well as passenger protection system of a vehicle |
| CN106428000A (en) * | 2016-09-07 | 2017-02-22 | 清华大学 | Vehicle speed control device and method |
| CN108022026A (en) * | 2018-01-25 | 2018-05-11 | 戴姆勒股份公司 | Traffic safety early warning method and device and system comprising device |
-
2018
- 2018-08-01 CN CN201810861624.0A patent/CN109063642A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120163206A1 (en) * | 2010-08-24 | 2012-06-28 | Kenneth Man-Kin Leung | Method and apparatus for analysis of user traffic within a predefined area |
| CN103640532A (en) * | 2013-11-29 | 2014-03-19 | 大连理工大学 | Pedestrian anti-collision early warning method based on recognition of braking and accelerating intention of driver |
| CN104802793A (en) * | 2014-01-23 | 2015-07-29 | 罗伯特·博世有限公司 | Method and device for classifying a behavior of a pedestrian when crossing a roadway of a vehicle as well as passenger protection system of a vehicle |
| CN106428000A (en) * | 2016-09-07 | 2017-02-22 | 清华大学 | Vehicle speed control device and method |
| CN108022026A (en) * | 2018-01-25 | 2018-05-11 | 戴姆勒股份公司 | Traffic safety early warning method and device and system comprising device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108995654B (en) | Driver state identification method and system | |
| Berndt et al. | Continuous driver intention recognition with hidden markov models | |
| CN110203202B (en) | Lane changing auxiliary early warning method and device based on driver intention recognition | |
| CN106647776B (en) | Method and device for judging lane changing trend of vehicle and computer storage medium | |
| CN109145719B (en) | Driver fatigue state identification method and system | |
| US7209574B2 (en) | Eye tracking apparatus, eye tracking method, eye state judging apparatus, eye state judging method and computer memory product | |
| CN110688877B (en) | Danger early warning method, device, equipment and storage medium | |
| US20190057604A1 (en) | Method, device and system for processing startup of preceding vehicle | |
| DE102011078615B4 (en) | OBJECT DETECTION DEVICE AND OBJECT DETECTION PROGRAM | |
| US9352746B2 (en) | Lane relative position estimation method and system for driver assistance systems | |
| CN105956548A (en) | Driver fatigue state detection method and device | |
| US10929715B2 (en) | Semantic segmentation using driver attention information | |
| CN106339692B (en) | A kind of fatigue driving state information determines method and system | |
| CN109562763A (en) | The control method and control device of automatic driving vehicle | |
| CN112287910B (en) | Road abnormal area detection method, road abnormal area detection device, electronic equipment and storage medium | |
| CN107944425A (en) | The recognition methods of road sign and device | |
| CN112863245B (en) | Vehicle track change real-time prediction method based on deep neural network | |
| CN108926352A (en) | A kind of driving fatigue detection method and system | |
| CN109389060A (en) | A kind of vehicle week anti-collision warning method of view-based access control model | |
| CN110276398A (en) | A kind of video abnormal behaviour automatic judging method | |
| CN113581199A (en) | Vehicle control method and device | |
| CN109278759B (en) | Vehicle safe driving auxiliary system | |
| CN111091131A (en) | Self-adaptive license plate character recognition system and recognition method based on multi-task learning | |
| EP3605385A1 (en) | Automated parking system | |
| CN109063642A (en) | The prediction technique and system that pedestrian based on HMM algorithm goes across the road |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181221 |