CN108596146A - Road multi-target classification method - Google Patents
Road multi-target classification method Download PDFInfo
- Publication number
- CN108596146A CN108596146A CN201810438507.3A CN201810438507A CN108596146A CN 108596146 A CN108596146 A CN 108596146A CN 201810438507 A CN201810438507 A CN 201810438507A CN 108596146 A CN108596146 A CN 108596146A
- Authority
- CN
- China
- Prior art keywords
- target
- variable
- bayesian network
- road
- discrete
- 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
- 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
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
Abstract
The invention discloses a kind of road Multi-Target Classification Methods, are learnt to Hybrid Bayesian Network structure using the NPC algorithms based on constraint by importing training sample；Then in network structure discrete variable and continuous variable carry out parameter learning respectively and merged to obtain the distribution of each node in network, then by parameter, finally by test sample for Bayesian network reasoning and road target is classified.On the one hand demand that this method has been abandoned to high-resolution and close shot image greatly reduces calculation amount by using the simple low level feature of road target.The structure of another aspect Hybrid Bayesian Network structure, which avoids, to be all considered as discrete variable by all variables in traditional BAYESIAN NETWORK CLASSIFIER and easily causes target information loss, while leading to search space and calculation amount sharply increase in the processing and analysis of road Multi-target Data.The present invention more meets reality using the Bayesian network that continuous nodes and discrete nodes coexist.
Description
Technical field
The present invention relates to a kind of intelligent Video Surveillance Technology, more particularly to a kind of road Multi-Target Classification Method.
Background technology
The research of intelligent automobile is in the ascendant, and Environment identification is that the basic module of intelligent automobile and intelligent vehicle are carried out certainly
The premise of main driving.One basic task of Environment identification is detection and identification to barrier.Existing research is laid particular emphasis on pair
The detection of single type barrier, and it is less to the research for detecting and identifying of polymorphic type barrier.And for travelling in city
For intelligent vehicle under area's road environment, not only objects ahead is detected, and also to recognize their type.
Grader general at present has decision tree, neural network, support vector machines, Adaboost etc..Its categorised decision according to
According to both from the study to sample data, therefore need to be support with a large amount of class sample.Diversity, ring in view of targeted species
The complexity in border and the uncertainty of shape only rely on the classifying rules obtained to sample learning by sample size and space point
Cloth is affected.And Bayesian network then has the ability to express of powerful uncertain knowledge, it, can during categorised decision
Information of both making full use of priori and statistical learning, keeps inference rule more flexible and effective, this is but also in number
Effective grader can be still established in the case of according to missing or no sample data.But traditional Bayesian network point
Class device is all variables to be all considered as discrete variable, but inevitably existence information lacks the sliding-model control of variable,
And in the processing and analysis of road Multi-target Data, the discretization of continuous variable can lead to search space and calculation amount drastically
Increase.Therefore road multiple target classification problem is still a problem in personal vehicle system.
Invention content
In order to solve above-mentioned problem, not only there are discrete nodes but also there are the mixing shellfishes of continuous nodes the present invention provides a kind of
The road Multi-Target Classification Method of this network of leaf.
To achieve the above object, present invention employs following technical solutions：
A kind of road Multi-Target Classification Method, this method are：By training sample using the NPC algorithms based on constraint to mixing
Bayesian network structure is learnt, in network structure discrete variable and continuous variable carry out respectively parameter learning obtain net
The distribution of each node in network merges two class parameters, and test sample is finally used for the reasoning of Bayesian network simultaneously
Road multiple target is classified.
Specifically comprise the following steps：
1) data set is established, includes training dataset for training sorting algorithm model and for testing sorting algorithm
Test data set；
2) target classification classification to be identified is set as m classes, extracts n characteristic of division of target, characteristic of division is divided into
Two class of discrete variable and continuous variable；
AmnIt is the two-dimensional matrix for including each feature all categories information：
Wherein ai,jFor the value of i-th j-th of feature of class target, by { A1,A2,…,AnAs phase in Hybrid Bayesian Network
Answer the value of node；
3) training dataset is imported, using the Structure learning method NPC algorithms based on constraint to Hybrid Bayesian Network
Structure is learnt；Distribution P (the X of each node are obtained using Bayesian network parameters learning methodi| C), for discrete
Variable and continuous variable carry out parameter learning respectively；
Two class parameters of above-mentioned acquisition are merged, Hybrid Bayesian Network is obtained；
4) test data set is imported in the Hybrid Bayesian Network obtained, is classified to urban road target.
In above-mentioned technical proposal, further, the characteristic of division described in the step 2) is the circumstance of occlusion of target
(occlude), length (length), width (width), height (height) and five features of observation angle (alpha),
In, the circumstance of occlusion of target is discrete variable, and length, width, height and the observation angle of target are continuous variable.
Further, the relationship between node is determined in the step 3) by dependency analysis, if there is dependence between node
Relationship then retains this nonoriented edge, otherwise removes this edge between node；Specifically, by chi-square statistics amount do hypothesis testing come
The presence or absence on side is determined, so that it is determined that the frame of Bayesian network, then true further according to the segmentation collection generated in independence test
The direction of deckle；Progress can be interacted when seeming perverse or analytic fuzzy region for any connection with user, user can be with
The directionality of undirected connection is determined using this chance and solves its fuzzy region.
Further, for discrete variable, conditional probability table CPT is indicated using multi-dimensional matrix, is：E(θi,j,k|D,BS,
ξ)
=(Ni,j,k+1)(Ni,j+ri-Ni,j,k-1)/(Ni,j+ri)2(Ni,j+ri+1)
Wherein, θi,j,kIndicate conditional probability table；ξ indicates several hypothesis；D is data set；BSFor network structure；riIt is discrete
Stochastic variable XiAll possible value number, if using ωi,jIndicate variable XiJ-th of father node, vi,kFor variable XiTake
It is worth, then Ni,j,kFor variable X in data set DiValue is vi,kFather node is ω simultaneouslyi,jSample occur number, Ni,jCalculating
Formula is
For each feature of discrete variable, parameter learning the result is that the two-dimensional matrix of m × k size：M1=
CPTm×k；
For continuous variable, conditional probability distribution is：
Wherein, C represents class label, and continuous variable meets unitary normal distribution, i.e. Xi~N (μi,σi), normal distribution
Mean μiAnd variances sigmaiTwo important parameters are directly calculated from training sample；When sample data is incomplete, Ke Yitong
EM algorithms are crossed to solve the problem concerning study of Hybrid Bayesian Network parameter；
For each feature of continuous variable, parameter learning the result is that the two-dimensional matrix M of a size of m × 22=
CPDm×2=<μi, σi>,i∈{1,2,…,m}。
To two class parameter M1And M2Merge, obtain parameter network θ=<M1, M2>；Wherein, M1For the condition of m × k sizes
Probability tables CPTm×kTwo-dimensional matrix；M2For the conditional probability distribution CPD of the sizes of m × 2m×2Two-dimensional matrix.
The beneficial effects of the invention are as follows：
According to the road Multi-Target Classification Method of the present invention based on Hybrid Bayesian Network, by based on constraint
NPC algorithms learn the structure of Hybrid Bayesian Network, then carry out parameter learning respectively to discrete variable and continuous variable
The distribution for obtaining each node in network, then merges parameter, and test sample is finally used for the network by city
Road target is classified.On the one hand it has abandoned the demand to high-resolution and close shot image, by using road target letter
Single low level feature, such as：Highly, width and observation angle etc. greatly reduce calculation amount, and can real time execution.It is another
The structure of aspect Hybrid Bayesian Network structure avoid in traditional BAYESIAN NETWORK CLASSIFIER by all variables be all considered as from
Variable is dissipated, target information will be caused to lose in this way, while leading to search space in the processing and analysis of road Multi-target Data
And calculation amount sharply increase.And the Bayesian network that continuous nodes and discrete nodes coexist just more meets reality.
Description of the drawings
Fig. 1 is the logic schematic diagram of the embodiment of the present invention；
Fig. 2 is the Hybrid Bayesian Network structural schematic diagram of the embodiment of the present invention.
Specific implementation mode
For a further understanding of the present invention, the preferred embodiment of the invention is described with reference to embodiment, still
It should be appreciated that these descriptions are only the feature and advantage further illustrated the present invention, rather than to the claims in the present invention
Limitation.
A kind of road Multi-Target Classification Method is present embodiments provided, as shown in Fig. 1, by importing training sample application
NPC algorithms based on constraint learn Hybrid Bayesian Network structure；Then in network structure discrete variable and company
Continuous variable carries out parameter learning and is merged to obtain the distribution of each node in network, then by parameter respectively, will finally survey
Sample sheet is used for the reasoning of Bayesian network and road target is divided into eight classes.Specific implementation step is as follows：
1) it is tested using the 3D target detection reference data set pair one's duty class methods of KITTI.The data set is all
True roadway scene, data acquisition scenarios are abundant, and eight class barriers are marked out by professional mark personnel in every frame point cloud,：
Pedestrian, Car, Van, Truck, Cyclist, Person_sitting, Tram and Misc.And it is also various degrees of
It blocks and blocks.It is collated that entire data set is divided into two parts --- 1 training dataset and 1 test data set.Its
Middle training dataset is used for the training of algorithm model, about 20000 total containing target obstacle；Test data set is used for sorting algorithm
Test, containing target obstacle altogether about 20570.
2) it is as shown in table 1 to recognize target design two sets of plan, target segments are 8 classes by scheme one, respectively pedestrian, seat
People, car, truck, lorry, bicyclist, electric car and the hybrid vehicle.Target rough segmentation is 3 classes by scheme two, i.e., by scheme one
In the people being seated be merged into pedestrian's classification, lorry, truck are merged into car classification, and tramcar and hybrid car are neglected
Slightly.
Table 1 recognizes target protocol
3) experiment is carried out under Hugin expert platforms, is chosen training data and is concentrated 20000 road target information
Training sample of the data as Bayes classifier.Which includes 14035 cars (Car), 2263 pedestrians
(Pedestrian), 1461 lorries (Van), 528 trucies (Truck), 834 bicyclists (Cyclist), 117 are seated
People (Person_sitting), 265 electric cars (Tram) and 497 hybrid vehicles (Misc).
3) structure of Bayesian network is vital on the levels of precision of model.Learn the best of Bayesian network
Structure needs the exponential time, because the number of the largely possible structure of one group of given node is super index in the number of node
's.We learn network structure using the Structure learning method NPC algorithms based on constraint, and the priori of all variables is arranged
Probability distribution randomly generates, and using the statistic of chi square distribution structural environment independence test, significance is set as 0.05.EM
Parameter learning iterations are set as 10, using bayesian information criterion (BIC) function as scoring functions.The mixing pattra leaves of generation
This network structure is as shown in Figure 2:For the Hybrid Bayesian Network model of five layers of six node composition.Each node is detailed in figure
Carefully it is described as follows：It is assumed that variable C is discrete variable, value is all possible target type, is indicated with root node, to be identified
Target type it is as shown in table 1, observational variable is the feature for the moving target that laser radar and visual sensor observe, with son
Node indicates.Respectively discrete variable occlude (whether target blocks), continuous variable length (target length), width
(target width), height (object height) and alpha (target observation angle, range：- π~π).
3) the step of Hybrid Bayesian Network parameter learning is as follows：
(1) from training sample set D={ D1,D2,…,DmIn obtain road target character subset X=X1, X2 ...,
Xn}；
(2) target signature is extracted：AmnIt is the two-dimensional matrix for including each feature all categories information：
Wherein ai,jIt is general by { A for the value of i-th j-th of feature of class target1,A2,…,AnAs respective nodes in network
Value；
(3) the class label C={ C of Different categories of samples are obtained1,C2,…,Cm}；
(4) types of variables for confirming each node, i.e., be divided into discrete nodes collection X by XdiscreteWith continuous nodes collection
XcontinualTwo subsets；
(5) to XdiscreteSliding-model control is done by the method for discrete Bayesian network；
(6) parameter learning for carrying out discrete nodes and continuous nodes respectively, to XdiscreteIn each feature, parameter learning
The result is that the two-dimensional matrix M of m × k size1=CPTmk(conditional probability table).To XcontinualIn each feature, parameter
Study the result is that the two-dimensional matrix M of a size of m × 22=CPD=<μi, σi>, i ∈ 1,2 ..., and m } (conditional probability point
Cloth).
(7) by parameter M1And M2Merge, obtain parameter network θ=<M1, M2>。
4) statistics obtains the mean value of continuous variable characteristic value and variance and the conditional probability of discrete variable in training sample
Table.
5) test sample is input in trained network structure and is classified to test target.
Test data set is made of 20570 test targets, including 8 targets in 1 scheme one of table, respectively by
14707 cars, 2224 pedestrians, 1453 lorries, 566 trucies, 793 bicyclists, 105 people being seated, 246
Electric car and 476 hybrid railcars at.The confusion matrix of generation is as shown in table 2, it shows of the state and prediction that observe
With degree, error rate 4.08%.The first row illustrates that 2224 pedestrian samples have 2213 correctly to be classified, and 6 mistakes are divided into mixed
Vehicle is closed, 5 mistakes are divided into the people being seated.The accuracy rate of class object and the comparison of other methods are as shown in table 3.It can be seen that we
Method is preferable to the classifying quality of pedestrian, automobile, electric car and bicyclist, and to the classification accuracy of lorry, truck and hybrid vehicle
It is relatively low.And this method is applied to 3 targets in 1 scheme two of table, the confusion matrix of generation is as shown in table 4, and error rate is
0.08%, 16727 samples of automotive-type are all correctly classified.Characteristic of division is apparent in this way, so nicety of grading obviously carries
It is high.
2 eight classification target confusion matrix of table
The comparison of 3 accuracy rate of table (Precision) and other methods
Involved other methods are referring specifically to Subsequent literature in this table.
The confusion matrix of 4 tertiary target of table
[1] intelligent vehicle multiclass obstacle identification [J] the computer engineering of Shen Zhixi, yellow seat Yue etc. based on Boosting
.2009,35(14):241-242
[2]Silviu Bota,Sergiu Nedevschi.Matthias Konig.A Framework for Object
Detection,Tracking and Classification in Urban Traffic Scenarios Using
Stereovision[J].IEEE International Conference on Intelligent Computer
Communication&Processing,2009:153-156.
[3]Chug-Wei Liang and Chia-Feng Juang.Moving Object Classification
Using a Combination of Static Appearance Features and Spatial and Temporal
Entropy Values of Optical Flows[J].IEEE TRANSACTIONS OF INTELLIGENT
TRANSPORTATION SYSTEMS,2015,16(6):3453-3464.
[4]Mehran Kafai and Bir Bhanu.Dynamic Bayesian Networks for Vehicle
Classification in Video[C].IEEE transactions on industrial informatics,2012,
VOL.8,NO.1:100-109.
The explanation of above example is only intended to facilitate the understanding of the method and its core concept of the invention.It should be pointed out that pair
For those skilled in the art, without departing from the principle of the present invention, the present invention can also be carried out
Some improvements and modifications, these improvement and modification are also fallen within the protection scope of the claims of the present invention.
Claims (7)
1. a kind of road Multi-Target Classification Method, which is characterized in that this method is：Training sample is applied into the NPC based on constraint
Algorithm learns Hybrid Bayesian Network structure, in network structure discrete variable and continuous variable carry out parameter respectively
Study obtains the distribution of each node in network, and two class parameters are merged, and test sample is finally used for Bayesian network
The reasoning of network simultaneously classifies road multiple target.
2. road Multi-Target Classification Method according to claim 1, which is characterized in that specifically comprise the following steps：
1) data set is established, includes the training dataset for training sorting algorithm model and the survey for testing sorting algorithm
Try data set；
2) target classification classification to be identified is set as m classes, extracts n characteristic of division of target, characteristic of division is divided into discrete
Two class of variable and continuous variable；
AmnIt is the two-dimensional matrix for including each feature all categories information：
Wherein ai,jFor the value of i-th j-th of feature of class target, by { A1,A2,…,AnBe used as in Hybrid Bayesian Network and accordingly save
The value of point；
3) training dataset is imported, using the Structure learning method NPC algorithms based on constraint to the structure of Hybrid Bayesian Network
Learnt；Distribution P (the X of each node are obtained using Bayesian network parameters learning methodi| C), for discrete variable
Parameter learning is carried out respectively with continuous variable；
Two class parameters of above-mentioned acquisition are merged, Hybrid Bayesian Network is obtained；
4) test data set is imported in the Hybrid Bayesian Network obtained, is classified to urban road target.
3. road Multi-Target Classification Method according to claim 1, which is characterized in that point described in the step 2)
Category feature is circumstance of occlusion (occlude), length (length), width (width), height (height) and the view angle of target
Spend (alpha) five features, wherein the circumstance of occlusion of target is discrete variable, length, width, height and the view angle of target
Degree is continuous variable.
4. road Multi-Target Classification Method according to claim 1, which is characterized in that by card side in the step 3)
Statistic does hypothesis testing to determine the presence or absence on side, so that it is determined that the frame of Bayesian network, then further according to independent inspection
The segmentation collection for testing middle generation determines the direction on side；It determines the directionality of undirected connection in addition, can be interacted by user and solves it
Fuzzy region.
5. road Multi-Target Classification Method according to claim 1, which is characterized in that for discrete in the step 3)
Variable, conditional probability table CPT is indicated using multi-dimensional matrix, is：
E(θi,j,k|D,BS, ξ) and=(Ni,j,k+1)(Ni,j+ri-Ni,j,k-1)/(Ni,j+ri)2(Ni,j+ri+1)
Wherein, θi,j,kIndicate conditional probability table；ξ indicates several hypothesis；D is data set；BSFor network structure；riFor Discrete Stochastic
Variable XiAll possible value number, if using ωi,jIndicate variable XiJ-th of father node, vi,kFor variable XiValue,
Then Ni,j,kFor variable X in data set DiValue is vi,kFather node is ω simultaneouslyi,jSample occur number, Ni,jCalculating it is public
Formula is
For each feature of discrete variable, parameter learning the result is that the two-dimensional matrix of m × k size：M1=
CPTm×k；
6. road Multi-Target Classification Method according to claim 1, which is characterized in that for continuous in the step 3)
Variable, conditional probability distribution are：
Wherein, C represents class label, and continuous variable meets unitary normal distribution, i.e. Xi~N (μi,σi), normal distribution it is equal
Value μiAnd variances sigmaiTwo important parameters are directly calculated from training sample；When sample data is incomplete, EM can be passed through
Algorithm solves the problem concerning study of Hybrid Bayesian Network parameter；
For each feature of continuous variable, parameter learning the result is that the two-dimensional matrix M of a size of m × 22=CPDm×2=<
μi, σi>,i∈{1,2,…,m}。
7. road Multi-Target Classification Method according to claim 1, which is characterized in that join to two classes in the step 3)
Number M1And M2Merge, obtain parameter network θ=<M1, M2>；Wherein, M1For the conditional probability table CPT of m × k sizesm×kTwo
Tie up matrix；M2For the conditional probability distribution CPD of the sizes of m × 2m×2Two-dimensional matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810438507.3A CN108596146A (en) | 2018-05-09 | 2018-05-09 | Road multi-target classification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810438507.3A CN108596146A (en) | 2018-05-09 | 2018-05-09 | Road multi-target classification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108596146A true CN108596146A (en) | 2018-09-28 |
Family
ID=63636578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810438507.3A Pending CN108596146A (en) | 2018-05-09 | 2018-05-09 | Road multi-target classification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596146A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112721935A (en) * | 2021-01-19 | 2021-04-30 | 西人马帝言(北京)科技有限公司 | Vehicle control model training method, vehicle control method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6336108B1 (en) * | 1997-12-04 | 2002-01-01 | Microsoft Corporation | Speech recognition with mixtures of bayesian networks |
| CN106124175A (en) * | 2016-06-14 | 2016-11-16 | 电子科技大学 | A kind of compressor valve method for diagnosing faults based on Bayesian network |
-
2018
- 2018-05-09 CN CN201810438507.3A patent/CN108596146A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6336108B1 (en) * | 1997-12-04 | 2002-01-01 | Microsoft Corporation | Speech recognition with mixtures of bayesian networks |
| CN106124175A (en) * | 2016-06-14 | 2016-11-16 | 电子科技大学 | A kind of compressor valve method for diagnosing faults based on Bayesian network |
Non-Patent Citations (4)
| Title |
|---|
| HARALD. STECK: "Constraint-Based Structural Learning in Bayesian Networks using Finite Data Sets", 《INSTITUT FÜR INFORMATIK DER TECHNISCHEN UNIVERSITAT MÜNCHEN》 * |
| 刘洋: "贝叶斯网络在民航飞行安全评价中的应用", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
| 李凤等: "面向土地利用分类的多源遥感数据混合贝叶斯网络分类器", 《国土资源遥感》 * |
| 陶旭东等: "基于机器视觉的汽车仪表盘功能测试方法的研究", 《科技创新与应用》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112721935A (en) * | 2021-01-19 | 2021-04-30 | 西人马帝言(北京)科技有限公司 | Vehicle control model training method, vehicle control method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Husain et al. | Vehicle detection in intelligent transport system under a hazy environment: a survey | |
| Masmoudi et al. | Object detection learning techniques for autonomous vehicle applications | |
| CN105335702B (en) | A kind of bayonet model recognizing method based on statistical learning | |
| Huo et al. | Vehicle type classification and attribute prediction using multi-task RCNN | |
| Kuang et al. | Feature selection based on tensor decomposition and object proposal for night-time multiclass vehicle detection | |
| Li et al. | A novel spatial-temporal graph for skeleton-based driver action recognition | |
| CN105956568A (en) | Abnormal behavior detecting and early warning method based on monitored object identification | |
| Park et al. | Real-time signal light detection | |
| Quiros et al. | Machine vision of traffic state estimation using fuzzy logic | |
| Garcia et al. | Identification of ghost moving detections in automotive scenarios with deep learning | |
| Al Mamun et al. | Lane marking detection using simple encode decode deep learning technique: SegNet | |
| CN109993058A (en) | The recognition methods of road signs based on multi-tag classification | |
| CN115984537A (en) | Image processing method and device and related equipment | |
| Naik et al. | Implementation of YOLOv4 algorithm for multiple object detection in image and video dataset using deep learning and artificial intelligence for urban traffic video surveillance application | |
| CN113450573A (en) | Traffic monitoring method and traffic monitoring system based on unmanned aerial vehicle image recognition | |
| Quinn et al. | Traffic flow monitoring in crowded cities | |
| CN108596146A (en) | Road multi-target classification method | |
| WO2023071397A1 (en) | Detection method and system for dangerous driving behavior | |
| KR102035184B1 (en) | Method and Apparatus for detecting abnormal behavior | |
| Delavarian et al. | Multi‐camera multiple vehicle tracking in urban intersections based on multilayer graphs | |
| Raj et al. | Real-time vehicle and pedestrian detection through SSD in Indian traffic conditions | |
| Arce et al. | Efficient lane detection based on artificial neural networks | |
| Sreeja et al. | Traffic Sign Detection using Transfer learning and a Comparison Between Different Techniques | |
| Dhanakshirur et al. | A framework for lane prediction on unstructured roads | |
| Khan et al. | Multiple moving vehicle speed estimation using Blob analysis |
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: 20180928 |