CN109284699A - A kind of deep learning method being applicable in vehicle collision - Google Patents
A kind of deep learning method being applicable in vehicle collision Download PDFInfo
- Publication number
- CN109284699A CN109284699A CN201811026630.0A CN201811026630A CN109284699A CN 109284699 A CN109284699 A CN 109284699A CN 201811026630 A CN201811026630 A CN 201811026630A CN 109284699 A CN109284699 A CN 109284699A
- Authority
- CN
- China
- Prior art keywords
- layer
- vehicle
- data
- deep learning
- matrix
- 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 OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to car crass detection fields, and in particular to a kind of deep learning method suitable for vehicle collision includes the following steps: S1: recording the travelling data of current vehicle and aforementioned vehicle by sensor, and is uploaded to cloud platform；S2: the prediction model based on convolutional neural networks is established according to the data of S1；S3: obtaining the travelling data of current vehicle, and is uploaded to the prediction model based on convolutional neural networks；S4: judge whether to collide according to the prediction model based on convolutional neural networks that S2 is established.The present invention overcomes the technological deficiency that the data sampling frequency of existing method is high, computation complexity is high, provides a kind of deep learning method suitable for vehicle collision.
Description
Technical field
The present invention relates to car crass detection fields, and in particular to a kind of deep learning method suitable for vehicle collision.
Background technique
With the fast development of internet, car networking also becomes increasingly to weigh as a kind of important means of intelligent transportation
It wants.Car networking is the Internet being made of information such as vehicle location, speed and routes, and vehicle passes through GPS, RFID, sensor
The acquisition of itself environment and status information can be completed with camera etc., then by dividing information of vehicles computer technology
Analysis and processing, finally calculate the best route of different vehicle and report without delay road conditions.Car networking system is by vehicle instrument
Table platform installs vehicle-mounted terminal equipment, realizes to the acquisition of vehicle all working and quiet multidate information, storage and sends, car networking one
As all have the function of outdoor scene, can use mobile network realize people-car interaction.
Existing vehicle net system mainly passes through vehicle collision avoidance algorithm and adopts respectively in both the horizontal and vertical directions
Judge whether to collide with prediction algorithm prediction.But the data sampling frequency of existing method is high, computation complexity
It is high.
Summary of the invention
The present invention overcomes the data sampling frequency of the learning method of above-mentioned existing vehicle collision height, computation complexity are high
Technological deficiency provides a kind of deep learning method suitable for vehicle collision.
In order to solve the above technical problems, technical scheme is as follows:
A kind of deep learning method suitable for vehicle collision, includes the following steps:
S1: the travelling data of current vehicle and aforementioned vehicle is recorded by sensor, and is uploaded to cloud platform；
S2: the prediction model based on convolutional neural networks is established according to the data of S1；
S3: obtaining the travelling data of current vehicle, and is uploaded to the prediction model based on convolutional neural networks；
S4: judge whether to collide according to the prediction model based on convolutional neural networks that S2 is established.
S1 records the travelling data of current vehicle and aforementioned vehicle, and the travelling data for the aforementioned vehicle for needing to obtain is at least
The travelling data of 50 aforementioned vehicles, the G-sensor sensor that the sensor is 50 hertz, for recording vehicle in cross
To the stress in, longitudinal, vertical three directions.
The prediction model based on convolutional neural networks of S2, convolutional neural networks include four layers, respectively input layer, convolution
Layer, pond layer and output layer, the data that the data of the input layer are acquired from sensor；The convolutional layer is to input layer
Data carry out convolution；The pond layer screens to obtain matrix to the matrix that convolutional layer obtains.
S1 is the value that the stress in lateral, longitudinal, vertical three directions is obtained by sensor, wherein lateral stress value table
Show that the stress of vehicle horizontal direction, longitudinal value indicate that vehicle stress in the front-back direction, vertical value indicate vehicle up and down direction
Stress.
Prediction model of the S3 based on convolutional neural networks, which is characterized in that the first layer is the square that input layer is 50*4
Battle array；The second layer is convolutional layer, is multiplied using the convolution matrix of 5*5*4 with input layer, and the matrix of 5*10*1 is obtained；Third layer is pond
Change layer, the matrix that the second layer is obtained passes through maximum pond；4th layer is output layer, judges hair according to the matrix that third layer obtains
The probability of raw collision.
S4 judges the grade of collision according to the size of g value according to the prediction model of convolutional neural networks,
Output layer uses logistic function, is greater than threshold value for judging whether, if more than threshold value, then judgement collides
And triggering collision warning；Otherwise, judgement does not collide and does not trigger alarm.
Compared with prior art, the beneficial effect of technical solution of the present invention is: overcoming prior art data acquiring frequency
The technological deficiency of high, computation complexity height, the situation that needs to collide in view of different directions, improves the effect of vehicle collision prediction
Rate and accuracy.
Detailed description of the invention
Fig. 1 is flow chart when vehicle of the present invention collides；
Fig. 2 is convolutional neural networks prediction model of the present invention.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent；
In order to better illustrate this embodiment, the certain components of attached drawing have omission, zoom in or out, and do not represent actual product
Size；
To those skilled in the art, it is to be understood that certain known features and its explanation, which may be omitted, in attached drawing
's.
The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
Embodiment 1:
As shown in Figure 1, a kind of deep learning method suitable for vehicle collision, includes the following steps:
S1: the travelling data of current vehicle and aforementioned vehicle is recorded by sensor, and is uploaded to cloud platform；
S2: the prediction model based on convolutional neural networks is established according to the data of S1；
S3: obtaining the travelling data of current vehicle, and is uploaded to the prediction model based on convolutional neural networks；
S4: judge whether to collide according to the prediction model based on convolutional neural networks that S2 is established.
S1 records the travelling data of current vehicle and aforementioned vehicle, and the travelling data for the aforementioned vehicle for needing to obtain is at least
The travelling data of 50 aforementioned vehicles, the G-sensor sensor that the sensor is 50 hertz, for recording vehicle in cross
To the stress in, longitudinal, vertical three directions.50 hertz of G-sensor sensor record in real time vehicle in the horizontal direction, front and back
The stress data in three direction, up and down direction directions, namely 150 data sample points of record.
Wherein, based on the prediction model of convolutional neural networks, convolutional neural networks include four layers, respectively input layer, volume
Lamination, pond layer and output layer, the data that the data of the input layer are acquired from sensor；The convolutional layer is to input layer
Data carry out convolution；The pond layer screens to obtain matrix to the matrix that convolutional layer obtains.Wherein, input layer uses 50 rows 4
The matrix of column, convolutional layer carry out convolution to the matrix of input layer 50*4, and Convolution uses the matrix of a 5*4, by first layer
The matrix multiple of the matrix of input layer 50*4 and the second convolutional layer 5*4, obtains the matrix of a 10*1 size.Use different 5*
4 convolution matrix, is repeated 5 times, and may finally obtain the matrix of 5*10*1, and wherein the parameter of convolution matrix can be by trained
It arrives；The pond layer size of third layer can amount to 5, each matrix by the matrix-split of 5*10*1 at the matrix of 10*1 with 10*1
A maximum value is selected, the matrix of second layer 5*10*1 can finally obtain the matrix of 5*1 by pond layer；It is defeated at the 4th layer
Layer out, by the matrix of third layer 5*1, to calculate threshold value.
S1 is the value that the stress in lateral, longitudinal, vertical three directions is obtained by sensor, wherein lateral stress value table
Show that the stress of vehicle horizontal direction, longitudinal value indicate that vehicle stress in the front-back direction, vertical value indicate vehicle up and down direction
Stress.
Prediction model of the S3 based on convolutional neural networks, which is characterized in that the first layer is the square that input layer is 50*4
Battle array；The second layer is convolutional layer, is multiplied using the convolution matrix of 5*5*4 with input layer, and the matrix of 5*10*1 is obtained；Third layer is pond
Change layer, the matrix that the second layer is obtained passes through maximum pond；4th layer is output layer, judges hair according to the matrix that third layer obtains
The probability of raw collision.
S4 judges the grade of collision according to the size of g value according to the prediction model of convolutional neural networks,The stress condition of level, front and back, up and down direction is measured using g value, with it is collected by
Whether the size of force value is more than the threshold value of sensor settings, to determine whether colliding, is reported to cloud platform.
Output layer uses logistic function, is greater than threshold value for judging whether, if more than threshold value, then judgement collides
And triggering collision warning；Otherwise, judgement does not collide and does not trigger alarm.
The terms describing the positional relationship in the drawings are only for illustration, should not be understood as the limitation to this patent；
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (7)
1. a kind of deep learning method suitable for vehicle collision, which comprises the steps of:
S1: the travelling data of current vehicle and aforementioned vehicle is recorded by sensor, and is uploaded to cloud platform；
S2: the prediction model based on convolutional neural networks is established according to the data of S1；
S3: obtaining the travelling data of current vehicle, and is uploaded to the prediction model based on convolutional neural networks；
S4: judge whether to collide according to the prediction model based on convolutional neural networks that S2 is established.
2. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the S1 record is worked as
The travelling data of vehicle in front and aforementioned vehicle, the travelling data for the aforementioned vehicle for needing to obtain are at least the row of 50 aforementioned vehicles
Car data, the G-sensor sensor that the sensor is 50 hertz, for recording vehicle lateral, longitudinal, three vertical
The stress in direction.
3. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the S2 based on
The prediction model of convolutional neural networks, convolutional neural networks include four layers, respectively input layer, convolutional layer, pond layer and output
Layer, the data that the data of the input layer are acquired from sensor；The convolutional layer carries out convolution to the data of input layer；Institute
It states pond layer the matrix that convolutional layer obtains is screened to obtain matrix.
4. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the S1 is to pass through
Sensor obtains the value of the stress in lateral, longitudinal, vertical three directions, wherein lateral stress value indicates vehicle horizontal direction
Stress, longitudinal value indicate that vehicle stress in the front-back direction, vertical value indicate the stress of vehicle up and down direction.
5. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the S3 is based on volume
The prediction model of product neural network, which is characterized in that the first layer is the matrix that input layer is 50*4；The second layer is convolution
Layer, is multiplied with input layer using the convolution matrix of 5*5*4, obtains the matrix of 5*10*1；Third layer is pond layer, and the second layer is obtained
The matrix obtained passes through maximum pond；4th layer is output layer, and the matrix obtained according to third layer judges the probability to collide.
6. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the S4 is according to volume
The prediction model of product neural network, the grade of collision is judged according to the size of g value,
7. the deep learning method according to claim 1 suitable for vehicle collision, which is characterized in that the output layer is adopted
With logistic function, it is greater than threshold value for judging whether, if more than threshold value, then judgement collides and triggers collision warning；
Otherwise, judgement does not collide and does not trigger alarm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811026630.0A CN109284699A (en) | 2018-09-04 | 2018-09-04 | A kind of deep learning method being applicable in vehicle collision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811026630.0A CN109284699A (en) | 2018-09-04 | 2018-09-04 | A kind of deep learning method being applicable in vehicle collision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109284699A true CN109284699A (en) | 2019-01-29 |
Family
ID=65183994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811026630.0A Pending CN109284699A (en) | 2018-09-04 | 2018-09-04 | A kind of deep learning method being applicable in vehicle collision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109284699A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111846046A (en) * | 2019-04-03 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | System, method and device for detecting safety of bicycle |
| CN111982256A (en) * | 2020-08-18 | 2020-11-24 | 上海陆杰电子科技有限公司 | Automobile balance pressure sensor fault early warning method based on multichannel CNN model |
| CN115398496A (en) * | 2020-02-06 | 2022-11-25 | 美光科技公司 | Artificial intelligence-based persistence of vehicle black box data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934876A (en) * | 2017-03-16 | 2017-07-07 | 广东翼卡车联网服务有限公司 | A kind of recognition methods of vehicle abnormality driving event and system |
| US20170262750A1 (en) * | 2016-03-11 | 2017-09-14 | Panasonic Intellectual Property Corporation Of America | Risk prediction method |
| 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 |
| CN107703937A (en) * | 2017-09-22 | 2018-02-16 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and its conflict evading method based on convolutional neural networks |
| CN107972662A (en) * | 2017-10-16 | 2018-05-01 | 华南理工大学 | To anti-collision warning method before a kind of vehicle based on deep learning |
| US10007269B1 (en) * | 2017-06-23 | 2018-06-26 | Uber Technologies, Inc. | Collision-avoidance system for autonomous-capable vehicle |
| CN108268866A (en) * | 2016-12-30 | 2018-07-10 | 纵目科技（上海）股份有限公司 | A kind of vehicle checking method and system |
| US20180211403A1 (en) * | 2017-01-20 | 2018-07-26 | Ford Global Technologies, Llc | Recurrent Deep Convolutional Neural Network For Object Detection |
-
2018
- 2018-09-04 CN CN201811026630.0A patent/CN109284699A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170262750A1 (en) * | 2016-03-11 | 2017-09-14 | Panasonic Intellectual Property Corporation Of America | Risk prediction method |
| CN108268866A (en) * | 2016-12-30 | 2018-07-10 | 纵目科技（上海）股份有限公司 | A kind of vehicle checking method and system |
| US20180211403A1 (en) * | 2017-01-20 | 2018-07-26 | Ford Global Technologies, Llc | Recurrent Deep Convolutional Neural Network For Object Detection |
| CN106934876A (en) * | 2017-03-16 | 2017-07-07 | 广东翼卡车联网服务有限公司 | A kind of recognition methods of vehicle abnormality driving event and system |
| US10007269B1 (en) * | 2017-06-23 | 2018-06-26 | Uber Technologies, Inc. | Collision-avoidance system for autonomous-capable vehicle |
| 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 |
| CN107703937A (en) * | 2017-09-22 | 2018-02-16 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and its conflict evading method based on convolutional neural networks |
| CN107972662A (en) * | 2017-10-16 | 2018-05-01 | 华南理工大学 | To anti-collision warning method before a kind of vehicle based on deep learning |
Non-Patent Citations (3)
| Title |
|---|
| JUNGHWAN PYO: "Front collision warning based on vehicle detection using CNN", 《2016 INTERNATIONAL SOC DESIGN CONFERENCE (ISOCC)》 * |
| 史周华: "《中医药统计学与软件应用》", 31 December 2017, 中国中医药出版社 * |
| 杨晓康: "基于毫米波雷达和视觉的车辆检测识别方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111846046A (en) * | 2019-04-03 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | System, method and device for detecting safety of bicycle |
| CN115398496A (en) * | 2020-02-06 | 2022-11-25 | 美光科技公司 | Artificial intelligence-based persistence of vehicle black box data |
| CN111982256A (en) * | 2020-08-18 | 2020-11-24 | 上海陆杰电子科技有限公司 | Automobile balance pressure sensor fault early warning method based on multichannel CNN model |
| CN111982256B (en) * | 2020-08-18 | 2021-11-16 | 上海陆杰电子科技有限公司 | Automobile balance pressure sensor fault early warning method based on multichannel CNN model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Saiprasert et al. | Driver behaviour profiling using smartphone sensory data in a V2I environment | |
| CN106570444B (en) | The vehicle intelligent based reminding method and system of Behavior-based control identification | |
| CN109284699A (en) | A kind of deep learning method being applicable in vehicle collision | |
| Fox et al. | Crowdsourcing undersampled vehicular sensor data for pothole detection | |
| CN108830488B (en) | A road area risk assessment method | |
| Zhang et al. | Comparative analysis of pedestrian, bicycle and car traffic moving in circuits | |
| CN106056903B (en) | The detection method in the road congestion region based on GPS data | |
| Yijing et al. | Intelligent algorithms for incident detection and management in smart transportation systems | |
| JP6776373B2 (en) | Methods, devices, and systems for detecting reverse-way drivers | |
| CN105869402B (en) | Speed correction method of expressway section based on multi-type floating car data | |
| Chaovalit et al. | A method for driving event detection using SAX on smartphone sensors | |
| Aichinger et al. | Using low-cost smartphone sensor data for locating crash risk spots in a road network | |
| Zhao et al. | Modeling lateral interferences between motor vehicles and non-motor vehicles: A survival analysis based approach | |
| Sharma et al. | S-road assist: Road surface conditions and driving behavior analysis using smartphones | |
| Schendzielorz et al. | Infrastructure-based vehicle maneuver estimation at urban intersections | |
| CN109979194A (en) | Heavy vehicle ramp based on Markov chain is creeped Activity recognition method | |
| CN116434523B (en) | Vehicle active safety control method and device based on constraint degree in information perception scene | |
| CN115063976B (en) | Vehicle conflict risk assessment and prediction method based on multi-channel convolutional neural network | |
| Lu et al. | Quantifying the severity of traffic conflict by assuming moving elements as rectangles at intersection | |
| CN118884442B (en) | Flood event detection method and device, storage medium and electronic equipment | |
| Das et al. | Why slammed the brakes on? auto-annotating driving behaviors from adaptive causal modeling | |
| CN116580564B (en) | Traffic flow prediction method and device | |
| CN107256382A (en) | Virtual bumper control method and system based on image recognition | |
| Nam et al. | Traffic density estimation using radar sensor data from probe vehicles | |
| Oh et al. | Anonymous vehicle tracking for real-time traffic surveillance and performance on signalized arterials |
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: 20190129 |