CN109214438A - A kind of building method of the driving behavior identifying system based on convolutional neural networks - Google Patents
A kind of building method of the driving behavior identifying system based on convolutional neural networks Download PDFInfo
- Publication number
- CN109214438A CN109214438A CN201810959518.6A CN201810959518A CN109214438A CN 109214438 A CN109214438 A CN 109214438A CN 201810959518 A CN201810959518 A CN 201810959518A CN 109214438 A CN109214438 A CN 109214438A
- Authority
- CN
- China
- Prior art keywords
- driving behavior
- convolutional neural
- neural networks
- system based
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Traffic Control Systems (AREA)
Abstract
The building method for the driving behavior identifying system based on convolutional neural networks that the invention discloses a kind of, successively follows the steps below: acquisition driving behavior sample data；Data filtering；Data format is regular；Driving behavior identification, will be regular after driving behavior sample data as input, be input to the convolutional neural networks put up, by pond, export as driving behavior type to be identified；Driving behavior is trained by convolutional neural networks.The present invention solves and relies on artificial selection when conventional machines identify driving behavior in driving behavior identification field, accuracy of identification is not high, the problem of conventional machines can not handle large-scale data simultaneously, a kind of building method of driving behavior identifying system based on convolutional neural networks is provided, it is needed automatically to extract feature according to system based on convolutional neural networks when it is applied, improve the accuracy of identification of driving behavior, and extensive driving behavior data set can be efficiently used, identify a greater variety of driving behaviors.
Description
Technical field
The present invention relates to deep learning fields, and in particular to a kind of driving behavior identifying system based on convolutional neural networks
Building method.
Background technique
With the rapid development of economic society, China's vehicle guaranteeding organic quantity is in increase situation rapidly.Vehicles number it is fast
Speed increases the traffic accident for also resulting in further serious traffic congestion while bringing trip convenience and taking place frequently.
It shows according to statistics from the Traffic Management Bureau of the Ministry of Public Security, 90% fatal traffic accident is due to human factor.Based on considerations above, the present invention will
The driving behavior of people, such as acceleration, brake, turning, lane change, continuous lane change and overtaking other vehicles identify, and are fed back to driving
Personnel or related traffic control department, remind driver to drive with caution, and related traffic control department is prompted to cope in time.
With MEMS constantly drilling towards directions such as small size, light weight, low-power consumption, low cost and high integration
Into integrated multiple sensors just provide data for the research of different field and support, driving behavior identifies that field is no exception.
Using inertial sensors such as accelerometer and gyroscopes, inertia sensing number needed for description motion state of automobile can be collected
According to.When handling these data, there are certain limitations for traditional machine learning method, first is that system performance depends on Characteristic Vectors
The selection of amount, i.e., the data characteristics manually extracted is different, and obtained driving behavior accuracy of identification is also different；Second is that traditional machine
Learning method is had too many difficulties to cope with when handling large-scale dataset, cannot efficiently use all sample informations.
Above-mentioned limitation is then not present in convolutional neural networks, can not be by the constraint of conventional statistics feature, can be according to system
It needs automatically to extract feature, this characteristic is but also it yields unusually brilliant results in fields such as image procossing, video processing.Specific to driving
Activity recognition field is sailed, convolutional neural networks not only can extract and screen in time domain, frequency domain or wavelet field to avoid artificially special
Sign, and extensive driving behavior data set can be efficiently used, identify a greater variety of driving behaviors.
Summary of the invention
The present invention solves and relies on artificial selection when conventional machines identify driving behavior in driving behavior identification field, identifies
The problem of precision is not high, while conventional machines can not handle large-scale data provides a kind of driving based on convolutional neural networks
The building method of Activity recognition system is needed automatically to extract feature according to system, be mentioned when applying based on convolutional neural networks
The accuracy of identification of high driving behavior, and extensive driving behavior data set can be efficiently used, identify a greater variety of driving
Behavior.
The present invention is achieved through the following technical solutions:
A kind of building method of the driving behavior identifying system based on convolutional neural networks, successively follows the steps below:
A, driving behavior sample data is acquired；
B, data filtering is analyzed the noise composition in driving behavior sample data, is filtered by filter to data,
Eliminate influence of the noise to system；
C, data format is regular, and filtered driving behavior sample data is regular at m row × n column matrix, to meet
The input requirements of convolutional neural networks；
D, driving behavior identify, will be regular after driving behavior sample data matrix be input to convolutional neural networks, to sample
Notebook data matrix carries out pondization and samples, and is first carried out 1 × 2 pond, the concrete operations in pond be exactly with 1 × 2 window m ×
The characteristic layer of n carries out the movement of step-length 1 × 2, and the maximum value in window is taken to obtain as output every timePond layer it is defeated
Out, 1 × 3 pond is then executed, the concrete operations in pond are exactly to exist with 1 × 3 windowCharacteristic layer carry out step-length 1
× 3 movement takes the maximum value in window to obtain as output every timePond layer output, finally output drive row
For type；；
E, driving behavior is trained by convolutional neural networks, completes building for entire driving behavior identifying system.
Traditional driving behavior identification, using inertial sensors such as accelerometer and gyroscopes, can collect description vapour
Inertia sensing data needed for vehicle motion state.When handling these data, there are certain limits for traditional machine learning method
System, first is that system performance depends on the selection of characteristic vector, i.e., the data characteristics manually extracted is different, and obtained driving behavior is known
Other precision is also different；Second is that traditional machine learning method is had too many difficulties to cope with when handling large-scale dataset, cannot efficiently use
All sample informations.
The essence in pond is sampling, and pond layer selects certain mode (maximum value or average value) for the feature of input
It is compressed, the meaning of pond layer is it is clear that one is that can reduce parameter, and to feature progress dimensionality reduction, one is to increase
Strong feature aluminium bar keeps output constant, there is certain Anti-Jamming for there is the input feature vector of micro-displacement.
The present invention is based on the driving behavior systems of convolutional neural networks, and above-mentioned limitation is then not present, can not be by conventional statistics
The constraint of feature can need automatically to extract feature according to system, identify field specific to driving behavior, convolutional neural networks are not
Only and extensive driving can be efficiently used to avoid artificially extracting and screening feature in time domain, frequency domain or wavelet field
Behavioral data collection identifies a greater variety of driving behaviors.Under true road conditions, the acquisition of driving behavior sample will receive noise shadow
It rings, first is that the system noise of MEMS, second is that the noise as caused by vehicle vibration, as vehicle by deceleration strip or starts
Machine caused vibration when working.In above two noise, system noise amplitude is much smaller compared to vehicle motion data, can
To ignore；Shaking noise compared to vehicle motion data is high-frequency noise, is increased on the basis of vehicle motion data
Random bias, usable low-pass filter effectively filter out vehicle vibration noise.So by collected driving behavior sample data
It is filtered, then carries out that data format is regular, filtered driving behavior sample data is regular at m row × n column square
Battle array, to meet the input requirements of convolutional neural networks；Carry out driving behavior identification process again, will be regular after driving behavior sample
Data are input to the convolutional neural networks put up, export as driving behavior type to be identified as input；Pass through convolution mind
Driving behavior is trained through network, completes building for entire driving behavior identifying system.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, in the step A
Driving behavior sample data include acceleration information and angular velocity data.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, in the step B
Filter be low-pass filter.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, in the step D
Convolutional neural networks include input layer, convolutional layer one, convolutional layer two, full articulamentum one, full articulamentum two and output layer.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, in the step D
Convolutional neural networks in, Batch value be 64.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, in the step D
Convolutional neural networks in, using using ReLU as activation primitive.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, using cross entropy
Function is as loss function.
Further, a kind of building method of the driving behavior identifying system based on convolutional neural networks, the step E
In: driving behavior is trained by convolutional neural networks, is trained by the way of stochastic gradient optimization.
Compared with prior art, the present invention having the following advantages and benefits:
1, it needs automatically to extract feature according to system based on convolutional neural networks when present invention application, improves driving behavior
Accuracy of identification, by driving behavior acquisition, data filtering, data format be regular, driving behavior identification, pondization sample, most
Driving behavior type is exported afterwards, pondization can reduce parameter, carry out dimensionality reduction to feature, and one is Enhanced feature aluminium bar, for
There is the input feature vector of micro-displacement, keeps output constant, have certain Anti-Jamming, to enhance driving behavior of the present invention
Accuracy of identification.
2, the present invention can efficiently use extensive driving behavior data set, identify a greater variety of driving behaviors.
3, the present invention is high by the overall discrimination of convolutional neural networks training, is conducive to identify more driving behaviors, and will
It feeds back to driver or related traffic control department, and driver is reminded to drive with caution, and related traffic control department is prompted to cope in time.
Detailed description of the invention
Attached drawing described herein is used to provide to further understand the embodiment of the present invention, constitutes one of the application
Point, do not constitute the restriction to the embodiment of the present invention.In the accompanying drawings:
Fig. 1 is the structural diagram of the present invention；
Fig. 2 is the waveform diagram of data filtering process of the present invention；
Fig. 3 is the system building structure chart of convolutional neural networks in the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below with reference to embodiment and attached drawing, to this
Invention is described in further detail, and exemplary embodiment of the invention and its explanation for explaining only the invention, are not made
For limitation of the invention.
Embodiment
As shown in Figure 1 to Figure 3, a kind of building method of the driving behavior identifying system based on convolutional neural networks, successively
It follows the steps below:
A, driving behavior sample data is acquired, driving behavior sample data includes acceleration information and angular velocity data；
B, data filtering analyzes the noise composition in driving behavior sample data, as shown in Figure 2 by filter to data
It is filtered, eliminates influence of the noise to system, filter is low-pass filter；
C, data format is regular, and filtered driving behavior sample data is regular at m row × n column matrix, to meet
The input requirements of convolutional neural networks；
D, driving behavior identify, will be regular after driving behavior sample data matrix be input to convolutional neural networks, to sample
Notebook data matrix carries out pondization and samples, and is first carried out 1 × 2 pond, the concrete operations in pond be exactly with 1 × 2 window m ×
The characteristic layer of n carries out the movement of step-length 1 × 2, and the maximum value in window is taken to obtain as output every timePond layer it is defeated
Out, 1 × 3 pond is then executed, the concrete operations in pond are exactly to exist with 1 × 3 windowCharacteristic layer carry out step-length 1
× 3 movement takes the maximum value in window to obtain as output every timePond layer output, finally output drive row
For type, as shown in figure 3, convolutional neural networks include input layer, convolutional layer one, convolutional layer two, full articulamentum one, full articulamentum
Two and output layer, in convolutional neural networks, Batch value is 64, activation primitive is used as using using ReLU, using intersection entropy function
As loss function；
E, driving behavior is trained by convolutional neural networks, completes building for entire driving behavior identifying system,
Driving behavior is trained by convolutional neural networks, is trained by the way of stochastic gradient optimization.
The present embodiment is illustrated in conjunction with a driving behavior database.The database includes acceleration, brake, turning, lane change, surpasses
9 kinds of vehicle etc. act total 4495 driving behavior samples, specific signal such as table 1.It should be noted that each driving behavior sample
3-axis acceleration, three axis angular rate information when including vehicle movement.When building convolutional neural networks, the platform used is
TensorFlow。
1 driving behavior database of table
| Specimen types | Number of samples |
| Accelerate | 200 |
| Brake | 178 |
| Turn left | 599 |
| It turns right | 741 |
| It turns around | 802 |
| Left lane change | 555 |
| Right lane change | 694 |
| Continuous lane change | 480 |
| It overtakes other vehicles | 246 |
Data filtering, under true road conditions, the acquisition of driving behavior sample will receive influence of noise, first is that MEMS
System noise, second is that the noise as caused by vehicle vibration, when passing through deceleration strip or engine operation such as vehicle caused by shake.
In above two noise, system noise amplitude is much smaller compared to vehicle motion data, can be ignored；Shake noise
It is high-frequency noise compared to vehicle motion data, is to increase random bias on the basis of vehicle motion data, low pass can be used
Filter effectively filters out vehicle vibration noise.Low-pass filter filtering performance is as shown in Figure 2.Wherein, red curve represents vehicle
Longitudinal acceleration when acceleration, blue curve represent the output of low-pass filter, it can be seen that low-pass filter can be filtered effectively
Except data noise, retain data variation tendency.
Data format is regular, it is clear that the different driving behavior duration is different, leads to collected driving behavior sample
Length is different.Specific to the present embodiment, driving behavior data format is 6 rows × n column.Every row is respectively X-axis from top to bottom in 6 rows
Acceleration, Y-axis acceleration, Z axis acceleration, X-axis angular speed, Y-axis angular speed and Z axis angular speed；When the numerical value of n and movement continue
Between (number of sampling points) it is corresponding.Convolutional neural networks require the format of input data to be consistent, and in this embodiment, take
The mode of front and back zero padding adjusts data length to consistent.Specifically, the maximum driving behavior sample of n value is first looked for, and
It is L by its length mark, other samples is mended into [(L-n)/2] (being rounded downwards) and L- [(L-n)/2] a 0 in front and back respectively.Most
Eventually, all driving behavior sample formats are 6 rows × L column.For the driving behavior database in table one, the value of L is 2758.
Driving behavior identifies that in the present embodiment, the convolutional neural networks built are as shown in figure 3, by input layer, convolutional layer
One, convolutional layer two, full articulamentum one, full articulamentum two and output layer composition.Input the matrix that layer data is 6 × 2758；First
A convolutional layer has 32 convolution kernels, and convolution kernel then executes 1 × 2 pond having a size of 6 × 6；Second convolutional layer has 64 volumes
Product core, convolution kernel then execute 1 × 3 pond having a size of 6 × 6；First full articulamentum number of nodes is 1024；Second
Full articulamentum number of nodes is 256；Output layer number of nodes is 9, corresponding 9 kinds of driving behaviors to be identified.In the network, Batch
Value is 64, is instructed using entropy function is intersected as loss function using stochastic gradient optimization using ReLU as activation primitive
Practice.
The data for randomly selecting in table one 80% are used as verifying to collect as training set, remaining 20%.Under this arrangement, on
State the overall discrimination that convolutional neural networks have reached 89.6%.
Traditional driving behavior identification, using inertial sensors such as accelerometer and gyroscopes, can collect description vapour
Inertia sensing data needed for vehicle motion state.When handling these data, there are certain limits for traditional machine learning method
System, first is that system performance depends on the selection of characteristic vector, i.e., the data characteristics manually extracted is different, and obtained driving behavior is known
Other precision is also different；Second is that traditional machine learning method is had too many difficulties to cope with when handling large-scale dataset, cannot efficiently use
All sample informations.The present invention is based on the driving behavior systems of convolutional neural networks, and above-mentioned limitation is then not present, and can not be passed
The constraint of system statistical nature, can need automatically to extract feature according to system, identify field, convolutional Neural specific to driving behavior
Network not only but also can efficiently use big rule to avoid artificially extracting and screening feature in time domain, frequency domain or wavelet field
Mould driving behavior data set, identifies a greater variety of driving behaviors.Under true road conditions, the acquisition of driving behavior sample, which will receive, makes an uproar
Sound shadow is rung, first is that the system noise of MEMS, second is that the noise as caused by vehicle vibration, such as vehicle by deceleration strip or
Caused vibration when engine operation.In above two noise, system noise amplitude wants small very compared to vehicle motion data
It is more, it can be ignored；Shaking noise compared to vehicle motion data is high-frequency noise, is on the basis of vehicle motion data
Increase random bias, usable low-pass filter effectively filters out vehicle vibration noise.So by collected driving behavior sample
Data are filtered, then to carry out data format regular, are arranged at m row × n by filtered driving behavior sample data is regular
Matrix, to meet the input requirements of convolutional neural networks；Carry out driving behavior identification process again, will be regular after driving behavior
Sample data is input to the convolutional neural networks put up, exports as driving behavior type to be identified as input；Pass through volume
Product neural network is trained driving behavior, completes building for entire driving behavior identifying system.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not intended to limit the present invention the foregoing is merely a specific embodiment of the invention
Protection scope, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.
Claims (8)
1. a kind of building method of the driving behavior identifying system based on convolutional neural networks, which is characterized in that successively carry out with
Lower step:
A, driving behavior sample data is acquired；
B, data filtering analyzes the noise composition in driving behavior sample data, is filtered, is eliminated to data by filter
Influence of the noise to system；
C, data format is regular, by the regular matrix at m × n of filtered driving behavior sample data, to meet convolutional Neural
The input requirements of network；
D, driving behavior identify, will be regular after driving behavior sample data matrix be input to convolutional neural networks, to sample number
Pondization sampling is carried out according to matrix, 1 × 2 pond is first carried out, the concrete operations in pond are exactly with 1 × 2 window in m × n
Characteristic layer carries out the movement of step-length 1 × 2, and the maximum value in window is taken to obtain as output every timePond layer output, and
1 × 3 pond is executed afterwards, and the concrete operations in pond are exactly to exist with 1 × 3 windowCharacteristic layer carry out step-length 1 × 3
It is mobile, take the maximum value in window to obtain as output every timePond layer output, finally export driving behavior type；
E, driving behavior is trained by convolutional neural networks, completes building for entire driving behavior identifying system.
2. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, the driving behavior sample data in the step A includes acceleration information and angular velocity data.
3. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, the filter in the step B is low-pass filter.
4. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, the convolutional neural networks in the step D include input layer, convolutional layer one, convolutional layer two, full articulamentum one, complete
Articulamentum two and output layer.
5. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, in the convolutional neural networks in the step D, Batch value is 64.
6. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, in the convolutional neural networks in the step D, using use ReLU as activation primitive.
7. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, using intersection entropy function as loss function.
8. a kind of building method of driving behavior identifying system based on convolutional neural networks according to claim 1,
It is characterized in that, in the step E: being trained by convolutional neural networks to driving behavior, the side optimized using stochastic gradient
Formula is trained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810959518.6A CN109214438A (en) | 2018-08-22 | 2018-08-22 | A kind of building method of the driving behavior identifying system based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810959518.6A CN109214438A (en) | 2018-08-22 | 2018-08-22 | A kind of building method of the driving behavior identifying system based on convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109214438A true CN109214438A (en) | 2019-01-15 |
Family
ID=64988902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810959518.6A Pending CN109214438A (en) | 2018-08-22 | 2018-08-22 | A kind of building method of the driving behavior identifying system based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109214438A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378397A (en) * | 2019-06-27 | 2019-10-25 | 深圳大学 | A kind of driving style recognition methods and device |
| CN114492525A (en) * | 2022-01-25 | 2022-05-13 | 浙江大学 | Method for monitoring drunk driving by using wireless network and storage medium |
| CN115195757A (en) * | 2022-09-07 | 2022-10-18 | 郑州轻工业大学 | Electric bus starting driving behavior modeling and recognition training method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170032222A1 (en) * | 2015-07-30 | 2017-02-02 | Xerox Corporation | Cross-trained convolutional neural networks using multimodal images |
| CN106503673A (en) * | 2016-11-03 | 2017-03-15 | 北京文安智能技术股份有限公司 | A kind of recognition methodss of traffic driving behavior, device and a kind of video acquisition device |
| CN106611169A (en) * | 2016-12-31 | 2017-05-03 | 中国科学技术大学 | Dangerous driving behavior real-time detection method based on deep learning |
-
2018
- 2018-08-22 CN CN201810959518.6A patent/CN109214438A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170032222A1 (en) * | 2015-07-30 | 2017-02-02 | Xerox Corporation | Cross-trained convolutional neural networks using multimodal images |
| CN106503673A (en) * | 2016-11-03 | 2017-03-15 | 北京文安智能技术股份有限公司 | A kind of recognition methodss of traffic driving behavior, device and a kind of video acquisition device |
| CN106611169A (en) * | 2016-12-31 | 2017-05-03 | 中国科学技术大学 | Dangerous driving behavior real-time detection method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 王忠民等: "SF-CNN在驾驶行为识别中的应用研究", 《计算机工程与应用》 * |
| 王忠民等: "基于滑动窗特征融合的深信度网络驾驶行为识别", 《计算机应用研究》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378397A (en) * | 2019-06-27 | 2019-10-25 | 深圳大学 | A kind of driving style recognition methods and device |
| CN114492525A (en) * | 2022-01-25 | 2022-05-13 | 浙江大学 | Method for monitoring drunk driving by using wireless network and storage medium |
| CN115195757A (en) * | 2022-09-07 | 2022-10-18 | 郑州轻工业大学 | Electric bus starting driving behavior modeling and recognition training method |
| CN115195757B (en) * | 2022-09-07 | 2023-08-04 | 郑州轻工业大学 | Electric bus starting driving behavior modeling and recognition training method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109214438A (en) | A kind of building method of the driving behavior identifying system based on convolutional neural networks | |
| CN108717790B (en) | Vehicle travel analysis method based on checkpoint license plate recognition data | |
| CN110329271A (en) | A kind of multisensor vehicle driving detection system and method based on machine learning | |
| CN108280415A (en) | Driving behavior recognition methods based on intelligent mobile terminal | |
| CN106203330A (en) | A kind of vehicle classification method based on convolutional neural networks | |
| CN105160309B (en) | Three lanes detection method based on morphological image segmentation and region growing | |
| CN106570477A (en) | Vehicle model recognition model construction method based on depth learning and vehicle model recognition method based on depth learning | |
| CN107153871A (en) | Fall detection method based on convolutional neural networks and mobile phone sensor data | |
| EP2973211A1 (en) | Video stream evaluation | |
| CN104166548B (en) | Deep learning method based on Mental imagery eeg data | |
| CN105956625A (en) | Method and system for recognizing automobile motion state based on given physical model | |
| CN109376594A (en) | Visual perception method, apparatus, equipment and medium based on automatic driving vehicle | |
| CN105930565A (en) | Method for calibrating traffic simulation model parameters based on cross entropy algorithm of linear strategy | |
| CN103578227A (en) | Fatigue driving detection method based on GPS positioning information | |
| CN110194041A (en) | The adaptive bodywork height adjusting method of Multi-source Information Fusion | |
| CN103489010A (en) | Fatigue driving detecting method based on driving behaviors | |
| CN110378397A (en) | A kind of driving style recognition methods and device | |
| CN109871789A (en) | Vehicle checking method under a kind of complex environment based on lightweight neural network | |
| CN108694408A (en) | A kind of driving behavior recognition methods based on depth sparseness filtering convolutional neural networks | |
| CN108388871A (en) | A kind of vehicle checking method returned based on vehicle body | |
| CN110458214A (en) | Driver replaces recognition methods and device | |
| CN104200670A (en) | Vehicle model identifying method based on vehicle earth induction characteristics | |
| CN106485727A (en) | The dividing method of object and device in a kind of video | |
| CN110263836B (en) | Bad driving state identification method based on multi-feature convolutional neural network | |
| Silva et al. | Finding manoeuvre motifs in vehicle telematics |
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: 20190115 |